Module thunderfish.pulseplots
Plot and save key steps in pulses.py for visualizing the alorithm.
Expand source code
"""
Plot and save key steps in pulses.py for visualizing the alorithm.
"""
import glob
import numpy as np
from scipy import stats
from matplotlib import gridspec, ticker
import matplotlib.pyplot as plt
try:
from matplotlib.colors import colorConverter as cc
except ImportError:
import matplotlib.colors as cc
try:
from matplotlib.colors import to_hex
except ImportError:
from matplotlib.colors import rgb2hex as to_hex
from matplotlib.patches import ConnectionPatch, Rectangle
from matplotlib.lines import Line2D
import warnings
def warn(*args, **kwargs):
"""
Ignore all warnings.
"""
pass
warnings.warn=warn
# plotting parameters and colors:
cmap = plt.get_cmap("Dark2")
c_g = cmap(0)
c_o = cmap(1)
c_grey = cmap(7)
cmap_pts = [cmap(2), cmap(3)]
def darker(color, saturation):
""" Make a color darker.
From bendalab/plottools package.
Parameters
----------
color: dict or matplotlib color spec
A matplotlib color (hex string, name color string, rgb tuple)
or a dictionary with an 'color' or 'facecolor' key.
saturation: float
The smaller the saturation, the darker the returned color.
A saturation of 0 returns black.
A saturation of 1 leaves the color untouched.
A saturation of 2 returns white.
Returns
-------
color: string or dictionary
The darker color as a hexadecimal RGB string (e.g. '#rrggbb').
If `color` is a dictionary, a copy of the dictionary is returned
with the value of 'color' or 'facecolor' set to the darker color.
"""
try:
c = color['color']
cd = dict(**color)
cd['color'] = darker(c, saturation)
return cd
except (KeyError, TypeError):
try:
c = color['facecolor']
cd = dict(**color)
cd['facecolor'] = darker(c, saturation)
return cd
except (KeyError, TypeError):
if saturation > 2:
sauration = 2
if saturation > 1:
return lighter(color, 2.0-saturation)
if saturation < 0:
saturation = 0
r, g, b = cc.to_rgb(color)
rd = r*saturation
gd = g*saturation
bd = b*saturation
return to_hex((rd, gd, bd)).upper()
def lighter(color, lightness):
"""Make a color lighter
From bendalab/plottools package.
Parameters
----------
color: dict or matplotlib color spec
A matplotlib color (hex string, name color string, rgb tuple)
or a dictionary with an 'color' or 'facecolor' key.
lightness: float
The smaller the lightness, the lighter the returned color.
A lightness of 0 returns white.
A lightness of 1 leaves the color untouched.
A lightness of 2 returns black.
Returns
-------
color: string or dict
The lighter color as a hexadecimal RGB string (e.g. '#rrggbb').
If `color` is a dictionary, a copy of the dictionary is returned
with the value of 'color' or 'facecolor' set to the lighter color.
"""
try:
c = color['color']
cd = dict(**color)
cd['color'] = lighter(c, lightness)
return cd
except (KeyError, TypeError):
try:
c = color['facecolor']
cd = dict(**color)
cd['facecolor'] = lighter(c, lightness)
return cd
except (KeyError, TypeError):
if lightness > 2:
lightness = 2
if lightness > 1:
return darker(color, 2.0-lightness)
if lightness < 0:
lightness = 0
r, g, b = cc.to_rgb(color)
rl = r + (1.0-lightness)*(1.0 - r)
gl = g + (1.0-lightness)*(1.0 - g)
bl = b + (1.0-lightness)*(1.0 - b)
return to_hex((rl, gl, bl)).upper()
def xscalebar(ax, x, y, width, wunit=None, wformat=None, ha='left', va='bottom',
lw=None, color=None, capsize=None, clw=None, **kwargs):
"""Horizontal scale bar with label.
From bendalab/plottools package.
Parameters
----------
ax: matplotlib axes
Axes where to draw the scale bar.
x: float
x-coordinate where to draw the scale bar in relative units of the axes.
y: float
y-coordinate where to draw the scale bar in relative units of the axes.
width: float
Length of the scale bar in units of the data's x-values.
wunit: string or None
Optional unit of the data's x-values.
wformat: string or None
Optional format string for formatting the label of the scale bar
or simply a string used for labeling the scale bar.
ha: 'left', 'right', or 'center'
Scale bar aligned left, right, or centered to (x, y)
va: 'top' or 'bottom'
Label of the scale bar either above or below the scale bar.
lw: int, float, None
Line width of the scale bar.
color: matplotlib color
Color of the scalebar.
capsize: float or None
If larger then zero draw cap lines at the ends of the bar.
The length of the lines is given in points (same unit as linewidth).
clw: int, float, None
Line width of the cap lines.
kwargs: key-word arguments
Passed on to `ax.text()` used to print the scale bar label.
"""
ax.autoscale(False)
# ax dimensions:
pixelx = np.abs(np.diff(ax.get_window_extent().get_points()[:,0]))[0]
pixely = np.abs(np.diff(ax.get_window_extent().get_points()[:,1]))[0]
xmin, xmax = ax.get_xlim()
ymin, ymax = ax.get_ylim()
unitx = xmax - xmin
unity = ymax - ymin
dxu = np.abs(unitx)/pixelx
dyu = np.abs(unity)/pixely
# transform x, y from relative units to axis units:
x = xmin + x*unitx
y = ymin + y*unity
# bar length:
if wformat is None:
wformat = '%.0f'
if width < 1.0:
wformat = '%.1f'
try:
ls = wformat % width
width = float(ls)
except TypeError:
ls = wformat
# bar:
if ha == 'left':
x0 = x
x1 = x+width
elif ha == 'right':
x0 = x-width
x1 = x
else:
x0 = x-0.5*width
x1 = x+0.5*width
# line width:
if lw is None:
lw = 2
# color:
if color is None:
color = 'k'
# scalebar:
lh = ax.plot([x0, x1], [y, y], '-', color=color, lw=lw,
solid_capstyle='butt', clip_on=False)
# get y position of line in figure pixel coordinates:
ly = np.array(lh[0].get_window_extent(ax.get_figure().canvas.get_renderer()))[0,1]
# caps:
if capsize is None:
capsize = 0
if clw is None:
clw = 0.5
if capsize > 0.0:
dy = capsize*dyu
ax.plot([x0, x0], [y-dy, y+dy], '-', color=color, lw=clw,
solid_capstyle='butt', clip_on=False)
ax.plot([x1, x1], [y-dy, y+dy], '-', color=color, lw=clw,
solid_capstyle='butt', clip_on=False)
# label:
if wunit:
ls += u'\u2009%s' % wunit
if va == 'top':
th = ax.text(0.5*(x0+x1), y, ls, clip_on=False,
ha='center', va='bottom', **kwargs)
# get y coordinate of text bottom in figure pixel coordinates:
ty = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[0,1]
dty = ly+0.5*lw + 2.0 - ty
else:
th = ax.text(0.5*(x0+x1), y, ls, clip_on=False,
ha='center', va='top', **kwargs)
# get y coordinate of text bottom in figure pixel coordinates:
ty = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[1,1]
dty = ly-0.5*lw - 2.0 - ty
th.set_position((0.5*(x0+x1), y+dyu*dty))
return x0, x1, y
def yscalebar(ax, x, y, height, hunit=None, hformat=None, ha='left', va='bottom',
lw=None, color=None, capsize=None, clw=None, **kwargs):
"""Vertical scale bar with label.
