Module thunderfish.multivariateexplorer
Simple GUI for viewing and exploring multivariate data.
class MultiVariateExplorer: simple matplotlib-based GUI for viewing and exploring multivariate data.categorize(): convert categorial string data into integer categories.
Expand source code
"""Simple GUI for viewing and exploring multivariate data.
- `class MultiVariateExplorer`: simple matplotlib-based GUI for viewing and exploring multivariate data.
- `categorize()`: convert categorial string data into integer categories.
"""
import numpy as np
from sklearn import decomposition
from sklearn import preprocessing
import matplotlib.pyplot as plt
import matplotlib.patches as patches
import matplotlib.widgets as widgets
import argparse
from .version import __version__, __year__
from .tabledata import TableData
class MultivariateExplorer(object):
"""Simple matplotlib-based GUI for viewing and exploring multivariate data.
Shown are scatter plots of all pairs of variables or PCA axis.
Points in the scatter plots are colored according to the values of one of the variables.
Data points can be selected and optionally corresponding waveforms are shown.
First you initialize the explorer with the data. Then you optionally
specify how to colorize the data and provide waveform data
associated with the data. Finally you show the figure:
```
expl = MultivariateExplorer(data)
expl.set_colors(2)
expl.set_wave_data(waveforms, 'Time [s]', 'Sine')
expl.show()
```
The `compute_pca() function computes a principal component analysis (PCA)
on the input data, and `save_pca()` writes the principal components to a file.
Customize the appearance and information provided by subclassing
MultivariateExplorer and reimplementing the functions
- fix_scatter_plot()
- fix_waveform_plot()
- list_selection()
- analyze_selection()
See the documentation of these functions for details.
"""
mouse_actions = (('left click', 'select data points'),
('left and drag', 'rectangular selection and zoom of data points'),
('shift + left click/drag', 'add data points to selection'),
('ctrl + left click/drag', 'remove data points from selection'))
key_actions = (('l', 'list selected EOD waveforms on console'),
('p,P', 'toggle between data columns, PC, and scaled PC axis'),
('<, pageup', 'decrease number of displayed data columns/PC axis'),
('>, pagedown', 'increase number of displayed data columns/PC axis'),
('w', 'toggle maximized waveform plot'),
('o, z', 'toggle zoom mode on or off'),
('backspace', 'zoom back'),
('ctrl + a', 'select all'),
('+, -', 'increase, decrease pick radius'),
('0', 'reset pick radius'),
('n, N', 'decrease, increase number of bins of histograms'),
('h', 'toggle between scatter plot and 2D histogram'),
('c, C', 'cycle color map trough data columns'),
('left, right, up, down', 'show and move magnified scatter plot'),
('escape', 'close magnified scatter plot'))
def __init__(self, data, labels=None, title=None):
"""Initialize explorer with scatter-plot data.
Parameters
----------
data: TableData, 2D array, or list of 1D arrays
The data to be explored. Each column is a variable.
For the 2D array the columns are the second dimension,
for a list of 1D arrays, the list goes over columns,
i.e. each 1D array is one column.
labels: list of string
If data is not a TableData, then this provides labels
for the data columns.
title: string
Title for the window.
"""
# data and labels:
if isinstance(data, TableData):
self.categories = []
for c, col in enumerate(data):
if not isinstance(col[0], (int, float,
np.integer, np.floating)):
# categorial data:
cats, data[:,c] = categorize(col)
self.categories.append(cats)
else:
self.categories.append(None)
self.raw_data = data.array()
if labels is None:
self.raw_labels = []
for c in range(len(data)):
if len(data.unit(c)) > 0 and not data.unit(c) in ['-', '1']:
self.raw_labels.append('%s [%s]' % (data.label(c), data.unit(c)))
else:
self.raw_labels.append(data.label(c))
else:
self.raw_labels = labels
else:
if isinstance(data, np.ndarray):
self.raw_data = data
else:
self.categories = []
for c, col in enumerate(data):
if not isinstance(col[0], (int, float,
np.integer, np.floating)):
# categorial data:
cats, data[c] = categorize(col)
self.categories.append(cats)
else:
self.categories.append(None)
self.raw_data = np.asarray(data).T
self.raw_labels = labels
self.title = title if title is not None else 'MultivariateExplorer'
# no pca data yet:
self.all_data = [self.raw_data, None, None]
self.all_labels = [self.raw_labels, None, None]
self.all_maxcols = [self.raw_data.shape[1], None, None]
self.all_titles = ['data', 'PCA', 'scaled PCA']
# pca:
self.pca_tables = [None, None]
self._pca_header(data, labels)
# start showing raw data:
self.show_mode = 0
self.data = self.all_data[self.show_mode]
self.labels = self.all_labels[self.show_mode]
self.maxcols = self.all_maxcols[self.show_mode]
if self.maxcols > 6:
self.maxcols = 6
# waveform data:
self.wave_data = []
self.wave_nested = False
self.wave_has_xticks = []
self.wave_xlabels = []
self.wave_ylabels = []
self.wave_title = False
# colors:
self.color_map = plt.get_cmap('jet')
self.extra_colors = None
self.extra_color_label = None
self.extra_categories = None
self.color_values = None
self.color_set_index = 0
self.color_index = 0
self.color_label = None
self.color_set_index = 0
self.color_index = 0
self.data_colors = None
self.color_vmin = None
self.color_vmax = None
self.color_ticks = None
self.cbax = None
# figure variables:
self.plt_params = {}
for k in ['toolbar', 'keymap.quit', 'keymap.back', 'keymap.forward',
'keymap.zoom', 'keymap.pan', 'keymap.xscale', 'keymap.yscale']:
self.plt_params[k] = plt.rcParams[k]
if k != 'toolbar':
plt.rcParams[k] = ''
self.xborder = 70.0 # pixel for ylabels
self.yborder = 50.0 # pixel for xlabels
self.spacing = 10.0 # pixel between plots
self.pick_radius = 4.0
# histogram plots:
self.hist_ax = []
self.hist_indices = []
self.hist_selector = []
self.hist_nbins = 30
# scatter plots:
self.scatter_ax = []
self.scatter_indices = []
self.scatter_artists = []
self.scatter_selector = []
self.scatter = True
self.mark_data = []
self.select_zooms = False
self.zoom_stack = []
# magnified scatter plot:
self.magnified_on = False
self.magnified_backdrop = None
self.magnified_size = np.array([0.5, 0.5])
# waveform plots:
self.wave_ax = []
def set_wave_data(self, data, xlabels='', ylabels=[], title=False):
"""Add waveform data to explorer.
Parameters
----------
data: list of (list of) 2D arrays
Waveform data associated with each row of the data.
Elements of the outer list correspond to the rows of the data.
The inner 2D arrays contain a common x-axes (first column)
and one or more corresponding y-values (second and optional higher columns).
Each column for y-values is plotted in its own axes on top of each other,
from top to bottom.
The optional inner list of 2D arrays contains several 2D arrays as ascribed above
each with its own common x-axes.
xlabel: string or list of strings
The xlabels for the waveform plots. If only a string is given, then
there will be a common xaxis for all the plots, and only the lowest
one gets a labeled xaxis. If a list of strings is given, each waveform
plot gets its own labeled x-axis.
ylabels: list of strings
The ylabels for each of the waveform plots.
title: bool or string
If True or a string, povide space on top of the waveform plots for a title.
If string, set this as the title for the waveform plots.
"""
self.wave_data = []
if data is not None and len(data) > 0:
self.wave_data = data
self.wave_has_xticks = []
self.wave_nested = isinstance(data[0], (list, tuple))
if self.wave_nested:
for data in self.wave_data[0]:
for k in range(data.shape[1]-2):
self.wave_has_xticks.append(False)
self.wave_has_xticks.append(True)
else:
for k in range(self.wave_data[0].shape[1]-2):
self.wave_has_xticks.append(False)
self.wave_has_xticks.append(True)
if isinstance(xlabels, (list, tuple)):
self.wave_xlabels = xlabels
else:
self.wave_xlabels = [xlabels]
self.wave_ylabels = ylabels
self.wave_title = title
self.wave_ax = []
def set_colors(self, colors=0, color_label=None, color_map=None):
"""Set data column used to color scatter plots.
Parameters
----------
colors: int or 1D array
Index to colum in data to be used for coloring scatter plots.
-2 for coloring row index of data.
Or data array used to color scalar plots.
color_label: string
If colors is an array, this is a label describing the data.
It is used to label the color bar.
color_map: string or None
Name of a matplotlib color map.
If None 'jet' is used.
"""
if isinstance(colors, (np.integer, int)):
if colors < 0:
self.color_set_index = -1
self.color_index = 0
else:
self.color_set_index = 0
self.color_index = colors
else:
if not isinstance(colors[0], (int, float,
np.integer, np.floating)):
# categorial data:
self.extra_categories, self.extra_colors = categorize(colors)
else:
self.extra_colors = colors
self.extra_color_label = color_label
self.color_set_index = -1
self.color_index = 1
self.color_map = plt.get_cmap(color_map if color_map else 'jet')
def show(self):
"""Show interactive scatter plots for exploration.
"""
plt.ioff()
plt.rcParams['toolbar'] = 'None'
plt.rcParams['keymap.quit'] = 'ctrl+w, alt+q, q'
self.fig = plt.figure(facecolor='white')
self.fig.canvas.manager.set_window_title(self.title + ': ' + self.all_titles[self.show_mode])
self.fig.canvas.mpl_connect('key_press_event', self._on_key)
self.fig.canvas.mpl_connect('resize_event', self._on_resize)
self.fig.canvas.mpl_connect('pick_event', self._on_pick)
if self.color_map is None:
self.color_map = plt.get_cmap('jet')
self._set_color_column()
self._init_hist_plots()
self._init_scatter_plots()
self.wave_ax = []
if self.wave_data is not None and len(self.wave_data) > 0:
axx = None
xi = 0
for k, has_xticks in enumerate(self.wave_has_xticks):
ax = self.fig.add_subplot(1, len(self.wave_has_xticks), 1+k, sharex=axx)
self.wave_ax.append(ax)
if has_xticks:
if xi >= len(self.wave_xlabels):
self.wave_xlabels.append('')
ax.set_xlabel(self.wave_xlabels[xi])
xi += 1
axx = None
else:
plt.setp(ax.get_xticklabels(), visible=False)
if axx is None:
axx = ax
for ax, ylabel in zip(self.wave_ax, self.wave_ylabels):
ax.set_ylabel(ylabel)
if not isinstance(self.wave_title, bool) and self.wave_title:
self.wave_ax[0].set_title(self.wave_title)
self.fix_waveform_plot(self.wave_ax, self.mark_data)
self._plot_magnified_scatter()
plt.show()
def _pca_header(self, data, labels):
"""Set up header for the table of principal components. """
if isinstance(data, TableData):
header = data.table_header()
for c in reversed(range(data.columns())):
if not np.all(np.isfinite(data[:,c])):
header.remove(c)
else:
lbs = []
for l, d in zip(labels, data):
if not np.all(np.isfinite(d)):
continue
if '[' in l:
lbs.append(l.split('[')[0].strip())
elif '/' in l:
lbs.append(l.split('/')[0].strip())
else:
lbs.append(l)
header = TableData(header=lbs)
header.set_formats('%.3f')
header.insert(0, ['PC'] + ['-']*header.nsecs, '', '%d')
header.insert(1, 'variance', '%', '%.3f')
for k in range(len(self.pca_tables)):
self.pca_tables[k] = TableData(header)
def compute_pca(self, scale=False, write=False):
"""Compute PCA based on the data.
Parameters
----------
scale: boolean
If True standardize data before computing PCA, i.e. remove mean
of each variabel and divide by its standard deviation.
write: boolean
If True write PCA components to standard out.
"""
# select columns without NANs:
idxs = [i for i in range(self.raw_data.shape[1]) if np.all(np.isfinite(self.raw_data[:,i]))]
data = self.raw_data[:,idxs]
# pca:
pca = decomposition.PCA()
if scale:
scaler = preprocessing.StandardScaler()
scaler.fit(data)
pca.fit(scaler.transform(data))
pca_label = 'sPC'
else:
pca.fit(data)
pca_label = 'PC'
for k in range(len(pca.components_)):
if np.abs(np.min(pca.components_[k])) > np.max(pca.components_[k]):
pca.components_[k] *= -1.0
pca_data = pca.transform(data)
pca_labels = [('%s%d (%.1f%%)' if v > 0.01 else '%s%d (%.2f%%)') % (pca_label, k+1, 100.0*v)
for k, v in enumerate(pca.explained_variance_ratio_)]
if np.min(pca.explained_variance_ratio_) >= 0.01:
pca_maxcols = pca_data.shape[1]
else:
pca_maxcols = np.argmax(pca.explained_variance_ratio_ < 0.01)
if pca_maxcols < 2:
pca_maxcols = 2
if pca_maxcols > 6:
pca_maxcols = 6
# table with PCA feature weights:
pca_table = self.pca_tables[1] if scale else self.pca_tables[0]
pca_table.clear_data()
pca_table.set_section(pca_label, 0, pca_table.nsecs)
for k, comp in enumerate(pca.components_):
pca_table.append_data(k+1, 0)
pca_table.append_data(100.0*pca.explained_variance_ratio_[k])
pca_table.append_data(comp)
if write:
pca_table.write(table_format='out', unit_style='none')
# submit data:
if scale:
self.all_data[2] = pca_data
self.all_labels[2] = pca_labels
self.all_maxcols[2] = pca_maxcols
else:
self.all_data[1] = pca_data
self.all_labels[1] = pca_labels
self.all_maxcols[1] = pca_maxcols
def save_pca(self, file_name, scale, **kwargs):
"""Write PCA data to file.