From bendalab/plottools package.
Parameters
----------
ax: matplotlib axes
Axes where to draw the scale bar.
x: float
x-coordinate where to draw the scale bar in relative units of the axes.
y: float
y-coordinate where to draw the scale bar in relative units of the axes.
height: float
Length of the scale bar in units of the data's y-values.
hunit: string
Unit of the data's y-values.
hformat: string or None
Optional format string for formatting the label of the scale bar
or simply a string used for labeling the scale bar.
ha: 'left' or 'right'
Label of the scale bar either to the left or to the right
of the scale bar.
va: 'top', 'bottom', or 'center'
Scale bar aligned above, below, or centered on (x, y).
lw: int, float, None
Line width of the scale bar.
color: matplotlib color
Color of the scalebar.
capsize: float or None
If larger then zero draw cap lines at the ends of the bar.
The length of the lines is given in points (same unit as linewidth).
clw: int, float
Line width of the cap lines.
kwargs: key-word arguments
Passed on to `ax.text()` used to print the scale bar label.
"""
ax.autoscale(False)
# ax dimensions:
pixelx = np.abs(np.diff(ax.get_window_extent().get_points()[:,0]))[0]
pixely = np.abs(np.diff(ax.get_window_extent().get_points()[:,1]))[0]
xmin, xmax = ax.get_xlim()
ymin, ymax = ax.get_ylim()
unitx = xmax - xmin
unity = ymax - ymin
dxu = np.abs(unitx)/pixelx
dyu = np.abs(unity)/pixely
# transform x, y from relative units to axis units:
x = xmin + x*unitx
y = ymin + y*unity
# bar length:
if hformat is None:
hformat = '%.0f'
if height < 1.0:
hformat = '%.1f'
try:
ls = hformat % height
width = float(ls)
except TypeError:
ls = hformat
# bar:
if va == 'bottom':
y0 = y
y1 = y+height
elif va == 'top':
y0 = y-height
y1 = y
else:
y0 = y-0.5*height
y1 = y+0.5*height
# line width:
if lw is None:
lw = 2
# color:
if color is None:
color = 'k'
# scalebar:
lh = ax.plot([x, x], [y0, y1], '-', color=color, lw=lw,
solid_capstyle='butt', clip_on=False)
# get x position of line in figure pixel coordinates:
lx = np.array(lh[0].get_window_extent(ax.get_figure().canvas.get_renderer()))[0,0]
# caps:
if capsize is None:
capsize = 0
if clw is None:
clw = 0.5
if capsize > 0.0:
dx = capsize*dxu
ax.plot([x-dx, x+dx], [y0, y0], '-', color=color, lw=clw, solid_capstyle='butt',
clip_on=False)
ax.plot([x-dx, x+dx], [y1, y1], '-', color=color, lw=clw, solid_capstyle='butt',
clip_on=False)
# label:
if hunit:
ls += u'\u2009%s' % hunit
if ha == 'right':
th = ax.text(x, 0.5*(y0+y1), ls, clip_on=False, rotation=90.0,
ha='left', va='center', **kwargs)
# get x coordinate of text bottom in figure pixel coordinates:
tx = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[0,0]
dtx = lx+0.5*lw + 2.0 - tx
else:
th = ax.text(x, 0.5*(y0+y1), ls, clip_on=False, rotation=90.0,
ha='right', va='center', **kwargs)
# get x coordinate of text bottom in figure pixel coordinates:
tx = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[1,0]
dtx = lx-0.5*lw - 1.0 - tx
th.set_position((x+dxu*dtx, 0.5*(y0+y1)))
return x, y0, y1
def arrowed_spines(ax, ms=10):
""" Spine with arrow on the y-axis of a plot.
Parameters
----------
ax : matplotlib figure axis
Axis on which the arrow should be plot.
"""
xmin, xmax = ax.get_xlim()
ymin, ymax = ax.get_ylim()
ax.scatter([xmin], [ymax], s=ms, marker='^', clip_on=False, color='k')
ax.set_xlim(xmin, xmax)
ax.set_ylim(ymin, ymax)
def loghist(ax, x, bmin, bmax, n, c, orientation='vertical', label=''):
""" Plot histogram with logarithmic scale.
Parameters
----------
ax : matplotlib axis
Axis to plot the histogram on.
x : numpy array
Input data for histogram.
bmin : float
Minimum value for the histogram bins.
bmax : float
Maximum value for the histogram bins.
n : int
Number of bins.
c : matplotlib color
Color of histogram.
orientation : string (optional)
Histogram orientation.
Defaults to 'vertical'.
label : string (optional)
Label for x.
Defaults to '' (no label).
Returns
-------
n : array
The values of the histogram bins.
bins : array
The edges of the bins.
patches : BarContainer
Container of individual artists used to create the histogram.
"""
return ax.hist(x, bins=np.exp(np.linspace(np.log(bmin), np.log(bmax), n)),
color=c, orientation=orientation, label=label)
def plot_all(data, eod_p_times, eod_tr_times, fs, mean_eods):
"""Quick way to view the output of extract_pulsefish in a single plot.
Parameters
----------
data: array
Recording data.
eod_p_times: array of ints
EOD peak indices.
eod_tr_times: array of ints
EOD trough indices.
fs: float
Samplerate.
mean_eods: list of numpy arrays
Mean EODs of each pulsefish found in the recording.
"""
fig = plt.figure(figsize=(10, 5))
if len(eod_p_times) > 0:
gs = gridspec.GridSpec(2, len(eod_p_times))
ax = fig.add_subplot(gs[0,:])
ax.plot(np.arange(len(data))/fs, data, c='k', alpha=0.3)
for i, (pt, tt) in enumerate(zip(eod_p_times, eod_tr_times)):
ax.plot(pt, data[(pt*fs).astype('int')], 'o', label=i+1, ms=10, c=cmap(i))
ax.plot(tt, data[(tt*fs).astype('int')], 'o', label=i+1, ms=10, c=cmap(i))
ax.set_xlabel('time [s]')
ax.set_ylabel('amplitude [V]')
for i, m in enumerate(mean_eods):
ax = fig.add_subplot(gs[1,i])
ax.plot(1000*m[0], 1000*m[1], c='k')
ax.fill_between(1000*m[0], 1000*(m[1]-m[2]), 1000*(m[1]+m[2]), color=cmap(i))
ax.set_xlabel('time [ms]')
ax.set_ylabel('amplitude [mV]')
else:
plt.plot(np.arange(len(data))/fs, data, c='k', alpha=0.3)
plt.tight_layout()
def plot_clustering(samplerate, eod_widths, eod_hights, eod_shapes, disc_masks, merge_masks):
"""Plot all clustering steps.
Plot clustering steps on width, height and shape. Then plot the remaining EODs after
the EOD assessment step and the EODs after the merge step.
Parameters
----------
samplerate : float
Samplerate of EOD snippets.
eod_widths : list of three 1D numpy arrays
The first list entry gives the unique labels of all width clusters as a list of ints.
The second list entry gives the width values for each EOD in samples as a
1D numpy array of ints.
The third list entry gives the width labels for each EOD as a 1D numpy array of ints.
eod_hights : nested lists (2 layers) of three 1D numpy arrays
The first list entry gives the unique labels of all height clusters as a list of ints
for each width cluster.
The second list entry gives the height values for each EOD as a 1D numpy array
of floats for each width cluster.
The third list entry gives the height labels for each EOD as a 1D numpy array
of ints for each width cluster.
eod_shapes : nested lists (3 layers) of three 1D numpy arrays
The first list entry gives the raw EOD snippets as a 2D numpy array for each
height cluster in a width cluster.
The second list entry gives the snippet PCA values for each EOD as a 2D numpy array
of floats for each height cluster in a width cluster.
The third list entry gives the shape labels for each EOD as a 1D numpy array of ints
for each height cluster in a width cluster.
disc_masks : Nested lists (two layers) of 1D numpy arrays
The masks of EODs that are discarded by the discarding step of the algorithm.
The masks are 1D boolean arrays where
instances that are set to True are discarded by the algorithm. Discarding masks
are saved in nested lists that represent the width and height clusters.
merge_masks : Nested lists (two layers) of 2D numpy arrays
The masks of EODs that are discarded by the merging step of the algorithm.
The masks are 2D boolean arrays where
for each sample point `i` either `merge_mask[i,0]` or `merge_mask[i,1]` is set to True.
Here, merge_mask[:,0] represents the
peak-centered clusters and `merge_mask[:,1]` represents the trough-centered clusters.
Merge masks are saved in nested lists
that represent the width and height clusters.
"""
# create figure + transparant figure.
fig = plt.figure(figsize=(12, 7))
transFigure = fig.transFigure.inverted()
# set up the figure layout
outer = gridspec.GridSpec(1, 5, width_ratios=[1, 1, 2, 1, 2], left=0.05, right=0.95)
# set titles for each clustering step
titles = ['1. Widths', '2. Heights', '3. Shape', '4. Pulse EODs', '5. Merge']
for i, title in enumerate(titles):
title_ax = gridspec.GridSpecFromSubplotSpec(1, 1, subplot_spec = outer[i])
ax = fig.add_subplot(title_ax[0])
ax.text(0, 110, title, ha='center', va='bottom', clip_on=False)
ax.set_xlim(-100, 100)
ax.set_ylim(-100, 100)
ax.axis('off')
# compute sizes for each axis
w_size = 1
h_size = len(eod_hights[1])
shape_size = np.sum([len(sl) for sl in eod_shapes[0]])
# count required axes sized for the last two plot columns.
disc_size = 0
merge_size= 0
for shapelabel, dmasks, mmasks in zip(eod_shapes[2], disc_masks, merge_masks):
for sl, dm, mm in zip(shapelabel, dmasks, mmasks):
uld1 = np.unique((sl[0]+1)*np.invert(dm[0]))
uld2 = np.unique((sl[1]+1)*np.invert(dm[1]))
disc_size = disc_size+len(uld1[uld1>0])+len(uld2[uld2>0])
uld1 = np.unique((sl[0]+1)*mm[0])
uld2 = np.unique((sl[1]+1)*mm[1])
merge_size = merge_size+len(uld1[uld1>0])+len(uld2[uld2>0])
# set counters to keep track of the plot axes
disc_block = 0
merge_block = 0
shape_count = 0
# create all axes
width_hist_ax = gridspec.GridSpecFromSubplotSpec(w_size, 1, subplot_spec = outer[0])
hight_hist_ax = gridspec.GridSpecFromSubplotSpec(h_size, 1, subplot_spec = outer[1])
shape_ax = gridspec.GridSpecFromSubplotSpec(shape_size, 1, subplot_spec = outer[2])
shape_windows = [gridspec.GridSpecFromSubplotSpec(2, 2, hspace=0.0, wspace=0.0,
subplot_spec=shape_ax[i])
for i in range(shape_size)]
EOD_delete_ax = gridspec.GridSpecFromSubplotSpec(disc_size, 1, subplot_spec=outer[3])
EOD_merge_ax = gridspec.GridSpecFromSubplotSpec(merge_size, 1, subplot_spec=outer[4])
# plot width labels histogram
ax1 = fig.add_subplot(width_hist_ax[0])
# set axes features.
ax1.set_xscale('log')
ax1.spines['top'].set_visible(False)
ax1.spines['right'].set_visible(False)
ax1.spines['bottom'].set_visible(False)
ax1.axes.xaxis.set_visible(False)
ax1.set_yticklabels([])
# indices for plot colors (dark to light)
colidxsw = -np.linspace(-1.25, -0.5, h_size)
for i, (wl, colw, uhl, eod_h, eod_h_labs, w_snip, w_feat, w_lab, w_dm, w_mm) in enumerate(zip(eod_widths[0], colidxsw, eod_hights[0], eod_hights[1], eod_hights[2], eod_shapes[0], eod_shapes[1], eod_shapes[2], disc_masks, merge_masks)):
# plot width hist
hw, _, _ = ax1.hist(eod_widths[1][eod_widths[2]==wl],
bins=np.linspace(np.min(eod_widths[1]), np.max(eod_widths[1]), 100),
color=lighter(c_o, colw), orientation='horizontal')
# set arrow when the last hist is plot so the size of the axes are known.
if i == h_size-1:
arrowed_spines(ax1, ms=20)
# determine total size of the hight historgams now.
my, b = np.histogram(eod_h, bins=np.exp(np.linspace(np.min(np.log(eod_h)),
np.max(np.log(eod_h)), 100)))
maxy = np.max(my)