Parameters
----------
file_name: string
Name of ouput file.
scale: boolean
If True write PCA components of standardized PCA.
kwargs: dict
Additional parameter for TableData.write()
"""
if scale:
pca_file = file_name + '-pcacor'
pca_table = self.pca_tables[1]
else:
pca_file = file_name + '-pcacov'
pca_table = self.pca_tables[0]
if 'unit_style' in kwargs:
del kwargs['unit_style']
if 'table_format' in kwargs:
pca_table.write(pca_file, unit_style='none', **kwargs)
else:
pca_file += '.dat'
pca_table.write(pca_file, unit_style='none')
def _set_color_column(self):
"""Initialize variables used for colorization of scatter points."""
if self.color_set_index == -1:
if self.color_index == 0:
self.color_values = np.arange(self.data.shape[0], dtype=float)
self.color_label = 'row'
elif self.color_index == 1:
self.color_values = self.extra_colors
self.color_label = self.extra_color_label
else:
self.color_values = self.all_data[self.color_set_index][:,self.color_index]
self.color_label = self.all_labels[self.color_set_index][self.color_index]
self.color_vmin, self.color_vmax, self.color_ticks = \
self.fix_scatter_plot(self.cbax, self.color_values, self.color_label, 'c')
if self.color_ticks is None:
if self.color_set_index == 0 and \
self.categories[self.color_index] is not None:
self.color_ticks = np.arange(len(self.categories[self.color_index]))
elif self.color_set_index == -1 and \
self.color_index == 1 and \
self.extra_categories is not None:
self.color_ticks = np.arange(len(self.extra_categories))
self.data_colors = self.color_map((self.color_values - self.color_vmin)/(self.color_vmax - self.color_vmin))
def _plot_hist(self, ax, magnifiedax, keep_lims):
"""Plot and label a histogram."""
ax_xlim = ax.get_xlim()
ax_ylim = ax.get_ylim()
try:
idx = self.hist_ax.index(ax)
c = self.hist_indices[idx]
in_hist = True
except ValueError:
idx = self.scatter_ax.index(ax)
c = self.scatter_indices[-1][0]
in_hist = False
ax.clear()
ax.relim()
ax.autoscale(True)
x = self.data[:,c]
ax.hist(x[np.isfinite(x)], self.hist_nbins)
ax.set_xlabel(self.labels[c])
if self.categories[c] is not None:
ax.set_xticks(np.arange(len(self.categories[c])))
ax.set_xticklabels(self.categories[c])
self.fix_scatter_plot(ax, self.data[:,c], self.labels[c], 'x')
if magnifiedax:
ax.set_ylabel('count')
cax = self.hist_ax[self.scatter_indices[-1][0]]
ax.set_xlim(cax.get_xlim())
else:
if c == 0:
ax.set_ylabel('count')
else:
plt.setp(ax.get_yticklabels(), visible=False)
if keep_lims:
ax.set_xlim(*ax_xlim)
ax.set_ylim(*ax_ylim)
try:
selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1,
props=dict(facecolor='gray', edgecolor='gray', alpha=0.2, fill=True),
state_modifier_keys=dict(move='', clear='', square='', center=''))
except TypeError:
selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1)
if in_hist:
self.hist_selector[idx] = selector
else:
self.scatter_selector[idx] = selector
self.scatter_artists[idx] = None
if magnifiedax:
bbox = ax.get_tightbbox(self.fig.canvas.get_renderer())
if bbox is not None:
self.magnified_backdrop = patches.Rectangle((bbox.x0, bbox.y0),
bbox.width, bbox.height,
transform=None, clip_on=False,
facecolor='white', edgecolor='none', zorder=-5)
ax.add_patch(self.magnified_backdrop)
def _init_hist_plots(self):
"""Initial plots of the histograms."""
n = self.data.shape[1]
yax = None
self.hist_ax = []
for r in range(n):
ax = self.fig.add_subplot(n, n, (n-1)*n+r+1, sharey=yax)
self.hist_ax.append(ax)
self.hist_indices.append(r)
self.hist_selector.append(None)
self._plot_hist(ax, False, False)
yax = ax
def _plot_scatter(self, ax, magnifiedax, keep_lims, cax=None):
"""Plot a scatter plot."""
ax_xlim = ax.get_xlim()
ax_ylim = ax.get_ylim()
idx = self.scatter_ax.index(ax)
c, r = self.scatter_indices[idx]
sel = np.isfinite(self.data[:,c]) & np.isfinite(self.data[:,r])
if self.scatter:
ax.clear()
ax.relim()
ax.autoscale(True)
a = ax.scatter(self.data[sel,c], self.data[sel,r], c=self.color_values[sel],
cmap=self.color_map, vmin=self.color_vmin, vmax=self.color_vmax,
s=50, edgecolors='none', zorder=10)
if cax is not None:
self.fig.colorbar(a, cax=cax, ticks=self.color_ticks)
cax.set_ylabel(self.color_label)
self.color_vmin, self.color_vmax, self.color_ticks = \
self.fix_scatter_plot(self.cbax, self.color_values, self.color_label, 'c')
if self.color_ticks is None:
if self.color_set_index == 0 and \
self.categories[self.color_index] is not None:
cax.set_yticklabels(self.categories[self.color_index])
elif self.color_set_index == -1 and \
self.color_index == 1 and \
self.extra_categories is not None:
cax.set_yticklabels(self.extra_categories)
else:
ax.autoscale(True)
self.fix_scatter_plot(ax, self.data[sel,c], self.labels[c], 'x')
self.fix_scatter_plot(ax, self.data[sel,r], self.labels[r], 'y')
axrange = [ax.get_xlim(), ax.get_ylim()]
ax.clear()
ax.hist2d(self.data[sel,c], self.data[sel,r], self.hist_nbins, range=axrange,
cmap=plt.get_cmap('Greys'))
md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and
np.isfinite(self.data[m,r])]
a = ax.scatter(self.data[md,c], self.data[md,r], c=self.data_colors[md],
s=80, zorder=11)
self.scatter_artists[idx] = a
if self.categories[c] is not None:
ax.set_xticks(np.arange(len(self.categories[c])))
ax.set_xticklabels(self.categories[c])
if self.categories[r] is not None:
ax.set_yticks(np.arange(len(self.categories[r])))
ax.set_yticklabels(self.categories[r])
if magnifiedax:
ax.set_xlabel(self.labels[c])
ax.set_ylabel(self.labels[r])
cax = self.scatter_ax[self.scatter_indices[:-1].index(self.scatter_indices[-1])]
ax.set_xlim(cax.get_xlim())
ax.set_ylim(cax.get_ylim())
else:
if c == 0:
ax.set_ylabel(self.labels[r])
self.fix_scatter_plot(ax, self.data[sel,c], self.labels[c], 'x')
self.fix_scatter_plot(ax, self.data[sel,r], self.labels[r], 'y')
if not magnifiedax:
plt.setp(ax.get_xticklabels(), visible=False)
if c > 0:
plt.setp(ax.get_yticklabels(), visible=False)
if keep_lims:
ax.set_xlim(*ax_xlim)
ax.set_ylim(*ax_ylim)
if magnifiedax:
bbox = ax.get_tightbbox(self.fig.canvas.get_renderer())
if bbox is not None:
self.magnified_backdrop = patches.Rectangle((bbox.x0, bbox.y0),
bbox.width, bbox.height,
transform=None, clip_on=False,
facecolor='white', edgecolor='none', zorder=-5)
ax.add_patch(self.magnified_backdrop)
try:
selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1,
props=dict(facecolor='gray', edgecolor='gray', alpha=0.2, fill=True),
state_modifier_keys=dict(move='', clear='', square='', center=''))
except TypeError:
selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1)
self.scatter_selector[idx] = selector
def _init_scatter_plots(self):
"""Initial plots of scatter plots."""
self.cbax = self.fig.add_axes([0.5, 0.5, 0.1, 0.5])
cbax = self.cbax
n = self.data.shape[1]
for r in range(1, n):
yax = None
for c in range(r):
ax = self.fig.add_subplot(n, n, (r-1)*n+c+1, sharex=self.hist_ax[c], sharey=yax)
self.scatter_ax.append(ax)
self.scatter_indices.append([c, r])
self.scatter_artists.append(None)
self.scatter_selector.append(None)
self._plot_scatter(ax, False, False, cbax)
yax = ax
cbax = None
def _plot_magnified_scatter(self):
"""Initial plot of the magnified scatter plot."""
ax = self.fig.add_axes([0.5, 0.9, 0.05, 0.05])
ax.set_visible(False)
self.magnified_on = False
c = 0
r = 1
sel = np.isfinite(self.data[:,c]) & np.isfinite(self.data[:,r])
ax.scatter(self.data[sel,c], self.data[sel,r], c=self.data_colors[sel],
s=50, edgecolors='none')
md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and
np.isfinite(self.data[m,r])]
a = ax.scatter(self.data[md,c], self.data[md,r],
c=self.data_colors[md], s=80)
ax.set_xlabel(self.labels[c])
ax.set_ylabel(self.labels[r])
self.fix_scatter_plot(ax, self.data[sel,c], self.labels[c], 'x')
self.fix_scatter_plot(ax, self.data[sel,r], self.labels[r], 'y')
self.scatter_ax.append(ax)
self.scatter_indices.append([c, r])
self.scatter_artists.append(a)
self.scatter_selector.append(None)
def fix_scatter_plot(self, ax, data, label, axis):
"""Customize an axes of a scatter plot.
This function is called after a scatter plot has been plotted.
Once for the x axes, once for the y axis and once for the color bar.
Reimplement this function to set appropriate limits and ticks.
Return values are only used for the color bar (`axis='c'`).
Otherwise they are ignored.
For example, ticks for phase variables can be nicely labeled
using the unicode character for pi:
```
if 'phase' in label:
if axis == 'y':
ax.set_ylim(0.0, 2.0*np.pi)
ax.set_yticks(np.arange(0.0, 2.5*np.pi, 0.5*np.pi))
ax.set_yticklabels(['0', u'\u03c0/2', u'\u03c0', u'3\u03c0/2', u'2\u03c0'])
```
Parameters
----------
ax: matplotlib axes
Axes of the scatter plot or color bar to be worked on.
data: 1D array
Data array of the axes.
label: string
Label coresponding to the data array.
axis: str
'x', 'y': set properties of x or y axes of ax.
'c': set properies of color bar axes (note that ax can be None!)
and return vmin, vmax, and ticks.
Returns
-------
min: float
minimum value of color bar axis
max: float
maximum value of color bar axis
ticks: list of float
position of ticks for color bar axis
"""
return np.nanmin(data), np.nanmax(data), None
def fix_waveform_plot(self, axs, indices):
"""Customize waveform plots.
This function is called once after new data have been plotted
into the waveform plots. Reimplement this function to customize
these plots. In particular to set axis limits and labels, plot
title, etc.
You may even open a new figure (with non-blocking `show()`).
The following member variables might be usefull:
- `self.wave_data`: the full list of waveform data.
- `self.wave_nested`: True if the elements of `self.wave_data` are lists of 2D arrays. Otherwise the elements are 2D arrays. The first column of a 2D array contains the x-values, further columns y-values.
- `self.wave_has_xticks`: List of booleans for each axis. True if the axis has its own xticks.
- `self.wave_xlabels`: List of xlabels (only for the axis where the corresponding entry in `self.wave_has_xticks` is True).
- `self.wave_ylabels`: for each axis its ylabel
For example, you can set the linewidth of all plotted waveforms via:
```
for ax in axs:
for l in ax.lines:
l.set_linewidth(3.0)
```
or enable markers to be plotted:
```
for ax, yl in zip(axs, self.wave_ylabels):
if 'Power' in yl:
for l in ax.lines:
l.set_marker('.')
l.set_markersize(15.0)
l.set_markeredgewidth(0.5)
l.set_markeredgecolor('k')
l.set_markerfacecolor(l.get_color())
```
Usefull is to reduce the maximum number of y-ticks:
```
axs[0].yaxis.get_major_locator().set_params(nbins=7)
```
or
```
import matplotlib.ticker as ticker
axs[0].yaxis.set_major_locator(ticker.MaxNLocator(nbins=4))
```
Parameters
----------
axs: list of matplotlib axes
Axis of the waveform plots to be worked on.
indices: list of int
Indices of the waveforms that have been selected and plotted.
"""
pass
def list_selection(self, indices):
"""List information about the current selection of data points.
This function is called when 'l' is pressed. Reimplement this
function, for example, to print some meaningfull information
about the current selection of data points on console. You nay
do, however, whatever you want in this function.
Parameters
----------
indices: list of int
Indices of the data points that have been selected.
"""
for i in indices:
print(i)
def analyze_selection(self, index):
"""Provide further information about a single selected data point.
This function is called when a single data item was double
clicked. Reimplement this function to provide some further
details on this data point. This can be an additional figure
window. In this case show it non-blocking:
`plt.show(block=False)`
Parameters
----------
index: int
The index of the selected data point.