# set axes features for hight hist.
ax2 = fig.add_subplot(hight_hist_ax[h_size-i-1])
ax2.set_xscale('log')
ax2.spines['top'].set_visible(False)
ax2.spines['right'].set_visible(False)
ax2.spines['bottom'].set_visible(False)
ax2.set_xlim(0.9, maxy)
ax2.axes.xaxis.set_visible(False)
ax2.set_yscale('log')
ax2.yaxis.set_major_formatter(ticker.NullFormatter())
ax2.yaxis.set_minor_formatter(ticker.NullFormatter())
# define colors for plots
colidxsh = -np.linspace(-1.25, -0.5, len(uhl))
for n, (hl, hcol, snippets, features, labels, dmasks, mmasks) in enumerate(zip(uhl, colidxsh, w_snip, w_feat, w_lab, w_dm, w_mm)):
hh, _, _ = loghist(ax2, eod_h[eod_h_labs==hl], np.min(eod_h), np.max(eod_h), 100,
lighter(c_g, hcol), orientation='horizontal')
# set arrow spines only on last plot
if n == len(uhl)-1:
arrowed_spines(ax2, ms=10)
# plot line from the width histogram to the height histogram.
if n == 0:
coord1 = transFigure.transform(ax1.transData.transform([np.median(hw[hw!=0]),
np.median(eod_widths[1][eod_widths[2]==wl])]))
coord2 = transFigure.transform(ax2.transData.transform([0.9, np.mean(eod_h)]))
line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]),
transform=fig.transFigure, color='grey', linewidth=0.5)
fig.lines.append(line)
# compute sizes of the eod_discarding and merge steps
s1 = np.unique((labels[0]+1)*(~dmasks[0]))
s2 = np.unique((labels[1]+1)*(~dmasks[1]))
disc_block = disc_block + len(s1[s1>0]) + len(s2[s2>0])
s1 = np.unique((labels[0]+1)*(mmasks[0]))
s2 = np.unique((labels[1]+1)*(mmasks[1]))
merge_block = merge_block + len(s1[s1>0]) + len(s2[s2>0])
axs = []
disc_count = 0
merge_count = 0
# now plot the clusters for peak and trough centerings
for pt, cmap_pt in zip([0, 1], cmap_pts):
ax3 = fig.add_subplot(shape_windows[shape_size-1-shape_count][pt,0])
ax4 = fig.add_subplot(shape_windows[shape_size-1-shape_count][pt,1])
# remove axes
ax3.axes.xaxis.set_visible(False)
ax4.axes.yaxis.set_visible(False)
ax3.axes.yaxis.set_visible(False)
ax4.axes.xaxis.set_visible(False)
# set color indices
colidxss = -np.linspace(-1.25, -0.5, len(np.unique(labels[pt][labels[pt]>=0])))
j=0
for c in np.unique(labels[pt]):
if c<0:
# plot noise features + snippets
ax3.plot(features[pt][labels[pt]==c,0], features[pt][labels[pt]==c,1],
'.', color='lightgrey', label='-1', rasterized=True)
ax4.plot(snippets[pt][labels[pt]==c].T, linewidth=0.1,
color='lightgrey', label='-1', rasterized=True)
else:
# plot cluster features and snippets
ax3.plot(features[pt][labels[pt]==c,0], features[pt][labels[pt]==c,1],
'.', color=lighter(cmap_pt, colidxss[j]), label=c,
rasterized=True)
ax4.plot(snippets[pt][labels[pt]==c].T, linewidth=0.1,
color=lighter(cmap_pt, colidxss[j]), label=c, rasterized=True)
# check if the current cluster is an EOD, if yes, plot it.
if np.sum(dmasks[pt][labels[pt]==c]) == 0:
ax = fig.add_subplot(EOD_delete_ax[disc_size-disc_block+disc_count])
ax.axis('off')
# plot mean EOD snippet
ax.plot(np.mean(snippets[pt][labels[pt]==c], axis=0),
color=lighter(cmap_pt, colidxss[j]))
disc_count = disc_count + 1
# match colors and draw line..
coord1 = transFigure.transform(ax4.transData.transform([ax4.get_xlim()[1],
ax4.get_ylim()[0] + 0.5*(ax4.get_ylim()[1]-ax4.get_ylim()[0])]))
coord2 = transFigure.transform(ax.transData.transform([ax.get_xlim()[0],ax.get_ylim()[0] + 0.5*(ax.get_ylim()[1]-ax.get_ylim()[0])]))
line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]),
transform=fig.transFigure, color='grey',
linewidth=0.5)
fig.lines.append(line)
axs.append(ax)
# check if the current EOD survives the merge step
# if so, plot it.
if np.sum(mmasks[pt, labels[pt]==c])>0:
ax = fig.add_subplot(EOD_merge_ax[merge_size-merge_block+merge_count])
ax.axis('off')
ax.plot(np.mean(snippets[pt][labels[pt]==c], axis=0),
color=lighter(cmap_pt, colidxss[j]))
merge_count = merge_count + 1
j=j+1
if pt==0:
# draw line from hight cluster to EOD shape clusters.
coord1 = transFigure.transform(ax2.transData.transform([np.median(hh[hh!=0]),
np.median(eod_h[eod_h_labs==hl])]))
coord2 = transFigure.transform(ax3.transData.transform([ax3.get_xlim()[0],
ax3.get_ylim()[0]]))
line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]),
transform=fig.transFigure, color='grey', linewidth=0.5)
fig.lines.append(line)
shape_count = shape_count + 1
if len(axs)>0:
# plot lines that indicate the merged clusters.
coord1 = transFigure.transform(axs[0].transData.transform([axs[0].get_xlim()[1]+0.1*(axs[0].get_xlim()[1]-axs[0].get_xlim()[0]),
axs[0].get_ylim()[1]-0.25*(axs[0].get_ylim()[1]-axs[0].get_ylim()[0])]))
coord2 = transFigure.transform(axs[-1].transData.transform([axs[-1].get_xlim()[1]+0.1*(axs[-1].get_xlim()[1]-axs[-1].get_xlim()[0]),
axs[-1].get_ylim()[0]+0.25*(axs[-1].get_ylim()[1]-axs[-1].get_ylim()[0])]))
line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]),
transform=fig.transFigure, color='grey', linewidth=1)
fig.lines.append(line)
def plot_bgm(x, means, variances, weights, use_log, labels, labels_am, xlab):
"""Plot a BGM clustering step either on EOD width or height.
Parameters
----------
x : 1D numpy array of floats
BGM input values.
means : list of floats
BGM Gaussian means
variances : list of floats
BGM Gaussian variances.
weights : list of floats
BGM Gaussian weights.
use_log : boolean
True if the z-scored logarithm of the data was used as BGM input.
labels : 1D numpy array of ints
Labels defined by BGM model (before merging based on merge factor).
labels_am : 1D numpy array of ints
Labels defined by BGM model (after merging based on merge factor).
xlab : string
Label for plot (defines the units of the BGM data).
"""
if 'width' in xlab:
ccol = c_o
elif 'height' in xlab:
ccol = c_g
else:
ccol = 'b'
# get the transform that was used as BGM input
if use_log:
x_transform = stats.zscore(np.log(x))
xplot = np.exp(np.linspace(np.log(np.min(x)), np.log(np.max(x)), 1000))
else:
x_transform = stats.zscore(x)
xplot = np.linspace(np.min(x), np.max(x), 1000)
# compute the x values and gaussians
x2 = np.linspace(np.min(x_transform), np.max(x_transform), 1000)
gaussians = []
gmax = 0
for i, (w, m, std) in enumerate(zip(weights, means, variances)):
gaus = np.sqrt(w*stats.norm.pdf(x2, m, np.sqrt(std)))
gaussians.append(gaus)
gmax = max(np.max(gaus), gmax)
# compute classes defined by gaussian intersections
classes = np.argmax(np.vstack(gaussians), axis=0)
# find the minimum of any gaussian that is within its class
gmin = 100
for i, c in enumerate(np.unique(classes)):
gmin=min(gmin, np.min(gaussians[c][classes==c]))
# set up the figure
fig, ax1 = plt.subplots(figsize=(8, 4.8))
fig_ysize = 4
ax2 = ax1.twinx()
ax1.spines['top'].set_visible(False)
ax2.spines['top'].set_visible(False)
ax1.set_xlabel('x [a.u.]')
ax1.set_ylabel('#')
ax2.set_ylabel('Likelihood')
ax2.set_yscale('log')
ax1.set_yscale('log')
if use_log:
ax1.set_xscale('log')
ax1.set_xlabel(xlab)
# define colors for plotting gaussians
colidxs = -np.linspace(-1.25, -0.5, len(np.unique(classes)))
# plot the gaussians
for i, c in enumerate(np.unique(classes)):
ax2.plot(xplot, gaussians[c], c=lighter(c_grey, colidxs[i]), linewidth=2,
label=r'$N(\mu_%i, \sigma_%i)$'%(c, c))
# plot intersection lines
ax2.vlines(xplot[1:][np.diff(classes)!=0], 0, gmax/gmin, color='k', linewidth=2,
linestyle='--')
ax2.set_ylim(gmin, np.max(np.vstack(gaussians))*1.1)
# plot data distributions and classes
colidxs = -np.linspace(-1.25, -0.5, len(np.unique(labels)))
for i, l in enumerate(np.unique(labels)):
if use_log:
h, binn, _ = loghist(ax1, x[labels==l], np.min(x), np.max(x), 100,
lighter(ccol, colidxs[i]), label=r'$x_%i$'%l)
else:
h, binn, _ = ax1.hist(x[labels==l], bins=np.linspace(np.min(x), np.max(x), 100),
color=lighter(ccol, colidxs[i]), label=r'$x_%i$'%l)
# annotate merged clusters
for l in np.unique(labels_am):
maps = np.unique(labels[labels_am==l])
if len(maps) > 1:
x1 = x[labels==maps[0]]
x2 = x[labels==maps[1]]
print(np.median(x1))
print(np.median(x2))
print(gmax)
ax2.plot([np.median(x1), np.median(x2)], [1.2*gmax, 1.2*gmax], c='k', clip_on=False)
ax2.plot([np.median(x1), np.median(x1)], [1.1*gmax, 1.2*gmax], c='k', clip_on=False)
ax2.plot([np.median(x2), np.median(x2)], [1.1*gmax, 1.2*gmax], c='k', clip_on=False)
ax2.annotate(r'$\frac{|{\tilde{x}_%i-\tilde{x}_%i}|}{max(\tilde{x}_%i, \tilde{x}_%i)} < \epsilon$' % (maps[0], maps[1], maps[0], maps[1]), [np.median(x1)*1.1, gmax*1.2], xytext=(10, 10), textcoords='offset points', fontsize=12, annotation_clip=False, ha='center')
# add legends and plot.
ax2.legend(loc='lower left', frameon=False, bbox_to_anchor=(-0.05, 1.3),
ncol=len(np.unique(classes)))
ax1.legend(loc='upper left', frameon=False, bbox_to_anchor=(-0.05, 1.3),
ncol=len(np.unique(labels)))
plt.tight_layout()
def plot_feature_extraction(raw_snippets, normalized_snippets, features, labels, dt, pt):
"""Plot clustering step on EOD shape.
Parameters
----------
raw_snippets : 2D numpy array
Raw EOD snippets.
normalized_snippets : 2D numpy array
Normalized EOD snippets.
features : 2D numpy array
PCA values for each normalized EOD snippet.
labels : 1D numpy array of ints
Cluster labels.
dt : float
Sample interval of snippets.
pt : int
Set to 0 for peak-centered EODs and set to 1 for trough-centered EODs.
"""
ccol = cmap_pts[pt]
# set up the figure layout
fig = plt.figure(figsize=(((2+0.2)*4.8), 4.8))
outer = gridspec.GridSpec(1, 2, wspace=0.2, hspace=0)
x = np.arange(-dt*1000*raw_snippets.shape[1]/2, dt*1000*raw_snippets.shape[1]/2, dt*1000)
snip_ax = gridspec.GridSpecFromSubplotSpec(2, 1, subplot_spec = outer[0], hspace=0.35)
pc_ax = gridspec.GridSpecFromSubplotSpec(features.shape[1]-1, features.shape[1]-1,
subplot_spec = outer[1], hspace=0, wspace=0)
# 3 plots: raw snippets, normalized, pcs.
ax_raw_snip = fig.add_subplot(snip_ax[0])
ax_normalized_snip = fig.add_subplot(snip_ax[1])
colidxs = -np.linspace(-1.25, -0.5, len(np.unique(labels[labels>=0])))
j=0
for c in np.unique(labels):
if c<0:
color='lightgrey'
else:
color = lighter(ccol, colidxs[j])
j=j+1
ax_raw_snip.plot(x, raw_snippets[labels==c].T, color=color, label='-1',
rasterized=True, alpha=0.25)
ax_normalized_snip.plot(x, normalized_snippets[labels==c].T, color=color, alpha=0.25)
ax_raw_snip.spines['top'].set_visible(False)
ax_raw_snip.spines['right'].set_visible(False)
ax_raw_snip.get_xaxis().set_ticklabels([])
ax_raw_snip.set_title('Raw snippets')
ax_raw_snip.set_ylabel('Amplitude [a.u.]')
ax_normalized_snip.spines['top'].set_visible(False)
ax_normalized_snip.spines['right'].set_visible(False)
ax_normalized_snip.set_title('Normalized snippets')
ax_normalized_snip.set_ylabel('Amplitude [a.u.]')