"""
pass
def _set_magnified_pos(self, width, height):
"""Set position of magnified plot."""
if self.magnified_on:
xoffs = self.xborder/width
yoffs = self.yborder/height
if self.scatter_indices[-1][1] < self.data.shape[1]:
idx = self.scatter_indices[:-1].index(self.scatter_indices[-1])
pos = self.scatter_ax[idx].get_position().get_points()
else:
pos = self.hist_ax[self.scatter_indices[-1][0]].get_position().get_points()
pos[0] = np.mean(pos, 0) - 0.5*self.magnified_size
if pos[0][0] < xoffs: pos[0][0] = xoffs
if pos[0][1] < yoffs: pos[0][1] = yoffs
pos[1] = pos[0] + self.magnified_size
if pos[1][0] > 1.0-self.spacing/width: pos[1][0] = 1.0-self.spacing/width
if pos[1][1] > 1.0-self.spacing/height: pos[1][1] = 1.0-self.spacing/height
pos[0] = pos[1] - self.magnified_size
self.scatter_ax[-1].set_position([pos[0][0], pos[0][1],
self.magnified_size[0], self.magnified_size[1]])
self.scatter_ax[-1].set_visible(True)
else:
self.scatter_ax[-1].set_position([0.5, 0.9, 0.05, 0.05])
self.scatter_ax[-1].set_visible(False)
def _make_selection(self, ax, key, x0, x1, y0, y1):
"""Select points from a scatter or histogram plot."""
if not key in ['shift', 'control']:
self.mark_data = []
try:
axi = self.scatter_ax.index(ax)
# from scatter plots:
c, r = self.scatter_indices[axi]
if r < self.data.shape[1]:
# from scatter:
for ind, (x, y) in enumerate(zip(self.data[:,c], self.data[:,r])):
if np.isfinite(x) and np.isfinite(y) and \
x >= x0 and x <= x1 and y >= y0 and y <= y1:
if ind in self.mark_data:
if key == 'control':
self.mark_data.remove(ind)
elif key != 'control':
self.mark_data.append(ind)
else:
# from histogram:
for ind, x in enumerate(self.data[:,c]):
if np.isfinite(x) and x >= x0 and x <= x1:
if ind in self.mark_data:
if key == 'control':
self.mark_data.remove(ind)
elif key != 'control':
self.mark_data.append(ind)
except ValueError:
try:
r = self.hist_ax.index(ax)
# from histogram:
for ind, x in enumerate(self.data[:,r]):
if np.isfinite(x) and x >= x0 and x <= x1:
if ind in self.mark_data:
if key == 'control':
self.mark_data.remove(ind)
elif key != 'control':
self.mark_data.append(ind)
except ValueError:
return
def _update_selection(self):
"""Highlight select points in the scatter plots and plot corresponding waveforms."""
# update scatter plots:
for artist, (c, r) in zip(self.scatter_artists, self.scatter_indices):
if artist is not None:
md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and
np.isfinite(self.data[m,r])]
artist.set_offsets(list(zip(self.data[md,c], self.data[md,r])))
artist.set_facecolors(self.data_colors[md])
# waveform plots:
if len(self.wave_ax) > 0:
axdi = 0
axti = 1
for xi, ax in enumerate(self.wave_ax):
ax.clear()
if len(self.mark_data) > 0:
for idx in self.mark_data:
if self.wave_nested:
data = self.wave_data[idx][axdi]
else:
data = self.wave_data[idx]
if data is not None:
ax.plot(data[:,0], data[:,axti], c=self.data_colors[idx],
picker=self.pick_radius)
axti += 1
if self.wave_has_xticks[xi]:
ax.set_xlabel(self.wave_xlabels[axdi])
axti = 1
axdi += 1
else:
plt.setp(ax.get_xticklabels(), visible=False)
for ax, ylabel in zip(self.wave_ax, self.wave_ylabels):
ax.set_ylabel(ylabel)
if not isinstance(self.wave_title, bool) and self.wave_title:
self.wave_ax[0].set_title(self.wave_title)
self.fix_waveform_plot(self.wave_ax, self.mark_data)
self.fig.canvas.draw()
def _on_key(self, event):
"""Handle key events."""
#print('pressed', event.key)
plot_zoom = True
if event.key in ['left', 'right', 'up', 'down']:
if self.magnified_on:
if event.key == 'left':
if self.scatter_indices[-1][0] > 0:
self.scatter_indices[-1][0] -= 1
else:
plot_zoom = False
elif event.key == 'right':
if self.scatter_indices[-1][0] < self.scatter_indices[-1][1]-1 and \
self.scatter_indices[-1][0] < self.maxcols-1:
self.scatter_indices[-1][0] += 1
else:
plot_zoom = False
elif event.key == 'up':
if self.scatter_indices[-1][1] > 1:
if self.scatter_indices[-1][1] >= self.data.shape[1]:
self.scatter_indices[-1][1] = self.maxcols-1
else:
self.scatter_indices[-1][1] -= 1
if self.scatter_indices[-1][0] >= self.scatter_indices[-1][1]:
self.scatter_indices[-1][0] = self.scatter_indices[-1][1]-1
else:
plot_zoom = False
elif event.key == 'down':
if self.scatter_indices[-1][1] < self.maxcols:
self.scatter_indices[-1][1] += 1
if self.scatter_indices[-1][1] >= self.maxcols:
self.scatter_indices[-1][1] = self.data.shape[1]
else:
plot_zoom = False
else:
plot_zoom = False
if event.key == 'escape':
self.scatter_ax[-1].set_position([0.5, 0.9, 0.05, 0.05])
self.magnified_on = False
self.scatter_ax[-1].set_visible(False)
self.fig.canvas.draw()
elif event.key in 'oz':
self.select_zooms = not self.select_zooms
elif event.key == 'backspace':
if len(self.zoom_stack) > 0:
ax, xmin, xmax, ymin, ymax = self.zoom_stack.pop()
ax.set_xlim(xmin, xmax)
ax.set_ylim(ymin, ymax)
if ax in self.scatter_ax[:-1]:
axidx = self.scatter_ax[:-1].index(ax)
if self.scatter_indices[axidx][0] == self.scatter_indices[-1][0]:
self.scatter_ax[-1].set_xlim(xmin, xmax)
if self.scatter_indices[axidx][1] == self.scatter_indices[-1][1]:
self.scatter_ax[-1].set_ylim(ymin, ymax)
elif ax in self.hist_ax:
if self.scatter_indices[-1][1] == self.data.shape[1] and \
self.scatter_indices[-1][0] == self.hist_ax.index(ax):
self.scatter_ax[-1].set_xlim(xmin, xmax)
self.scatter_ax[-1].set_ylim(ymin, ymax)
self.fig.canvas.draw()
elif event.key in '+=':
self.pick_radius *= 1.5
elif event.key in '-':
if self.pick_radius > 5.0:
self.pick_radius /= 1.5
elif event.key in '0':
self.pick_radius = 4.0
elif event.key in ['pageup', 'pagedown', '<', '>']:
if event.key in ['pageup', '<'] and self.maxcols > 2:
self.maxcols -= 1
elif event.key in ['pagedown', '>'] and self.maxcols < self.raw_data.shape[1]:
self.maxcols += 1
self._update_layout()
elif event.key == 'w':
if self.maxcols > 0:
self.all_maxcols[self.show_mode] = self.maxcols
self.maxcols = 0
else:
self.maxcols = self.all_maxcols[self.show_mode]
self._set_layout(self.fig.get_window_extent().width,
self.fig.get_window_extent().height)
self.fig.canvas.draw()
elif event.key == 'ctrl+a':
self.mark_data = range(len(self.data))
self._update_selection()
elif event.key in 'cC':
if event.key in 'c':
first = True
while first or not np.all(np.isfinite(self.color_values)):
self.color_index -= 1
if self.color_index < 0:
self.color_set_index -= 1
if self.color_set_index < -1:
self.color_set_index = len(self.all_data)-1
if self.color_set_index >= 0:
if self.all_data[self.color_set_index] is None:
self.compute_pca(self.color_set_index>1, True)
self.color_index = self.all_data[self.color_set_index].shape[1]-1
else:
self.color_index = 0 if self.extra_colors is None else 1
self._set_color_column()
first = False
else:
first = True
while first or not np.all(np.isfinite(self.color_values)):
self.color_index += 1
if (self.color_set_index >= 0 and \
self.color_index >= self.all_data[self.color_set_index].shape[1]) or \
(self.color_set_index < 0 and \
self.color_index >= (1 if self.extra_colors is None else 2)):
self.color_index = 0
self.color_set_index += 1
if self.color_set_index >= len(self.all_data):
self.color_set_index = -1
elif self.all_data[self.color_set_index] is None:
self.compute_pca(self.color_set_index>1, True)
self._set_color_column()
first = False
for ax in self.scatter_ax:
if len(ax.collections) > 0:
idx = self.scatter_ax.index(ax)
c, r = self.scatter_indices[idx]
ax.collections[0].set_facecolors(self.data_colors[np.isfinite(self.data[:,c]) & np.isfinite(self.data[:,r])])
for a, (c, r) in zip(self.scatter_artists, self.scatter_indices):
if a is not None:
md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and
np.isfinite(self.data[m,r])]
a.set_facecolors(self.data_colors[md])
for ax in self.wave_ax:
for l, c in zip(ax.lines, self.data_colors[self.mark_data]):
l.set_color(c)
l.set_markerfacecolor(c)
self._plot_scatter(self.scatter_ax[0], False, True, self.cbax)
self.fix_scatter_plot(self.cbax, self.color_values,
self.color_label, 'c')
self.fig.canvas.draw()
elif event.key in 'nN':
if event.key in 'N':
self.hist_nbins = (self.hist_nbins*3)//2
elif self.hist_nbins >= 15:
self.hist_nbins = (self.hist_nbins*2)//3
for ax in self.hist_ax:
self._plot_hist(ax, False, True)
if self.scatter_indices[-1][1] >= self.data.shape[1]:
self._plot_hist(self.scatter_ax[-1], True, True)
elif not self.scatter:
self._plot_scatter(self.scatter_ax[-1], True, True)
if not self.scatter:
for ax in self.scatter_ax[:-1]:
self._plot_scatter(ax, False, True)
self.fig.canvas.draw()
elif event.key in 'h':
self.scatter = not self.scatter
for ax in self.scatter_ax[:-1]:
self._plot_scatter(ax, False, True)
if self.scatter_indices[-1][1] < self.data.shape[1]:
self._plot_scatter(self.scatter_ax[-1], True, True)
self.fig.canvas.draw()
elif event.key in 'pP':
self.all_maxcols[self.show_mode] = self.maxcols
if event.key == 'p':
self.show_mode += 1
if self.show_mode >= len(self.all_data):
self.show_mode = 0
else:
self.show_mode -= 1
if self.show_mode < 0:
self.show_mode = len(self.all_data)-1
if self.show_mode == 1:
print('principal components')
elif self.show_mode == 2:
print('scaled principal components')
else:
print('data')
if self.all_data[self.show_mode] is None:
self.compute_pca(self.show_mode>1, True)
self.data = self.all_data[self.show_mode]
self.labels = self.all_labels[self.show_mode]
self.maxcols = self.all_maxcols[self.show_mode]
self.zoom_stack = []
self.fig.canvas.manager.set_window_title(self.title + ': ' + self.all_titles[self.show_mode])
for ax in self.hist_ax[:self.maxcols]:
self._plot_hist(ax, False, False)
for ax in self.scatter_ax[:self.maxcols]:
self._plot_scatter(ax, False, False)
self._update_layout()
elif event.key in 'l':
if len(self.mark_data) > 0:
print('')
print('selected:')
self.list_selection(self.mark_data)
if plot_zoom:
for k in reversed(range(len(self.zoom_stack))):
if self.zoom_stack[k][0] == self.scatter_ax[-1]:
del self.zoom_stack[k]
self.scatter_ax[-1].clear()
self.scatter_ax[-1].set_visible(True)
self.magnified_on = True
self._set_magnified_pos(self.fig.get_window_extent().width,
self.fig.get_window_extent().height)
if self.scatter_indices[-1][1] < self.data.shape[1]:
self._plot_scatter(self.scatter_ax[-1], True, False)
else:
self._plot_hist(self.scatter_ax[-1], True, False)
self.fig.canvas.draw()
def _on_select(self, eclick, erelease):
"""Handle selection events."""
if eclick.dblclick:
if len(self.mark_data) > 0:
self.analyze_selection(self.mark_data[-1])
return
x0 = min(eclick.xdata, erelease.xdata)
x1 = max(eclick.xdata, erelease.xdata)
y0 = min(eclick.ydata, erelease.ydata)
y1 = max(eclick.ydata, erelease.ydata)
ax = erelease.inaxes
if ax is None:
ax = eclick.inaxes
xmin, xmax = ax.get_xlim()
ymin, ymax = ax.get_ylim()
dx = 0.02*(xmax-xmin)
dy = 0.02*(ymax-ymin)
if x1 - x0 < dx and y1 - y0 < dy:
bbox = ax.get_window_extent().transformed(self.fig.dpi_scale_trans.inverted())
width, height = bbox.width, bbox.height
width *= self.fig.dpi
height *= self.fig.dpi
dx = self.pick_radius*(xmax-xmin)/width
dy = self.pick_radius*(ymax-ymin)/height
x0 = erelease.xdata - dx
x1 = erelease.xdata + dx
y0 = erelease.ydata - dy
y1 = erelease.ydata + dy
elif self.select_zooms:
self.zoom_stack.append((ax, xmin, xmax, ymin, ymax))
ax.set_xlim(x0, x1)
ax.set_ylim(y0, y1)
self._make_selection(ax, erelease.key, x0, x1, y0, y1)
self._update_selection()
def _on_pick(self, event):
"""Handle pick events on waveforms."""
for ax in self.wave_ax:
for k, l in enumerate(ax.lines):
if l is event.artist:
self.mark_data = [self.mark_data[k]]
self._update_selection()
if event.mouseevent.dblclick:
if len(self.mark_data) > 0:
self.analyze_selection(self.mark_data[-1])
def _set_layout(self, width, height):
"""Update positions and visibility of all plots."""
xoffs = self.xborder/width
yoffs = self.yborder/height
xs = self.spacing/width
ys = self.spacing/height
if self.maxcols > 0:
dx = (1.0-xoffs)/self.maxcols
dy = (1.0-yoffs)/self.maxcols
xw = dx - xs
yw = dy - ys
# histograms:
for c, ax in enumerate(self.hist_ax):
if c < self.maxcols:
ax.set_position([xoffs+c*dx, yoffs, xw, yw])
ax.set_visible(True)
else:
ax.set_visible(False)
ax.set_position([0.99, 0.01, 0.01, 0.01])
# scatter plots:
for ax, (c, r) in zip(self.scatter_ax[:-1], self.scatter_indices[:-1]):
if r < self.maxcols:
ax.set_position([xoffs+c*dx, yoffs+(self.maxcols-r)*dy, xw, yw])
ax.set_visible(True)
else:
ax.set_visible(False)
ax.set_position([0.99, 0.01, 0.01, 0.01])
# color bar:
if self.maxcols > 0:
self.cbax.set_position([xoffs+dx, yoffs+(self.maxcols-1)*dy, 0.3*xoffs, yw])
self.cbax.set_visible(True)
else:
self.cbax.set_visible(False)
self.cbax.set_position([0.99, 0.01, 0.01, 0.01])
# magnified plot:
if self.maxcols > 0:
self._set_magnified_pos(width, height)
if self.magnified_backdrop is not None:
bbox = self.scatter_ax[-1].get_tightbbox(self.fig.canvas.get_renderer())
if bbox is not None:
self.magnified_backdrop.set_bounds(bbox.x0, bbox.y0, bbox.width, bbox.height)
else:
self.scatter_ax[-1].set_position([0.5, 0.9, 0.05, 0.05])
self.scatter_ax[-1].set_visible(False)
# waveform plots:
if len(self.wave_ax) > 0:
if self.maxcols > 0:
x0 = xoffs+((self.maxcols+1)//2)*dx
y0 = ((self.maxcols+1)//2)*dy
if self.maxcols%2 == 0:
x0 += xoffs
y0 += yoffs - ys
else:
y0 += ys
else:
x0 = xoffs
y0 = 0.0
yp = 1.0
dy = 1.0-y0
dy -= np.sum(self.wave_has_xticks)*yoffs
yp -= ys
dy -= ys
if self.wave_title:
yp -= 2*ys
dy -= 2*ys
dy /= len(self.wave_ax)
for ax, has_xticks in zip(self.wave_ax, self.wave_has_xticks):
yp -= dy
ax.set_position([x0, yp, 1.0-x0-xs, dy])
if has_xticks:
yp -= yoffs
else:
yp -= ys
def _update_layout(self):
"""Update content and position of magnified plot."""
if self.scatter_indices[-1][1] < self.data.shape[1]:
if self.scatter_indices[-1][1] >= self.maxcols:
self.scatter_indices[-1][1] = self.maxcols-1
if self.scatter_indices[-1][0] >= self.scatter_indices[-1][1]:
self.scatter_indices[-1][0] = self.scatter_indices[-1][1]-1
self._plot_scatter(self.scatter_ax[-1], True, False)
else:
if self.scatter_indices[-1][0] >= self.maxcols:
self.scatter_indices[-1][0] = self.maxcols-1
self._plot_hist(self.scatter_ax[-1], True, False)
self._set_layout(self.fig.get_window_extent().width,
self.fig.get_window_extent().height)
self.fig.canvas.draw()
def _on_resize(self, event):
"""Adapt layout of plots to new figure size."""
self._set_layout(event.width, event.height)
def categorize(data):
"""Convert categorial string data into integer categories.
Parameters
----------
data: list of string
A list of textual categories.
Returns
-------
categories: list of strings
A sorted unique list of the strings in data.
cdata: list of integers
A copy of the input `data` where each string value is replaced
by an integer number that is an index into the reurned `categories`.
"""
cats = sorted(set(data))
cdata = np.array([cats.index(x) for x in data], dtype=int)
return cats, cdata
def main():
# parse command line:
parser = argparse.ArgumentParser(add_help=True,
description='View and explore multivariate data.',
epilog='version %s by Benda-Lab (2019-%s)' % (__version__, __year__))
parser.add_argument('file', nargs='?', default='', type=str,
help='a file containing a table of data (csv file or similar)')
args = parser.parse_args()
if args.file:
# load data:
data = TableData(args.file)
# initialize explorer:
expl = MultivariateExplorer(data)
else:
# generate data:
n = 100
data = []
data.append(np.random.randn(n) + 2.0)
data.append(1.0+0.1*data[0] + 1.5*np.random.randn(n))
data.append(-3.0*data[0] - 2.0*data[1] + 1.8*np.random.randn(n))
idx = np.random.randint(0, 3, n)
names = ['aaa', 'bbb', 'ccc']
data.append([names[i] for i in idx])
# generate waveforms:
waveforms = []
time = np.arange(0.0, 10.0, 0.01)
for r in range(len(data[0])):
x = data[0][r]*np.sin(2.0*np.pi*data[1][r]*time + data[2][r])
y = data[0][r]*np.exp(-0.5*((time-data[1][r])/(0.2*data[2][r]))**2.0)
waveforms.append(np.column_stack((time, x, y)))
#waveforms.append([np.column_stack((time, x)), np.column_stack((time, y))])
# initialize explorer:
expl = MultivariateExplorer(data,
list(map(chr, np.arange(len(data))+ord('A'))),
'Explorer')
expl.set_wave_data(waveforms, 'Time', ['Sine', 'Gauss'])
# explore data:
expl.set_colors()
expl.show()
if __name__ == '__main__':
main()Functions
def categorize(data)-
Convert categorial string data into integer categories.
Parameters
data:listofstring- A list of textual categories.
Returns
categories:listofstrings- A sorted unique list of the strings in data.
cdata:listofintegers- A copy of the input
datawhere each string value is replaced by an integer number that is an index into the reurnedcategories.
Expand source code
def categorize(data): """Convert categorial string data into integer categories. Parameters ---------- data: list of string A list of textual categories. Returns ------- categories: list of strings A sorted unique list of the strings in data. cdata: list of integers A copy of the input `data` where each string value is replaced by an integer number that is an index into the reurned `categories`. """ cats = sorted(set(data)) cdata = np.array([cats.index(x) for x in data], dtype=int) return cats, cdata def main()-
Expand source code
def main(): # parse command line: parser = argparse.ArgumentParser(add_help=True, description='View and explore multivariate data.', epilog='version %s by Benda-Lab (2019-%s)' % (__version__, __year__)) parser.add_argument('file', nargs='?', default='', type=str, help='a file containing a table of data (csv file or similar)') args = parser.parse_args() if args.file: # load data: data = TableData(args.file) # initialize explorer: expl = MultivariateExplorer(data) else: # generate data: n = 100 data = [] data.append(np.random.randn(n) + 2.0) data.append(1.0+0.1*data[0] + 1.5*np.random.randn(n)) data.append(-3.0*data[0] - 2.0*data[1] + 1.8*np.random.randn(n)) idx = np.random.randint(0, 3, n) names = ['aaa', 'bbb', 'ccc'] data.append([names[i] for i in idx]) # generate waveforms: waveforms = [] time = np.arange(0.0, 10.0, 0.01) for r in range(len(data[0])): x = data[0][r]*np.sin(2.0*np.pi*data[1][r]*time + data[2][r]) y = data[0][r]*np.exp(-0.5*((time-data[1][r])/(0.2*data[2][r]))**2.0) waveforms.append(np.column_stack((time, x, y))) #waveforms.append([np.column_stack((time, x)), np.column_stack((time, y))]) # initialize explorer: expl = MultivariateExplorer(data, list(map(chr, np.arange(len(data))+ord('A'))), 'Explorer') expl.set_wave_data(waveforms, 'Time', ['Sine', 'Gauss']) # explore data: expl.set_colors() expl.show()
Classes
class MultivariateExplorer (data, labels=None, title=None)-
Simple matplotlib-based GUI for viewing and exploring multivariate data.
Shown are scatter plots of all pairs of variables or PCA axis. Points in the scatter plots are colored according to the values of one of the variables. Data points can be selected and optionally corresponding waveforms are shown.
First you initialize the explorer with the data. Then you optionally specify how to colorize the data and provide waveform data associated with the data. Finally you show the figure:
expl = MultivariateExplorer(data) expl.set_colors(2) expl.set_wave_data(waveforms, 'Time [s]', 'Sine') expl.show()The `compute_pca() function computes a principal component analysis (PCA) on the input data, and
save_pca()writes the principal components to a file.Customize the appearance and information provided by subclassing MultivariateExplorer and reimplementing the functions - fix_scatter_plot() - fix_waveform_plot() - list_selection() - analyze_selection() See the documentation of these functions for details.
Initialize explorer with scatter-plot data.
Parameters
data:TableData, 2D array,orlistof1D arrays- The data to be explored. Each column is a variable. For the 2D array the columns are the second dimension, for a list of 1D arrays, the list goes over columns, i.e. each 1D array is one column.
labels:listofstring- If data is not a TableData, then this provides labels for the data columns.
title:string- Title for the window.