ax_normalized_snip.set_xlabel('Time [ms]')
ax_raw_snip.axis('off')
ax_normalized_snip.axis('off')
ax_overlay = fig.add_subplot(pc_ax[:,:])
ax_overlay.set_title('Features')
ax_overlay.axis('off')
for n in range(features.shape[1]):
for m in range(n):
ax = fig.add_subplot(pc_ax[n-1,m])
ax.scatter(features[labels==c,m], features[labels==c,n], marker='.',
color=color, alpha=0.25)
ax.set_xlim(np.min(features), np.max(features))
ax.set_ylim(np.min(features), np.max(features))
ax.get_xaxis().set_ticklabels([])
ax.get_yaxis().set_ticklabels([])
ax.get_xaxis().set_ticks([])
ax.get_yaxis().set_ticks([])
if m==0:
ax.set_ylabel('PC %i'%(n+1))
if n==features.shape[1]-1:
ax.set_xlabel('PC %i'%(m+1))
ax = fig.add_subplot(pc_ax[0,features.shape[1]-2])
ax.set_xlim(np.min(features), np.max(features))
ax.set_ylim(np.min(features), np.max(features))
size = max(1, int(np.ceil(-np.log10(np.max(features)-np.min(features)))))
wbar = np.floor((np.max(features)-np.min(features))*10**size)/10**size
# should be smaller than the actual thing! so like x% of it?
xscalebar(ax, 0, 0, wbar, wformat='%%.%if'%size)
yscalebar(ax, 0, 0, wbar, hformat='%%.%if'%size)
ax.axis('off')
def plot_moving_fish(ws, dts, clusterss, ts, fishcounts, T, ignore_stepss):
"""Plot moving fish detection step.
Parameters
----------
ws : list of floats
Median width for each width cluster that the moving fish algorithm is computed on
(in seconds).
dts : list of floats
Sliding window size (in seconds) for each width cluster.
clusterss : list of 1D numpy int arrays
Cluster labels for each EOD cluster in a width cluster.
ts : list of 1D numpy float arrays
EOD emission times for each EOD in a width cluster.
fishcounts : list of lists
Sliding window timepoints and fishcounts for each width cluster.
T : float
Lenght of analyzed recording in seconds.
ignore_stepss : list of 1D int arrays
Mask for fishcounts that were ignored (ignored if True) in the moving_fish analysis.
"""
fig = plt.figure()
# create gridspec
outer = gridspec.GridSpec(len(ws), 1)
for i, (w, dt, clusters, t, fishcount, ignore_steps) in enumerate(zip(ws, dts, clusterss, ts, fishcounts, ignore_stepss)):
gs = gridspec.GridSpecFromSubplotSpec(2, 1, subplot_spec = outer[i])
# axis for clusters
ax1 = fig.add_subplot(gs[0])
# axis for fishcount
ax2 = fig.add_subplot(gs[1])
# plot clusters as eventplot
for cnum, c in enumerate(np.unique(clusters[clusters>=0])):
ax1.eventplot(t[clusters==c], lineoffsets=cnum, linelengths=0.5, color=cmap(i))
cnum = cnum + 1
# Plot the sliding window
rect=Rectangle((0, -0.5), dt, cnum, linewidth=1, linestyle='--', edgecolor='k',
facecolor='none', clip_on=False)
ax1.add_patch(rect)
ax1.arrow(dt+0.1, -0.5, 0.5, 0, head_width=0.1, head_length=0.1, facecolor='k',
edgecolor='k')
# plot parameters
ax1.set_title(r'$\tilde{w}_%i = %.3f ms$'%(i, 1000*w))
ax1.set_ylabel('cluster #')
ax1.set_yticks(range(0, cnum))
ax1.set_xlabel('time')
ax1.set_xlim(0, T)
ax1.axes.xaxis.set_visible(False)
ax1.spines['bottom'].set_visible(False)
ax1.spines['top'].set_visible(False)
ax1.spines['right'].set_visible(False)
ax1.spines['left'].set_visible(False)
# plot for fishcount
x = fishcount[0]
y = fishcount[1]
ax2 = fig.add_subplot(gs[1])
ax2.spines['top'].set_visible(False)
ax2.spines['right'].set_visible(False)
ax2.spines['bottom'].set_visible(False)
ax2.axes.xaxis.set_visible(False)
yplot = np.copy(y)
ax2.plot(x+dt/2, yplot, linestyle='-', marker='.', c=cmap(i), alpha=0.25)
yplot[ignore_steps.astype(bool)] = np.NaN
ax2.plot(x+dt/2, yplot, linestyle='-', marker='.', c=cmap(i))
ax2.set_ylabel('Fish count')
ax2.set_yticks(range(int(np.min(y)), 1+int(np.max(y))))
ax2.set_xlim(0, T)
if i < len(ws)-1:
ax2.axes.xaxis.set_visible(False)
else:
ax2.axes.xaxis.set_visible(False)
xscalebar(ax2, 1, 0, 1, wunit='s', ha='right')
con = ConnectionPatch([0, -0.5], [dt/2, y[0]], "data", "data",
axesA=ax1, axesB=ax2, color='k')
ax2.add_artist(con)
con = ConnectionPatch([dt, -0.5], [dt/2, y[0]], "data", "data",
axesA=ax1, axesB=ax2, color='k')
ax2.add_artist(con)
plt.xlim(0, T)Functions
def warn(*args, **kwargs)-
Ignore all warnings.
Expand source code
def warn(*args, **kwargs): """ Ignore all warnings. """ pass def darker(color, saturation)-
Make a color darker.
From bendalab/plottools package.
Parameters
color:dictormatplotlib color spec- A matplotlib color (hex string, name color string, rgb tuple) or a dictionary with an 'color' or 'facecolor' key.
saturation:float- The smaller the saturation, the darker the returned color. A saturation of 0 returns black. A saturation of 1 leaves the color untouched. A saturation of 2 returns white.
Returns
color:stringordictionary- The darker color as a hexadecimal RGB string (e.g. '#rrggbb').
If
coloris a dictionary, a copy of the dictionary is returned with the value of 'color' or 'facecolor' set to the darker color.
Expand source code
def darker(color, saturation): """ Make a color darker. From bendalab/plottools package. Parameters ---------- color: dict or matplotlib color spec A matplotlib color (hex string, name color string, rgb tuple) or a dictionary with an 'color' or 'facecolor' key. saturation: float The smaller the saturation, the darker the returned color. A saturation of 0 returns black. A saturation of 1 leaves the color untouched. A saturation of 2 returns white. Returns ------- color: string or dictionary The darker color as a hexadecimal RGB string (e.g. '#rrggbb'). If `color` is a dictionary, a copy of the dictionary is returned with the value of 'color' or 'facecolor' set to the darker color. """ try: c = color['color'] cd = dict(**color) cd['color'] = darker(c, saturation) return cd except (KeyError, TypeError): try: c = color['facecolor'] cd = dict(**color) cd['facecolor'] = darker(c, saturation) return cd except (KeyError, TypeError): if saturation > 2: sauration = 2 if saturation > 1: return lighter(color, 2.0-saturation) if saturation < 0: saturation = 0 r, g, b = cc.to_rgb(color) rd = r*saturation gd = g*saturation bd = b*saturation return to_hex((rd, gd, bd)).upper() def lighter(color, lightness)-
Make a color lighter
From bendalab/plottools package.