Expand source code
class MultivariateExplorer(object): """Simple matplotlib-based GUI for viewing and exploring multivariate data. Shown are scatter plots of all pairs of variables or PCA axis. Points in the scatter plots are colored according to the values of one of the variables. Data points can be selected and optionally corresponding waveforms are shown. First you initialize the explorer with the data. Then you optionally specify how to colorize the data and provide waveform data associated with the data. Finally you show the figure: ``` expl = MultivariateExplorer(data) expl.set_colors(2) expl.set_wave_data(waveforms, 'Time [s]', 'Sine') expl.show() ``` The `compute_pca() function computes a principal component analysis (PCA) on the input data, and `save_pca()` writes the principal components to a file. Customize the appearance and information provided by subclassing MultivariateExplorer and reimplementing the functions - fix_scatter_plot() - fix_waveform_plot() - list_selection() - analyze_selection() See the documentation of these functions for details. """ mouse_actions = (('left click', 'select data points'), ('left and drag', 'rectangular selection and zoom of data points'), ('shift + left click/drag', 'add data points to selection'), ('ctrl + left click/drag', 'remove data points from selection')) key_actions = (('l', 'list selected EOD waveforms on console'), ('p,P', 'toggle between data columns, PC, and scaled PC axis'), ('<, pageup', 'decrease number of displayed data columns/PC axis'), ('>, pagedown', 'increase number of displayed data columns/PC axis'), ('w', 'toggle maximized waveform plot'), ('o, z', 'toggle zoom mode on or off'), ('backspace', 'zoom back'), ('ctrl + a', 'select all'), ('+, -', 'increase, decrease pick radius'), ('0', 'reset pick radius'), ('n, N', 'decrease, increase number of bins of histograms'), ('h', 'toggle between scatter plot and 2D histogram'), ('c, C', 'cycle color map trough data columns'), ('left, right, up, down', 'show and move magnified scatter plot'), ('escape', 'close magnified scatter plot')) def __init__(self, data, labels=None, title=None): """Initialize explorer with scatter-plot data. Parameters ---------- data: TableData, 2D array, or list of 1D arrays The data to be explored. Each column is a variable. For the 2D array the columns are the second dimension, for a list of 1D arrays, the list goes over columns, i.e. each 1D array is one column. labels: list of string If data is not a TableData, then this provides labels for the data columns. title: string Title for the window. """ # data and labels: if isinstance(data, TableData): self.categories = [] for c, col in enumerate(data): if not isinstance(col[0], (int, float, np.integer, np.floating)): # categorial data: cats, data[:,c] = categorize(col) self.categories.append(cats) else: self.categories.append(None) self.raw_data = data.array() if labels is None: self.raw_labels = [] for c in range(len(data)): if len(data.unit(c)) > 0 and not data.unit(c) in ['-', '1']: self.raw_labels.append('%s [%s]' % (data.label(c), data.unit(c))) else: self.raw_labels.append(data.label(c)) else: self.raw_labels = labels else: if isinstance(data, np.ndarray): self.raw_data = data else: self.categories = [] for c, col in enumerate(data): if not isinstance(col[0], (int, float, np.integer, np.floating)): # categorial data: cats, data[c] = categorize(col) self.categories.append(cats) else: self.categories.append(None) self.raw_data = np.asarray(data).T self.raw_labels = labels self.title = title if title is not None else 'MultivariateExplorer' # no pca data yet: self.all_data = [self.raw_data, None, None] self.all_labels = [self.raw_labels, None, None] self.all_maxcols = [self.raw_data.shape[1], None, None] self.all_titles = ['data', 'PCA', 'scaled PCA'] # pca: self.pca_tables = [None, None] self._pca_header(data, labels) # start showing raw data: self.show_mode = 0 self.data = self.all_data[self.show_mode] self.labels = self.all_labels[self.show_mode] self.maxcols = self.all_maxcols[self.show_mode] if self.maxcols > 6: self.maxcols = 6 # waveform data: self.wave_data = [] self.wave_nested = False self.wave_has_xticks = [] self.wave_xlabels = [] self.wave_ylabels = [] self.wave_title = False # colors: self.color_map = plt.get_cmap('jet') self.extra_colors = None self.extra_color_label = None self.extra_categories = None self.color_values = None self.color_set_index = 0 self.color_index = 0 self.color_label = None self.color_set_index = 0 self.color_index = 0 self.data_colors = None self.color_vmin = None self.color_vmax = None self.color_ticks = None self.cbax = None # figure variables: self.plt_params = {} for k in ['toolbar', 'keymap.quit', 'keymap.back', 'keymap.forward', 'keymap.zoom', 'keymap.pan', 'keymap.xscale', 'keymap.yscale']: self.plt_params[k] = plt.rcParams[k] if k != 'toolbar': plt.rcParams[k] = '' self.xborder = 70.0 # pixel for ylabels self.yborder = 50.0 # pixel for xlabels self.spacing = 10.0 # pixel between plots self.pick_radius = 4.0 # histogram plots: self.hist_ax = [] self.hist_indices = [] self.hist_selector = [] self.hist_nbins = 30 # scatter plots: self.scatter_ax = [] self.scatter_indices = [] self.scatter_artists = [] self.scatter_selector = [] self.scatter = True self.mark_data = [] self.select_zooms = False self.zoom_stack = [] # magnified scatter plot: self.magnified_on = False self.magnified_backdrop = None self.magnified_size = np.array([0.5, 0.5]) # waveform plots: self.wave_ax = [] def set_wave_data(self, data, xlabels='', ylabels=[], title=False): """Add waveform data to explorer. Parameters ---------- data: list of (list of) 2D arrays Waveform data associated with each row of the data. Elements of the outer list correspond to the rows of the data. The inner 2D arrays contain a common x-axes (first column) and one or more corresponding y-values (second and optional higher columns). Each column for y-values is plotted in its own axes on top of each other, from top to bottom. The optional inner list of 2D arrays contains several 2D arrays as ascribed above each with its own common x-axes. xlabel: string or list of strings The xlabels for the waveform plots. If only a string is given, then there will be a common xaxis for all the plots, and only the lowest one gets a labeled xaxis. If a list of strings is given, each waveform plot gets its own labeled x-axis. ylabels: list of strings The ylabels for each of the waveform plots. title: bool or string If True or a string, povide space on top of the waveform plots for a title. If string, set this as the title for the waveform plots. """ self.wave_data = [] if data is not None and len(data) > 0: self.wave_data = data self.wave_has_xticks = [] self.wave_nested = isinstance(data[0], (list, tuple)) if self.wave_nested: for data in self.wave_data[0]: for k in range(data.shape[1]-2): self.wave_has_xticks.append(False) self.wave_has_xticks.append(True) else: for k in range(self.wave_data[0].shape[1]-2): self.wave_has_xticks.append(False) self.wave_has_xticks.append(True) if isinstance(xlabels, (list, tuple)): self.wave_xlabels = xlabels else: self.wave_xlabels = [xlabels] self.wave_ylabels = ylabels self.wave_title = title self.wave_ax = [] def set_colors(self, colors=0, color_label=None, color_map=None): """Set data column used to color scatter plots. Parameters ---------- colors: int or 1D array Index to colum in data to be used for coloring scatter plots. -2 for coloring row index of data. Or data array used to color scalar plots. color_label: string If colors is an array, this is a label describing the data. It is used to label the color bar. color_map: string or None Name of a matplotlib color map. If None 'jet' is used. """ if isinstance(colors, (np.integer, int)): if colors < 0: self.color_set_index = -1 self.color_index = 0 else: self.color_set_index = 0 self.color_index = colors else: if not isinstance(colors[0], (int, float, np.integer, np.floating)): # categorial data: self.extra_categories, self.extra_colors = categorize(colors) else: self.extra_colors = colors self.extra_color_label = color_label self.color_set_index = -1 self.color_index = 1 self.color_map = plt.get_cmap(color_map if color_map else 'jet') def show(self): """Show interactive scatter plots for exploration. """ plt.ioff() plt.rcParams['toolbar'] = 'None' plt.rcParams['keymap.quit'] = 'ctrl+w, alt+q, q' self.fig = plt.figure(facecolor='white') self.fig.canvas.manager.set_window_title(self.title + ': ' + self.all_titles[self.show_mode]) self.fig.canvas.mpl_connect('key_press_event', self._on_key) self.fig.canvas.mpl_connect('resize_event', self._on_resize) self.fig.canvas.mpl_connect('pick_event', self._on_pick) if self.color_map is None: self.color_map = plt.get_cmap('jet') self._set_color_column() self._init_hist_plots() self._init_scatter_plots() self.wave_ax = [] if self.wave_data is not None and len(self.wave_data) > 0: axx = None xi = 0 for k, has_xticks in enumerate(self.wave_has_xticks): ax = self.fig.add_subplot(1, len(self.wave_has_xticks), 1+k, sharex=axx) self.wave_ax.append(ax) if has_xticks: if xi >= len(self.wave_xlabels): self.wave_xlabels.append('') ax.set_xlabel(self.wave_xlabels[xi]) xi += 1 axx = None else: plt.setp(ax.get_xticklabels(), visible=False) if axx is None: axx = ax for ax, ylabel in zip(self.wave_ax, self.wave_ylabels): ax.set_ylabel(ylabel) if not isinstance(self.wave_title, bool) and self.wave_title: self.wave_ax[0].set_title(self.wave_title) self.fix_waveform_plot(self.wave_ax, self.mark_data) self._plot_magnified_scatter() plt.show() def _pca_header(self, data, labels): """Set up header for the table of principal components. """ if isinstance(data, TableData): header = data.table_header() for c in reversed(range(data.columns())): if not np.all(np.isfinite(data[:,c])): header.remove(c) else: lbs = [] for l, d in zip(labels, data): if not np.all(np.isfinite(d)): continue if '[' in l: lbs.append(l.split('[')[0].strip()) elif '/' in l: lbs.append(l.split('/')[0].strip()) else: lbs.append(l) header = TableData(header=lbs) header.set_formats('%.3f') header.insert(0, ['PC'] + ['-']*header.nsecs, '', '%d') header.insert(1, 'variance', '%', '%.3f') for k in range(len(self.pca_tables)): self.pca_tables[k] = TableData(header) def compute_pca(self, scale=False, write=False): """Compute PCA based on the data. Parameters ---------- scale: boolean If True standardize data before computing PCA, i.e. remove mean of each variabel and divide by its standard deviation. write: boolean If True write PCA components to standard out. """ # select columns without NANs: idxs = [i for i in range(self.raw_data.shape[1]) if np.all(np.isfinite(self.raw_data[:,i]))] data = self.raw_data[:,idxs] # pca: pca = decomposition.PCA() if scale: scaler = preprocessing.StandardScaler() scaler.fit(data) pca.fit(scaler.transform(data)) pca_label = 'sPC' else: pca.fit(data) pca_label = 'PC' for k in range(len(pca.components_)): if np.abs(np.min(pca.components_[k])) > np.max(pca.components_[k]): pca.components_[k] *= -1.0 pca_data = pca.transform(data) pca_labels = [('%s%d (%.1f%%)' if v > 0.01 else '%s%d (%.2f%%)') % (pca_label, k+1, 100.0*v) for k, v in enumerate(pca.explained_variance_ratio_)] if np.min(pca.explained_variance_ratio_) >= 0.01: pca_maxcols = pca_data.shape[1] else: pca_maxcols = np.argmax(pca.explained_variance_ratio_ < 0.01) if pca_maxcols < 2: pca_maxcols = 2 if pca_maxcols > 6: pca_maxcols = 6 # table with PCA feature weights: pca_table = self.pca_tables[1] if scale else self.pca_tables[0] pca_table.clear_data() pca_table.set_section(pca_label, 0, pca_table.nsecs) for k, comp in enumerate(pca.components_): pca_table.append_data(k+1, 0) pca_table.append_data(100.0*pca.explained_variance_ratio_[k]) pca_table.append_data(comp) if write: pca_table.write(table_format='out', unit_style='none') # submit data: if scale: self.all_data[2] = pca_data self.all_labels[2] = pca_labels self.all_maxcols[2] = pca_maxcols else: self.all_data[1] = pca_data self.all_labels[1] = pca_labels self.all_maxcols[1] = pca_maxcols def save_pca(self, file_name, scale, **kwargs): """Write PCA data to file. Parameters ---------- file_name: string Name of ouput file. scale: boolean If True write PCA components of standardized PCA. kwargs: dict Additional parameter for TableData.write() """ if scale: pca_file = file_name + '-pcacor' pca_table = self.pca_tables[1] else: pca_file = file_name + '-pcacov' pca_table = self.pca_tables[0] if 'unit_style' in kwargs: del kwargs['unit_style'] if 'table_format' in kwargs: pca_table.write(pca_file, unit_style='none', **kwargs) else: pca_file += '.dat' pca_table.write(pca_file, unit_style='none') def _set_color_column(self): """Initialize variables used for colorization of scatter points.""" if self.color_set_index == -1: if self.color_index == 0: self.color_values = np.arange(self.data.shape[0], dtype=float) self.color_label = 'row' elif self.color_index == 1: self.color_values = self.extra_colors self.color_label = self.extra_color_label else: self.color_values = self.all_data[self.color_set_index][:,self.color_index] self.color_label = self.all_labels[self.color_set_index][self.color_index] self.color_vmin, self.color_vmax, self.color_ticks = \ self.fix_scatter_plot(self.cbax, self.color_values, self.color_label, 'c') if self.color_ticks is None: if self.color_set_index == 0 and \ self.categories[self.color_index] is not None: self.color_ticks = np.arange(len(self.categories[self.color_index])) elif self.color_set_index == -1 and \ self.color_index == 1 and \ self.extra_categories is not None: self.color_ticks = np.arange(len(self.extra_categories)) self.data_colors = self.color_map((self.color_values - self.color_vmin)/(self.color_vmax - self.color_vmin)) def _plot_hist(self, ax, magnifiedax, keep_lims): """Plot and label a histogram.""" ax_xlim = ax.get_xlim() ax_ylim = ax.get_ylim() try: idx = self.hist_ax.index(ax) c = self.hist_indices[idx] in_hist = True except ValueError: idx = self.scatter_ax.index(ax) c = self.scatter_indices[-1][0] in_hist = False ax.clear() ax.relim() ax.autoscale(True) x = self.data[:,c] ax.hist(x[np.isfinite(x)], self.hist_nbins) ax.set_xlabel(self.labels[c]) if self.categories[c] is not None: ax.set_xticks(np.arange(len(self.categories[c]))) ax.set_xticklabels(self.categories[c]) self.fix_scatter_plot(ax, self.data[:,c], self.labels[c], 'x') if magnifiedax: ax.set_ylabel('count') cax = self.hist_ax[self.scatter_indices[-1][0]] ax.set_xlim(cax.get_xlim()) else: if c == 0: ax.set_ylabel('count') else: plt.setp(ax.get_yticklabels(), visible=False) if keep_lims: ax.set_xlim(*ax_xlim) ax.set_ylim(*ax_ylim) try: selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1, props=dict(facecolor='gray', edgecolor='gray', alpha=0.2, fill=True), state_modifier_keys=dict(move='', clear='', square='', center='')) except TypeError: selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1) if in_hist: self.hist_selector[idx] = selector else: self.scatter_selector[idx] = selector self.scatter_artists[idx] = None if magnifiedax: bbox = ax.get_tightbbox(self.fig.canvas.get_renderer()) if bbox is not None: self.magnified_backdrop = patches.Rectangle((bbox.x0, bbox.y0), bbox.width, bbox.height, transform=None, clip_on=False, facecolor='white', edgecolor='none', zorder=-5) ax.add_patch(self.magnified_backdrop) def _init_hist_plots(self): """Initial plots of the histograms.""" n = self.data.shape[1] yax = None self.hist_ax = [] for r in range(n): ax = self.fig.add_subplot(n, n, (n-1)*n+r+1, sharey=yax) self.hist_ax.append(ax) self.hist_indices.append(r) self.hist_selector.append(None) self._plot_hist(ax, False, False) yax = ax def _plot_scatter(self, ax, magnifiedax, keep_lims, cax=None): """Plot a scatter plot.""" ax_xlim = ax.get_xlim() ax_ylim = ax.get_ylim() idx = self.scatter_ax.index(ax) c, r = self.scatter_indices[idx] sel = np.isfinite(self.data[:,c]) & np.isfinite(self.data[:,r]) if self.scatter: ax.clear() ax.relim() ax.autoscale(True) a = ax.scatter(self.data[sel,c], self.data[sel,r], c=self.color_values[sel], cmap=self.color_map, vmin=self.color_vmin, vmax=self.color_vmax, s=50, edgecolors='none', zorder=10) if cax is not None: self.fig.colorbar(a, cax=cax, ticks=self.color_ticks) cax.set_ylabel(self.color_label) self.color_vmin, self.color_vmax, self.color_ticks = \ self.fix_scatter_plot(self.cbax, self.color_values, self.color_label, 'c') if self.color_ticks is None: if self.color_set_index == 0 and \ self.categories[self.color_index] is not None: cax.set_yticklabels(self.categories[self.color_index]) elif self.color_set_index == -1 and \ self.color_index == 1 and \ self.extra_categories is not None: cax.set_yticklabels(self.extra_categories) else: ax.autoscale(True) self.fix_scatter_plot(ax, self.data[sel,c], self.labels[c], 'x') self.fix_scatter_plot(ax, self.data[sel,r], self.labels[r], 'y') axrange = [ax.get_xlim(), ax.get_ylim()] ax.clear() ax.hist2d(self.data[sel,c], self.data[sel,r], self.hist_nbins, range=axrange, cmap=plt.get_cmap('Greys')) md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and np.isfinite(self.data[m,r])] a = ax.scatter(self.data[md,c], self.data[md,r], c=self.data_colors[md], s=80, zorder=11) self.scatter_artists[idx] = a if self.categories[c] is not None: ax.set_xticks(np.arange(len(self.categories[c]))) ax.set_xticklabels(self.categories[c]) if self.categories[r] is not None: ax.set_yticks(np.arange(len(self.categories[r]))) ax.set_yticklabels(self.categories[r]) if magnifiedax: ax.set_xlabel(self.labels[c]) ax.set_ylabel(self.labels[r]) cax = self.scatter_ax[self.scatter_indices[:-1].index(self.scatter_indices[-1])] ax.set_xlim(cax.get_xlim()) ax.set_ylim(cax.get_ylim()) else: if c == 0: ax.set_ylabel(self.labels[r]) self.fix_scatter_plot(ax, self.data[sel,c], self.labels[c], 'x') self.fix_scatter_plot(ax, self.data[sel,r], self.labels[r], 'y') if not magnifiedax: plt.setp(ax.get_xticklabels(), visible=False) if c > 0: plt.setp(ax.get_yticklabels(), visible=False) if keep_lims: ax.set_xlim(*ax_xlim) ax.set_ylim(*ax_ylim) if magnifiedax: bbox = ax.get_tightbbox(self.fig.canvas.get_renderer()) if bbox is not None: self.magnified_backdrop = patches.Rectangle((bbox.x0, bbox.y0), bbox.width, bbox.height, transform=None, clip_on=False, facecolor='white', edgecolor='none', zorder=-5) ax.add_patch(self.magnified_backdrop) try: selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1, props=dict(facecolor='gray', edgecolor='gray', alpha=0.2, fill=True), state_modifier_keys=dict(move='', clear='', square='', center='')) except TypeError: selector = widgets.RectangleSelector(ax, self._on_select, useblit=True, button=1) self.scatter_selector[idx] = selector def _init_scatter_plots(self): """Initial plots of scatter plots.""" self.cbax = self.fig.add_axes([0.5, 0.5, 0.1, 0.5]) cbax = self.cbax n = self.data.shape[1] for r in range(1, n): yax = None for c in range(r): ax = self.fig.add_subplot(n, n, (r-1)*n+c+1, sharex=self.hist_ax[c], sharey=yax) self.scatter_ax.append(ax) self.scatter_indices.append([c, r]) self.scatter_artists.append(None) self.scatter_selector.append(None) self._plot_scatter(ax, False, False, cbax) yax = ax cbax = None def _plot_magnified_scatter(self): """Initial plot of the magnified scatter plot.""" ax = self.fig.add_axes([0.5, 0.9, 0.05, 0.05]) ax.set_visible(False) self.magnified_on = False c = 0 r = 1 sel = np.isfinite(self.data[:,c]) & np.isfinite(self.data[:,r]) ax.scatter(self.data[sel,c], self.data[sel,r], c=self.data_colors[sel], s=50, edgecolors='none') md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and np.isfinite(self.data[m,r])] a = ax.scatter(self.data[md,c], self.data[md,r], c=self.data_colors[md], s=80) ax.set_xlabel(self.labels[c]) ax.set_ylabel(self.labels[r]) self.fix_scatter_plot(ax, self.data[sel,c], self.labels[c], 'x') self.fix_scatter_plot(ax, self.data[sel,r], self.labels[r], 'y') self.scatter_ax.append(ax) self.scatter_indices.append([c, r]) self.scatter_artists.append(a) self.scatter_selector.append(None) def fix_scatter_plot(self, ax, data, label, axis): """Customize an axes of a scatter plot. This function is called after a scatter plot has been plotted. Once for the x axes, once for the y axis and once for the color bar. Reimplement this function to set appropriate limits and ticks. Return values are only used for the color bar (`axis='c'`). Otherwise they are ignored. For example, ticks for phase variables can be nicely labeled using the unicode character for pi: ``` if 'phase' in label: if axis == 'y': ax.set_ylim(0.0, 2.0*np.pi) ax.set_yticks(np.arange(0.0, 2.5*np.pi, 0.5*np.pi)) ax.set_yticklabels(['0', u'\u03c0/2', u'\u03c0', u'3\u03c0/2', u'2\u03c0']) ``` Parameters ---------- ax: matplotlib axes Axes of the scatter plot or color bar to be worked on. data: 1D array Data array of the axes. label: string Label coresponding to the data array. axis: str 'x', 'y': set properties of x or y axes of ax. 'c': set properies of color bar axes (note that ax can be None!) and return vmin, vmax, and ticks. Returns ------- min: float minimum value of color bar axis max: float maximum value of color bar axis ticks: list of float position of ticks for color bar axis """ return np.nanmin(data), np.nanmax(data), None def fix_waveform_plot(self, axs, indices): """Customize waveform plots. This function is called once after new data have been plotted into the waveform plots. Reimplement this function to customize these plots. In particular to set axis limits and labels, plot title, etc. You may even open a new figure (with non-blocking `show()`). The following member variables might be usefull: - `self.wave_data`: the full list of waveform data. - `self.wave_nested`: True if the elements of `self.wave_data` are lists of 2D arrays. Otherwise the elements are 2D arrays. The first column of a 2D array contains the x-values, further columns y-values. - `self.wave_has_xticks`: List of booleans for each axis. True if the axis has its own xticks. - `self.wave_xlabels`: List of xlabels (only for the axis where the corresponding entry in `self.wave_has_xticks` is True). - `self.wave_ylabels`: for each axis its ylabel For example, you can set the linewidth of all plotted waveforms via: ``` for ax in axs: for l in ax.lines: l.set_linewidth(3.0) ``` or enable markers to be plotted: ``` for ax, yl in zip(axs, self.wave_ylabels): if 'Power' in yl: for l in ax.lines: l.set_marker('.') l.set_markersize(15.0) l.set_markeredgewidth(0.5) l.set_markeredgecolor('k') l.set_markerfacecolor(l.get_color()) ``` Usefull is to reduce the maximum number of y-ticks: ``` axs[0].yaxis.get_major_locator().set_params(nbins=7) ``` or ``` import matplotlib.ticker as ticker axs[0].yaxis.set_major_locator(ticker.MaxNLocator(nbins=4)) ``` Parameters ---------- axs: list of matplotlib axes Axis of the waveform plots to be worked on. indices: list of int Indices of the waveforms that have been selected and plotted. """ pass def list_selection(self, indices): """List information about the current selection of data points. This function is called when 'l' is pressed. Reimplement this function, for example, to print some meaningfull information about the current selection of data points on console. You nay do, however, whatever you want in this function. Parameters ---------- indices: list of int Indices of the data points that have been selected. """ for i in indices: print(i) def analyze_selection(self, index): """Provide further information about a single selected data point. This function is called when a single data item was double clicked. Reimplement this function to provide some further details on this data point. This can be an additional figure window. In this case show it non-blocking: `plt.show(block=False)` Parameters ---------- index: int The index of the selected data point. """ pass def _set_magnified_pos(self, width, height): """Set position of magnified plot.""" if self.magnified_on: xoffs = self.xborder/width yoffs = self.yborder/height if self.scatter_indices[-1][1] < self.data.shape[1]: idx = self.scatter_indices[:-1].index(self.scatter_indices[-1]) pos = self.scatter_ax[idx].get_position().get_points() else: pos = self.hist_ax[self.scatter_indices[-1][0]].get_position().get_points() pos[0] = np.mean(pos, 0) - 0.5*self.magnified_size if pos[0][0] < xoffs: pos[0][0] = xoffs if pos[0][1] < yoffs: pos[0][1] = yoffs pos[1] = pos[0] + self.magnified_size if pos[1][0] > 1.0-self.spacing/width: pos[1][0] = 1.0-self.spacing/width if pos[1][1] > 1.0-self.spacing/height: pos[1][1] = 1.0-self.spacing/height pos[0] = pos[1] - self.magnified_size self.scatter_ax[-1].set_position([pos[0][0], pos[0][1], self.magnified_size[0], self.magnified_size[1]]) self.scatter_ax[-1].set_visible(True) else: self.scatter_ax[-1].set_position([0.5, 0.9, 0.05, 0.05]) self.scatter_ax[-1].set_visible(False) def _make_selection(self, ax, key, x0, x1, y0, y1): """Select points from a scatter or histogram plot.""" if not key in ['shift', 'control']: self.mark_data = [] try: axi = self.scatter_ax.index(ax) # from scatter plots: c, r = self.scatter_indices[axi] if r < self.data.shape[1]: # from scatter: for ind, (x, y) in enumerate(zip(self.data[:,c], self.data[:,r])): if np.isfinite(x) and np.isfinite(y) and \ x >= x0 and x <= x1 and y >= y0 and y <= y1: if ind in self.mark_data: if key == 'control': self.mark_data.remove(ind) elif key != 'control': self.mark_data.append(ind) else: # from histogram: for ind, x in enumerate(self.data[:,c]): if np.isfinite(x) and x >= x0 and x <= x1: if ind in self.mark_data: if key == 'control': self.mark_data.remove(ind) elif key != 'control': self.mark_data.append(ind) except ValueError: try: r = self.hist_ax.index(ax) # from histogram: for ind, x in enumerate(self.data[:,r]): if np.isfinite(x) and x >= x0 and x <= x1: if ind in self.mark_data: if key == 'control': self.mark_data.remove(ind) elif key != 'control': self.mark_data.append(ind) except ValueError: return def _update_selection(self): """Highlight select points in the scatter plots and plot corresponding waveforms.""" # update scatter plots: for artist, (c, r) in zip(self.scatter_artists, self.scatter_indices): if artist is not None: md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and np.isfinite(self.data[m,r])] artist.set_offsets(list(zip(self.data[md,c], self.data[md,r]))) artist.set_facecolors(self.data_colors[md]) # waveform plots: if len(self.wave_ax) > 0: axdi = 0 axti = 1 for xi, ax in enumerate(self.wave_ax): ax.clear() if len(self.mark_data) > 0: for idx in self.mark_data: if self.wave_nested: data = self.wave_data[idx][axdi] else: data = self.wave_data[idx] if data is not None: ax.plot(data[:,0], data[:,axti], c=self.data_colors[idx], picker=self.pick_radius) axti += 1 if self.wave_has_xticks[xi]: ax.set_xlabel(self.wave_xlabels[axdi]) axti = 1 axdi += 1 else: plt.setp(ax.get_xticklabels(), visible=False) for ax, ylabel in zip(self.wave_ax, self.wave_ylabels): ax.set_ylabel(ylabel) if not isinstance(self.wave_title, bool) and self.wave_title: self.wave_ax[0].set_title(self.wave_title) self.fix_waveform_plot(self.wave_ax, self.mark_data) self.fig.canvas.draw() def _on_key(self, event): """Handle key events.""" #print('pressed', event.key) plot_zoom = True if event.key in ['left', 'right', 'up', 'down']: if self.magnified_on: if event.key == 'left': if self.scatter_indices[-1][0] > 0: self.scatter_indices[-1][0] -= 1 else: plot_zoom = False elif event.key == 'right': if self.scatter_indices[-1][0] < self.scatter_indices[-1][1]-1 and \ self.scatter_indices[-1][0] < self.maxcols-1: self.scatter_indices[-1][0] += 1 else: plot_zoom = False elif event.key == 'up': if self.scatter_indices[-1][1] > 1: if self.scatter_indices[-1][1] >= self.data.shape[1]: self.scatter_indices[-1][1] = self.maxcols-1 else: self.scatter_indices[-1][1] -= 1 if self.scatter_indices[-1][0] >= self.scatter_indices[-1][1]: self.scatter_indices[-1][0] = self.scatter_indices[-1][1]-1 else: plot_zoom = False elif event.key == 'down': if self.scatter_indices[-1][1] < self.maxcols: self.scatter_indices[-1][1] += 1 if self.scatter_indices[-1][1] >= self.maxcols: self.scatter_indices[-1][1] = self.data.shape[1] else: plot_zoom = False else: plot_zoom = False if event.key == 'escape': self.scatter_ax[-1].set_position([0.5, 0.9, 0.05, 0.05]) self.magnified_on = False self.scatter_ax[-1].set_visible(False) self.fig.canvas.draw() elif event.key in 'oz': self.select_zooms = not self.select_zooms elif event.key == 'backspace': if len(self.zoom_stack) > 0: ax, xmin, xmax, ymin, ymax = self.zoom_stack.pop() ax.set_xlim(xmin, xmax) ax.set_ylim(ymin, ymax) if ax in self.scatter_ax[:-1]: axidx = self.scatter_ax[:-1].index(ax) if self.scatter_indices[axidx][0] == self.scatter_indices[-1][0]: self.scatter_ax[-1].set_xlim(xmin, xmax) if self.scatter_indices[axidx][1] == self.scatter_indices[-1][1]: self.scatter_ax[-1].set_ylim(ymin, ymax) elif ax in self.hist_ax: if self.scatter_indices[-1][1] == self.data.shape[1] and \ self.scatter_indices[-1][0] == self.hist_ax.index(ax): self.scatter_ax[-1].set_xlim(xmin, xmax) self.scatter_ax[-1].set_ylim(ymin, ymax) self.fig.canvas.draw() elif event.key in '+=': self.pick_radius *= 1.5 elif event.key in '-': if self.pick_radius > 5.0: self.pick_radius /= 1.5 elif event.key in '0': self.pick_radius = 4.0 elif event.key in ['pageup', 'pagedown', '<', '>']: if event.key in ['pageup', '<'] and self.maxcols > 2: self.maxcols -= 1 elif event.key in ['pagedown', '>'] and self.maxcols < self.raw_data.shape[1]: self.maxcols += 1 self._update_layout() elif event.key == 'w': if self.maxcols > 0: self.all_maxcols[self.show_mode] = self.maxcols self.maxcols = 0 else: self.maxcols = self.all_maxcols[self.show_mode] self._set_layout(self.fig.get_window_extent().width, self.fig.get_window_extent().height) self.fig.canvas.draw() elif event.key == 'ctrl+a': self.mark_data = range(len(self.data)) self._update_selection() elif event.key in 'cC': if event.key in 'c': first = True while first or not np.all(np.isfinite(self.color_values)): self.color_index -= 1 if self.color_index < 0: self.color_set_index -= 1 if self.color_set_index < -1: self.color_set_index = len(self.all_data)-1 if self.color_set_index >= 0: if self.all_data[self.color_set_index] is None: self.compute_pca(self.color_set_index>1, True) self.color_index = self.all_data[self.color_set_index].shape[1]-1 else: self.color_index = 0 if self.extra_colors is None else 1 self._set_color_column() first = False else: first = True while first or not np.all(np.isfinite(self.color_values)): self.color_index += 1 if (self.color_set_index >= 0 and \ self.color_index >= self.all_data[self.color_set_index].shape[1]) or \ (self.color_set_index < 0 and \ self.color_index >= (1 if self.extra_colors is None else 2)): self.color_index = 0 self.color_set_index += 1 if self.color_set_index >= len(self.all_data): self.color_set_index = -1 elif self.all_data[self.color_set_index] is None: self.compute_pca(self.color_set_index>1, True) self._set_color_column() first = False for ax in self.scatter_ax: if len(ax.collections) > 0: idx = self.scatter_ax.index(ax) c, r = self.scatter_indices[idx] ax.collections[0].set_facecolors(self.data_colors[np.isfinite(self.data[:,c]) & np.isfinite(self.data[:,r])]) for a, (c, r) in zip(self.scatter_artists, self.scatter_indices): if a is not None: md = [m for m in self.mark_data if np.isfinite(self.data[m,c]) and np.isfinite(self.data[m,r])] a.set_facecolors(self.data_colors[md]) for ax in self.wave_ax: for l, c in zip(ax.lines, self.data_colors[self.mark_data]): l.set_color(c) l.set_markerfacecolor(c) self._plot_scatter(self.scatter_ax[0], False, True, self.cbax) self.fix_scatter_plot(self.cbax, self.color_values, self.color_label, 'c') self.fig.canvas.draw() elif event.key in 'nN': if event.key in 'N': self.hist_nbins = (self.hist_nbins*3)//2 elif self.hist_nbins >= 15: self.hist_nbins = (self.hist_nbins*2)//3 for ax in self.hist_ax: self._plot_hist(ax, False, True) if self.scatter_indices[-1][1] >= self.data.shape[1]: self._plot_hist(self.scatter_ax[-1], True, True) elif not self.scatter: self._plot_scatter(self.scatter_ax[-1], True, True) if not self.scatter: for ax in self.scatter_ax[:-1]: self._plot_scatter(ax, False, True) self.fig.canvas.draw() elif event.key in 'h': self.scatter = not self.scatter for ax in self.scatter_ax[:-1]: self._plot_scatter(ax, False, True) if self.scatter_indices[-1][1] < self.data.shape[1]: self._plot_scatter(self.scatter_ax[-1], True, True) self.fig.canvas.draw() elif event.key in 'pP': self.all_maxcols[self.show_mode] = self.maxcols if event.key == 'p': self.show_mode += 1 if self.show_mode >= len(self.all_data): self.show_mode = 0 else: self.show_mode -= 1 if self.show_mode < 0: self.show_mode = len(self.all_data)-1 if self.show_mode == 1: print('principal components') elif self.show_mode == 2: print('scaled principal components') else: print('data') if self.all_data[self.show_mode] is None: self.compute_pca(self.show_mode>1, True) self.data = self.all_data[self.show_mode] self.labels = self.all_labels[self.show_mode] self.maxcols = self.all_maxcols[self.show_mode] self.zoom_stack = [] self.fig.canvas.manager.set_window_title(self.title + ': ' + self.all_titles[self.show_mode]) for ax in self.hist_ax[:self.maxcols]: self._plot_hist(ax, False, False) for ax in self.scatter_ax[:self.maxcols]: self._plot_scatter(ax, False, False) self._update_layout() elif event.key in 'l': if len(self.mark_data) > 0: print('') print('selected:') self.list_selection(self.mark_data) if plot_zoom: for k in reversed(range(len(self.zoom_stack))): if self.zoom_stack[k][0] == self.scatter_ax[-1]: del self.zoom_stack[k] self.scatter_ax[-1].clear() self.scatter_ax[-1].set_visible(True) self.magnified_on = True self._set_magnified_pos(self.fig.get_window_extent().width, self.fig.get_window_extent().height) if self.scatter_indices[-1][1] < self.data.shape[1]: self._plot_scatter(self.scatter_ax[-1], True, False) else: self._plot_hist(self.scatter_ax[-1], True, False) self.fig.canvas.draw() def _on_select(self, eclick, erelease): """Handle selection events.""" if eclick.dblclick: if len(self.mark_data) > 0: self.analyze_selection(self.mark_data[-1]) return x0 = min(eclick.xdata, erelease.xdata) x1 = max(eclick.xdata, erelease.xdata) y0 = min(eclick.ydata, erelease.ydata) y1 = max(eclick.ydata, erelease.ydata) ax = erelease.inaxes if ax is None: ax = eclick.inaxes xmin, xmax = ax.get_xlim() ymin, ymax = ax.get_ylim() dx = 0.02*(xmax-xmin) dy = 0.02*(ymax-ymin) if x1 - x0 < dx and y1 - y0 < dy: bbox = ax.get_window_extent().transformed(self.fig.dpi_scale_trans.inverted()) width, height = bbox.width, bbox.height width *= self.fig.dpi height *= self.fig.dpi dx = self.pick_radius*(xmax-xmin)/width dy = self.pick_radius*(ymax-ymin)/height x0 = erelease.xdata - dx x1 = erelease.xdata + dx y0 = erelease.ydata - dy y1 = erelease.ydata + dy elif self.select_zooms: self.zoom_stack.append((ax, xmin, xmax, ymin, ymax)) ax.set_xlim(x0, x1) ax.set_ylim(y0, y1) self._make_selection(ax, erelease.key, x0, x1, y0, y1) self._update_selection() def _on_pick(self, event): """Handle pick events on waveforms.""" for ax in self.wave_ax: for k, l in enumerate(ax.lines): if l is event.artist: self.mark_data = [self.mark_data[k]] self._update_selection() if event.mouseevent.dblclick: if len(self.mark_data) > 0: self.analyze_selection(self.mark_data[-1]) def _set_layout(self, width, height): """Update positions and visibility of all plots.""" xoffs = self.xborder/width yoffs = self.yborder/height xs = self.spacing/width ys = self.spacing/height if self.maxcols > 0: dx = (1.0-xoffs)/self.maxcols dy = (1.0-yoffs)/self.maxcols xw = dx - xs yw = dy - ys # histograms: for c, ax in enumerate(self.hist_ax): if c < self.maxcols: ax.set_position([xoffs+c*dx, yoffs, xw, yw]) ax.set_visible(True) else: ax.set_visible(False) ax.set_position([0.99, 0.01, 0.01, 0.01]) # scatter plots: for ax, (c, r) in zip(self.scatter_ax[:-1], self.scatter_indices[:-1]): if r < self.maxcols: ax.set_position([xoffs+c*dx, yoffs+(self.maxcols-r)*dy, xw, yw]) ax.set_visible(True) else: ax.set_visible(False) ax.set_position([0.99, 0.01, 0.01, 0.01]) # color bar: if self.maxcols > 0: self.cbax.set_position([xoffs+dx, yoffs+(self.maxcols-1)*dy, 0.3*xoffs, yw]) self.cbax.set_visible(True) else: self.cbax.set_visible(False) self.cbax.set_position([0.99, 0.01, 0.01, 0.01]) # magnified plot: if self.maxcols > 0: self._set_magnified_pos(width, height) if self.magnified_backdrop is not None: bbox = self.scatter_ax[-1].get_tightbbox(self.fig.canvas.get_renderer()) if bbox is not None: self.magnified_backdrop.set_bounds(bbox.x0, bbox.y0, bbox.width, bbox.height) else: self.scatter_ax[-1].set_position([0.5, 0.9, 0.05, 0.05]) self.scatter_ax[-1].set_visible(False) # waveform plots: if len(self.wave_ax) > 0: if self.maxcols > 0: x0 = xoffs+((self.maxcols+1)//2)*dx y0 = ((self.maxcols+1)//2)*dy if self.maxcols%2 == 0: x0 += xoffs y0 += yoffs - ys else: y0 += ys else: x0 = xoffs y0 = 0.0 yp = 1.0 dy = 1.0-y0 dy -= np.sum(self.wave_has_xticks)*yoffs yp -= ys dy -= ys if self.wave_title: yp -= 2*ys dy -= 2*ys dy /= len(self.wave_ax) for ax, has_xticks in zip(self.wave_ax, self.wave_has_xticks): yp -= dy ax.set_position([x0, yp, 1.0-x0-xs, dy]) if has_xticks: yp -= yoffs else: yp -= ys def _update_layout(self): """Update content and position of magnified plot.""" if self.scatter_indices[-1][1] < self.data.shape[1]: if self.scatter_indices[-1][1] >= self.maxcols: self.scatter_indices[-1][1] = self.maxcols-1 if self.scatter_indices[-1][0] >= self.scatter_indices[-1][1]: self.scatter_indices[-1][0] = self.scatter_indices[-1][1]-1 self._plot_scatter(self.scatter_ax[-1], True, False) else: if self.scatter_indices[-1][0] >= self.maxcols: self.scatter_indices[-1][0] = self.maxcols-1 self._plot_hist(self.scatter_ax[-1], True, False) self._set_layout(self.fig.get_window_extent().width, self.fig.get_window_extent().height) self.fig.canvas.draw() def _on_resize(self, event): """Adapt layout of plots to new figure size.""" self._set_layout(event.width, event.height)Subclasses
Class variables
var mouse_actionsvar key_actions
Methods
def set_wave_data(self, data, xlabels='', ylabels=[], title=False)-
Add waveform data to explorer.
Parameters
data:listof(list of) 2D arrays- Waveform data associated with each row of the data. Elements of the outer list correspond to the rows of the data. The inner 2D arrays contain a common x-axes (first column) and one or more corresponding y-values (second and optional higher columns). Each column for y-values is plotted in its own axes on top of each other, from top to bottom. The optional inner list of 2D arrays contains several 2D arrays as ascribed above each with its own common x-axes.
xlabel:stringorlistofstrings- The xlabels for the waveform plots. If only a string is given, then there will be a common xaxis for all the plots, and only the lowest one gets a labeled xaxis. If a list of strings is given, each waveform plot gets its own labeled x-axis.
ylabels:listofstrings- The ylabels for each of the waveform plots.
title:boolorstring- If True or a string, povide space on top of the waveform plots for a title. If string, set this as the title for the waveform plots.
Expand source code
def set_wave_data(self, data, xlabels='', ylabels=[], title=False): """Add waveform data to explorer. Parameters ---------- data: list of (list of) 2D arrays Waveform data associated with each row of the data. Elements of the outer list correspond to the rows of the data. The inner 2D arrays contain a common x-axes (first column) and one or more corresponding y-values (second and optional higher columns). Each column for y-values is plotted in its own axes on top of each other, from top to bottom. The optional inner list of 2D arrays contains several 2D arrays as ascribed above each with its own common x-axes. xlabel: string or list of strings The xlabels for the waveform plots. If only a string is given, then there will be a common xaxis for all the plots, and only the lowest one gets a labeled xaxis. If a list of strings is given, each waveform plot gets its own labeled x-axis. ylabels: list of strings The ylabels for each of the waveform plots. title: bool or string If True or a string, povide space on top of the waveform plots for a title. If string, set this as the title for the waveform plots. """ self.wave_data = [] if data is not None and len(data) > 0: self.wave_data = data self.wave_has_xticks = [] self.wave_nested = isinstance(data[0], (list, tuple)) if self.wave_nested: for data in self.wave_data[0]: for k in range(data.shape[1]-2): self.wave_has_xticks.append(False) self.wave_has_xticks.append(True) else: for k in range(self.wave_data[0].shape[1]-2): self.wave_has_xticks.append(False) self.wave_has_xticks.append(True) if isinstance(xlabels, (list, tuple)): self.wave_xlabels = xlabels else: self.wave_xlabels = [xlabels] self.wave_ylabels = ylabels self.wave_title = title self.wave_ax = [] def set_colors(self, colors=0, color_label=None, color_map=None)-
Set data column used to color scatter plots.
Parameters
colors:intor1D array- Index to colum in data to be used for coloring scatter plots.
- -2 for coloring row index of data.
- Or data array used to color scalar plots.
color_label:string- If colors is an array, this is a label describing the data.
- It is used to label the color bar.
color_map:stringorNone- Name of a matplotlib color map. If None 'jet' is used.
Expand source code
def set_colors(self, colors=0, color_label=None, color_map=None): """Set data column used to color scatter plots. Parameters ---------- colors: int or 1D array Index to colum in data to be used for coloring scatter plots. -2 for coloring row index of data. Or data array used to color scalar plots. color_label: string If colors is an array, this is a label describing the data. It is used to label the color bar. color_map: string or None Name of a matplotlib color map. If None 'jet' is used. """ if isinstance(colors, (np.integer, int)): if colors < 0: self.color_set_index = -1 self.color_index = 0 else: self.color_set_index = 0 self.color_index = colors else: if not isinstance(colors[0], (int, float, np.integer, np.floating)): # categorial data: self.extra_categories, self.extra_colors = categorize(colors) else: self.extra_colors = colors self.extra_color_label = color_label self.color_set_index = -1 self.color_index = 1 self.color_map = plt.get_cmap(color_map if color_map else 'jet') def show(self)-
Show interactive scatter plots for exploration.
Expand source code
def show(self): """Show interactive scatter plots for exploration. """ plt.ioff() plt.rcParams['toolbar'] = 'None' plt.rcParams['keymap.quit'] = 'ctrl+w, alt+q, q' self.fig = plt.figure(facecolor='white') self.fig.canvas.manager.set_window_title(self.title + ': ' + self.all_titles[self.show_mode]) self.fig.canvas.mpl_connect('key_press_event', self._on_key) self.fig.canvas.mpl_connect('resize_event', self._on_resize) self.fig.canvas.mpl_connect('pick_event', self._on_pick) if self.color_map is None: self.color_map = plt.get_cmap('jet') self._set_color_column() self._init_hist_plots() self._init_scatter_plots() self.wave_ax = [] if self.wave_data is not None and len(self.wave_data) > 0: axx = None xi = 0 for k, has_xticks in enumerate(self.wave_has_xticks): ax = self.fig.add_subplot(1, len(self.wave_has_xticks), 1+k, sharex=axx) self.wave_ax.append(ax) if has_xticks: if xi >= len(self.wave_xlabels): self.wave_xlabels.append('') ax.set_xlabel(self.wave_xlabels[xi]) xi += 1 axx = None else: plt.setp(ax.get_xticklabels(), visible=False) if axx is None: axx = ax for ax, ylabel in zip(self.wave_ax, self.wave_ylabels): ax.set_ylabel(ylabel) if not isinstance(self.wave_title, bool) and self.wave_title: self.wave_ax[0].set_title(self.wave_title) self.fix_waveform_plot(self.wave_ax, self.mark_data) self._plot_magnified_scatter() plt.show() def compute_pca(self, scale=False, write=False)-
Compute PCA based on the data.
Parameters
scale:boolean- If True standardize data before computing PCA, i.e. remove mean of each variabel and divide by its standard deviation.
write:boolean- If True write PCA components to standard out.
Expand source code
def compute_pca(self, scale=False, write=False): """Compute PCA based on the data. Parameters ---------- scale: boolean If True standardize data before computing PCA, i.e. remove mean of each variabel and divide by its standard deviation. write: boolean If True write PCA components to standard out. """ # select columns without NANs: idxs = [i for i in range(self.raw_data.shape[1]) if np.all(np.isfinite(self.raw_data[:,i]))] data = self.raw_data[:,idxs] # pca: pca = decomposition.PCA() if scale: scaler = preprocessing.StandardScaler() scaler.fit(data) pca.fit(scaler.transform(data)) pca_label = 'sPC' else: pca.fit(data) pca_label = 'PC' for k in range(len(pca.components_)): if np.abs(np.min(pca.components_[k])) > np.max(pca.components_[k]): pca.components_[k] *= -1.0 pca_data = pca.transform(data) pca_labels = [('%s%d (%.1f%%)' if v > 0.01 else '%s%d (%.2f%%)') % (pca_label, k+1, 100.0*v) for k, v in enumerate(pca.explained_variance_ratio_)] if np.min(pca.explained_variance_ratio_) >= 0.01: pca_maxcols = pca_data.shape[1] else: pca_maxcols = np.argmax(pca.explained_variance_ratio_ < 0.01) if pca_maxcols < 2: pca_maxcols = 2 if pca_maxcols > 6: pca_maxcols = 6 # table with PCA feature weights: pca_table = self.pca_tables[1] if scale else self.pca_tables[0] pca_table.clear_data() pca_table.set_section(pca_label, 0, pca_table.nsecs) for k, comp in enumerate(pca.components_): pca_table.append_data(k+1, 0) pca_table.append_data(100.0*pca.explained_variance_ratio_[k]) pca_table.append_data(comp) if write: pca_table.write(table_format='out', unit_style='none') # submit data: if scale: self.all_data[2] = pca_data self.all_labels[2] = pca_labels self.all_maxcols[2] = pca_maxcols else: self.all_data[1] = pca_data self.all_labels[1] = pca_labels self.all_maxcols[1] = pca_maxcols def save_pca(self, file_name, scale, **kwargs)-
Write PCA data to file.
Parameters
file_name:string- Name of ouput file.
scale:boolean- If True write PCA components of standardized PCA.
kwargs:dict- Additional parameter for TableData.write()
Expand source code
def save_pca(self, file_name, scale, **kwargs): """Write PCA data to file. Parameters ---------- file_name: string Name of ouput file. scale: boolean If True write PCA components of standardized PCA. kwargs: dict Additional parameter for TableData.write() """ if scale: pca_file = file_name + '-pcacor' pca_table = self.pca_tables[1] else: pca_file = file_name + '-pcacov' pca_table = self.pca_tables[0] if 'unit_style' in kwargs: del kwargs['unit_style'] if 'table_format' in kwargs: pca_table.write(pca_file, unit_style='none', **kwargs) else: pca_file += '.dat' pca_table.write(pca_file, unit_style='none') def fix_scatter_plot(self, ax, data, label, axis)-
Customize an axes of a scatter plot.
This function is called after a scatter plot has been plotted. Once for the x axes, once for the y axis and once for the color bar. Reimplement this function to set appropriate limits and ticks.
Return values are only used for the color bar (
axis='c'). Otherwise they are ignored.For example, ticks for phase variables can be nicely labeled using the unicode character for pi:
if 'phase' in label: if axis == 'y': ax.set_ylim(0.0, 2.0*np.pi) ax.set_yticks(np.arange(0.0, 2.5*np.pi, 0.5*np.pi)) ax.set_yticklabels(['0', u'π/2', u'π', u'3π/2', u'2π'])Parameters
ax:matplotlib axes- Axes of the scatter plot or color bar to be worked on.
data:1D array- Data array of the axes.
label:string- Label coresponding to the data array.
axis:str- 'x', 'y': set properties of x or y axes of ax. 'c': set properies of color bar axes (note that ax can be None!) and return vmin, vmax, and ticks.
Returns
min: float minimum value of color bar axis max: float maximum value of color bar axis ticks: list of float position of ticks for color bar axisExpand source code
def fix_scatter_plot(self, ax, data, label, axis): """Customize an axes of a scatter plot. This function is called after a scatter plot has been plotted. Once for the x axes, once for the y axis and once for the color bar. Reimplement this function to set appropriate limits and ticks. Return values are only used for the color bar (`axis='c'`). Otherwise they are ignored. For example, ticks for phase variables can be nicely labeled using the unicode character for pi: ``` if 'phase' in label: if axis == 'y': ax.set_ylim(0.0, 2.0*np.pi) ax.set_yticks(np.arange(0.0, 2.5*np.pi, 0.5*np.pi)) ax.set_yticklabels(['0', u'\u03c0/2', u'\u03c0', u'3\u03c0/2', u'2\u03c0']) ``` Parameters ---------- ax: matplotlib axes Axes of the scatter plot or color bar to be worked on. data: 1D array Data array of the axes. label: string Label coresponding to the data array. axis: str 'x', 'y': set properties of x or y axes of ax. 'c': set properies of color bar axes (note that ax can be None!) and return vmin, vmax, and ticks. Returns ------- min: float minimum value of color bar axis max: float maximum value of color bar axis ticks: list of float position of ticks for color bar axis """ return np.nanmin(data), np.nanmax(data), None def fix_waveform_plot(self, axs, indices)-
Customize waveform plots.
This function is called once after new data have been plotted into the waveform plots. Reimplement this function to customize these plots. In particular to set axis limits and labels, plot title, etc. You may even open a new figure (with non-blocking
show()).The following member variables might be usefull: -
self.wave_data: the full list of waveform data. -self.wave_nested: True if the elements ofself.wave_dataare lists of 2D arrays. Otherwise the elements are 2D arrays. The first column of a 2D array contains the x-values, further columns y-values. -self.wave_has_xticks: List of booleans for each axis. True if the axis has its own xticks. -self.wave_xlabels: List of xlabels (only for the axis where the corresponding entry inself.wave_has_xticksis True). -self.wave_ylabels: for each axis its ylabelFor example, you can set the linewidth of all plotted waveforms via:
for ax in axs: for l in ax.lines: l.set_linewidth(3.0)or enable markers to be plotted:
for ax, yl in zip(axs, self.wave_ylabels): if 'Power' in yl: for l in ax.lines: l.set_marker('.') l.set_markersize(15.0) l.set_markeredgewidth(0.5) l.set_markeredgecolor('k') l.set_markerfacecolor(l.get_color())Usefull is to reduce the maximum number of y-ticks:
axs[0].yaxis.get_major_locator().set_params(nbins=7)or
import matplotlib.ticker as ticker axs[0].yaxis.set_major_locator(ticker.MaxNLocator(nbins=4))Parameters
axs:listofmatplotlib axes- Axis of the waveform plots to be worked on.
indices:listofint- Indices of the waveforms that have been selected and plotted.
Expand source code
def fix_waveform_plot(self, axs, indices): """Customize waveform plots. This function is called once after new data have been plotted into the waveform plots. Reimplement this function to customize these plots. In particular to set axis limits and labels, plot title, etc. You may even open a new figure (with non-blocking `show()`). The following member variables might be usefull: - `self.wave_data`: the full list of waveform data. - `self.wave_nested`: True if the elements of `self.wave_data` are lists of 2D arrays. Otherwise the elements are 2D arrays. The first column of a 2D array contains the x-values, further columns y-values. - `self.wave_has_xticks`: List of booleans for each axis. True if the axis has its own xticks. - `self.wave_xlabels`: List of xlabels (only for the axis where the corresponding entry in `self.wave_has_xticks` is True). - `self.wave_ylabels`: for each axis its ylabel For example, you can set the linewidth of all plotted waveforms via: ``` for ax in axs: for l in ax.lines: l.set_linewidth(3.0) ``` or enable markers to be plotted: ``` for ax, yl in zip(axs, self.wave_ylabels): if 'Power' in yl: for l in ax.lines: l.set_marker('.') l.set_markersize(15.0) l.set_markeredgewidth(0.5) l.set_markeredgecolor('k') l.set_markerfacecolor(l.get_color()) ``` Usefull is to reduce the maximum number of y-ticks: ``` axs[0].yaxis.get_major_locator().set_params(nbins=7) ``` or ``` import matplotlib.ticker as ticker axs[0].yaxis.set_major_locator(ticker.MaxNLocator(nbins=4)) ``` Parameters ---------- axs: list of matplotlib axes Axis of the waveform plots to be worked on. indices: list of int Indices of the waveforms that have been selected and plotted. """ pass def list_selection(self, indices)-
List information about the current selection of data points.
This function is called when 'l' is pressed. Reimplement this function, for example, to print some meaningfull information about the current selection of data points on console. You nay do, however, whatever you want in this function.
Parameters
indices:listofint- Indices of the data points that have been selected.
Expand source code
def list_selection(self, indices): """List information about the current selection of data points. This function is called when 'l' is pressed. Reimplement this function, for example, to print some meaningfull information about the current selection of data points on console. You nay do, however, whatever you want in this function. Parameters ---------- indices: list of int Indices of the data points that have been selected. """ for i in indices: print(i) def analyze_selection(self, index)-
Provide further information about a single selected data point.
This function is called when a single data item was double clicked. Reimplement this function to provide some further details on this data point. This can be an additional figure window. In this case show it non-blocking:
plt.show(block=False)Parameters
index:int- The index of the selected data point.
Expand source code
def analyze_selection(self, index): """Provide further information about a single selected data point. This function is called when a single data item was double clicked. Reimplement this function to provide some further details on this data point. This can be an additional figure window. In this case show it non-blocking: `plt.show(block=False)` Parameters ---------- index: int The index of the selected data point. """ pass