Parameters
color:dictormatplotlib color spec- A matplotlib color (hex string, name color string, rgb tuple) or a dictionary with an 'color' or 'facecolor' key.
lightness:float- The smaller the lightness, the lighter the returned color. A lightness of 0 returns white. A lightness of 1 leaves the color untouched. A lightness of 2 returns black.
Returns
color:stringordict- The lighter color as a hexadecimal RGB string (e.g. '#rrggbb').
If
coloris a dictionary, a copy of the dictionary is returned with the value of 'color' or 'facecolor' set to the lighter color.
Expand source code
def lighter(color, lightness): """Make a color lighter From bendalab/plottools package. Parameters ---------- color: dict or matplotlib color spec A matplotlib color (hex string, name color string, rgb tuple) or a dictionary with an 'color' or 'facecolor' key. lightness: float The smaller the lightness, the lighter the returned color. A lightness of 0 returns white. A lightness of 1 leaves the color untouched. A lightness of 2 returns black. Returns ------- color: string or dict The lighter color as a hexadecimal RGB string (e.g. '#rrggbb'). If `color` is a dictionary, a copy of the dictionary is returned with the value of 'color' or 'facecolor' set to the lighter color. """ try: c = color['color'] cd = dict(**color) cd['color'] = lighter(c, lightness) return cd except (KeyError, TypeError): try: c = color['facecolor'] cd = dict(**color) cd['facecolor'] = lighter(c, lightness) return cd except (KeyError, TypeError): if lightness > 2: lightness = 2 if lightness > 1: return darker(color, 2.0-lightness) if lightness < 0: lightness = 0 r, g, b = cc.to_rgb(color) rl = r + (1.0-lightness)*(1.0 - r) gl = g + (1.0-lightness)*(1.0 - g) bl = b + (1.0-lightness)*(1.0 - b) return to_hex((rl, gl, bl)).upper() def xscalebar(ax, x, y, width, wunit=None, wformat=None, ha='left', va='bottom', lw=None, color=None, capsize=None, clw=None, **kwargs)-
Horizontal scale bar with label.
From bendalab/plottools package.
Parameters
ax:matplotlib axes- Axes where to draw the scale bar.
x:float- x-coordinate where to draw the scale bar in relative units of the axes.
y:float- y-coordinate where to draw the scale bar in relative units of the axes.
width:float- Length of the scale bar in units of the data's x-values.
wunit:stringorNone- Optional unit of the data's x-values.
wformat:stringorNone- Optional format string for formatting the label of the scale bar or simply a string used for labeling the scale bar.
ha:'left', 'right',or'center'- Scale bar aligned left, right, or centered to (x, y)
va:'top'or'bottom'- Label of the scale bar either above or below the scale bar.
lw:int, float, None- Line width of the scale bar.
color:matplotlib color- Color of the scalebar.
capsize:floatorNone- If larger then zero draw cap lines at the ends of the bar. The length of the lines is given in points (same unit as linewidth).
clw:int, float, None- Line width of the cap lines.
kwargs:key-word arguments- Passed on to
ax.text()used to print the scale bar label.
Expand source code
def xscalebar(ax, x, y, width, wunit=None, wformat=None, ha='left', va='bottom', lw=None, color=None, capsize=None, clw=None, **kwargs): """Horizontal scale bar with label. From bendalab/plottools package. Parameters ---------- ax: matplotlib axes Axes where to draw the scale bar. x: float x-coordinate where to draw the scale bar in relative units of the axes. y: float y-coordinate where to draw the scale bar in relative units of the axes. width: float Length of the scale bar in units of the data's x-values. wunit: string or None Optional unit of the data's x-values. wformat: string or None Optional format string for formatting the label of the scale bar or simply a string used for labeling the scale bar. ha: 'left', 'right', or 'center' Scale bar aligned left, right, or centered to (x, y) va: 'top' or 'bottom' Label of the scale bar either above or below the scale bar. lw: int, float, None Line width of the scale bar. color: matplotlib color Color of the scalebar. capsize: float or None If larger then zero draw cap lines at the ends of the bar. The length of the lines is given in points (same unit as linewidth). clw: int, float, None Line width of the cap lines. kwargs: key-word arguments Passed on to `ax.text()` used to print the scale bar label. """ ax.autoscale(False) # ax dimensions: pixelx = np.abs(np.diff(ax.get_window_extent().get_points()[:,0]))[0] pixely = np.abs(np.diff(ax.get_window_extent().get_points()[:,1]))[0] xmin, xmax = ax.get_xlim() ymin, ymax = ax.get_ylim() unitx = xmax - xmin unity = ymax - ymin dxu = np.abs(unitx)/pixelx dyu = np.abs(unity)/pixely # transform x, y from relative units to axis units: x = xmin + x*unitx y = ymin + y*unity # bar length: if wformat is None: wformat = '%.0f' if width < 1.0: wformat = '%.1f' try: ls = wformat % width width = float(ls) except TypeError: ls = wformat # bar: if ha == 'left': x0 = x x1 = x+width elif ha == 'right': x0 = x-width x1 = x else: x0 = x-0.5*width x1 = x+0.5*width # line width: if lw is None: lw = 2 # color: if color is None: color = 'k' # scalebar: lh = ax.plot([x0, x1], [y, y], '-', color=color, lw=lw, solid_capstyle='butt', clip_on=False) # get y position of line in figure pixel coordinates: ly = np.array(lh[0].get_window_extent(ax.get_figure().canvas.get_renderer()))[0,1] # caps: if capsize is None: capsize = 0 if clw is None: clw = 0.5 if capsize > 0.0: dy = capsize*dyu ax.plot([x0, x0], [y-dy, y+dy], '-', color=color, lw=clw, solid_capstyle='butt', clip_on=False) ax.plot([x1, x1], [y-dy, y+dy], '-', color=color, lw=clw, solid_capstyle='butt', clip_on=False) # label: if wunit: ls += u'\u2009%s' % wunit if va == 'top': th = ax.text(0.5*(x0+x1), y, ls, clip_on=False, ha='center', va='bottom', **kwargs) # get y coordinate of text bottom in figure pixel coordinates: ty = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[0,1] dty = ly+0.5*lw + 2.0 - ty else: th = ax.text(0.5*(x0+x1), y, ls, clip_on=False, ha='center', va='top', **kwargs) # get y coordinate of text bottom in figure pixel coordinates: ty = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[1,1] dty = ly-0.5*lw - 2.0 - ty th.set_position((0.5*(x0+x1), y+dyu*dty)) return x0, x1, y def yscalebar(ax, x, y, height, hunit=None, hformat=None, ha='left', va='bottom', lw=None, color=None, capsize=None, clw=None, **kwargs)-
Vertical scale bar with label.
From bendalab/plottools package.
Parameters
ax:matplotlib axes- Axes where to draw the scale bar.
x:float- x-coordinate where to draw the scale bar in relative units of the axes.
y:float- y-coordinate where to draw the scale bar in relative units of the axes.
height:float- Length of the scale bar in units of the data's y-values.
hunit:string- Unit of the data's y-values.
hformat:stringorNone- Optional format string for formatting the label of the scale bar or simply a string used for labeling the scale bar.
ha:'left'or'right'- Label of the scale bar either to the left or to the right of the scale bar.
va:'top', 'bottom',or'center'- Scale bar aligned above, below, or centered on (x, y).
lw:int, float, None- Line width of the scale bar.
color:matplotlib color- Color of the scalebar.
capsize:floatorNone- If larger then zero draw cap lines at the ends of the bar. The length of the lines is given in points (same unit as linewidth).
clw:int, float- Line width of the cap lines.
kwargs:key-word arguments- Passed on to
ax.text()used to print the scale bar label.
Expand source code
def yscalebar(ax, x, y, height, hunit=None, hformat=None, ha='left', va='bottom', lw=None, color=None, capsize=None, clw=None, **kwargs): """Vertical scale bar with label. From bendalab/plottools package. Parameters ---------- ax: matplotlib axes Axes where to draw the scale bar. x: float x-coordinate where to draw the scale bar in relative units of the axes. y: float y-coordinate where to draw the scale bar in relative units of the axes. height: float Length of the scale bar in units of the data's y-values. hunit: string Unit of the data's y-values. hformat: string or None Optional format string for formatting the label of the scale bar or simply a string used for labeling the scale bar. ha: 'left' or 'right' Label of the scale bar either to the left or to the right of the scale bar. va: 'top', 'bottom', or 'center' Scale bar aligned above, below, or centered on (x, y). lw: int, float, None Line width of the scale bar. color: matplotlib color Color of the scalebar. capsize: float or None If larger then zero draw cap lines at the ends of the bar. The length of the lines is given in points (same unit as linewidth). clw: int, float Line width of the cap lines. kwargs: key-word arguments Passed on to `ax.text()` used to print the scale bar label. """ ax.autoscale(False) # ax dimensions: pixelx = np.abs(np.diff(ax.get_window_extent().get_points()[:,0]))[0] pixely = np.abs(np.diff(ax.get_window_extent().get_points()[:,1]))[0] xmin, xmax = ax.get_xlim() ymin, ymax = ax.get_ylim() unitx = xmax - xmin unity = ymax - ymin dxu = np.abs(unitx)/pixelx dyu = np.abs(unity)/pixely # transform x, y from relative units to axis units: x = xmin + x*unitx y = ymin + y*unity # bar length: if hformat is None: hformat = '%.0f' if height < 1.0: hformat = '%.1f' try: ls = hformat % height width = float(ls) except TypeError: ls = hformat # bar: if va == 'bottom': y0 = y y1 = y+height elif va == 'top': y0 = y-height y1 = y else: y0 = y-0.5*height y1 = y+0.5*height # line width: if lw is None: lw = 2 # color: if color is None: color = 'k' # scalebar: lh = ax.plot([x, x], [y0, y1], '-', color=color, lw=lw, solid_capstyle='butt', clip_on=False) # get x position of line in figure pixel coordinates: lx = np.array(lh[0].get_window_extent(ax.get_figure().canvas.get_renderer()))[0,0] # caps: if capsize is None: capsize = 0 if clw is None: clw = 0.5 if capsize > 0.0: dx = capsize*dxu ax.plot([x-dx, x+dx], [y0, y0], '-', color=color, lw=clw, solid_capstyle='butt', clip_on=False) ax.plot([x-dx, x+dx], [y1, y1], '-', color=color, lw=clw, solid_capstyle='butt', clip_on=False) # label: if hunit: ls += u'\u2009%s' % hunit if ha == 'right': th = ax.text(x, 0.5*(y0+y1), ls, clip_on=False, rotation=90.0, ha='left', va='center', **kwargs) # get x coordinate of text bottom in figure pixel coordinates: tx = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[0,0] dtx = lx+0.5*lw + 2.0 - tx else: th = ax.text(x, 0.5*(y0+y1), ls, clip_on=False, rotation=90.0, ha='right', va='center', **kwargs) # get x coordinate of text bottom in figure pixel coordinates: tx = np.array(th.get_window_extent(ax.get_figure().canvas.get_renderer()))[1,0] dtx = lx-0.5*lw - 1.0 - tx th.set_position((x+dxu*dtx, 0.5*(y0+y1))) return x, y0, y1 def arrowed_spines(ax, ms=10)-
Spine with arrow on the y-axis of a plot.
Parameters
ax:matplotlib figure axis- Axis on which the arrow should be plot.
Expand source code
def arrowed_spines(ax, ms=10): """ Spine with arrow on the y-axis of a plot. Parameters ---------- ax : matplotlib figure axis Axis on which the arrow should be plot. """ xmin, xmax = ax.get_xlim() ymin, ymax = ax.get_ylim() ax.scatter([xmin], [ymax], s=ms, marker='^', clip_on=False, color='k') ax.set_xlim(xmin, xmax) ax.set_ylim(ymin, ymax) def loghist(ax, x, bmin, bmax, n, c, orientation='vertical', label='')-
Plot histogram with logarithmic scale.
Parameters
ax:matplotlib axis- Axis to plot the histogram on.
x:numpy array- Input data for histogram.
bmin:float- Minimum value for the histogram bins.
bmax:float- Maximum value for the histogram bins.
n:int- Number of bins.
c:matplotlib color- Color of histogram.
orientation:string (optional)- Histogram orientation. Defaults to 'vertical'.
label:string (optional)- Label for x. Defaults to '' (no label).
Returns
n:array- The values of the histogram bins.
bins:array- The edges of the bins.
patches:BarContainer- Container of individual artists used to create the histogram.
Expand source code
def loghist(ax, x, bmin, bmax, n, c, orientation='vertical', label=''): """ Plot histogram with logarithmic scale. Parameters ---------- ax : matplotlib axis Axis to plot the histogram on. x : numpy array Input data for histogram. bmin : float Minimum value for the histogram bins. bmax : float Maximum value for the histogram bins. n : int Number of bins. c : matplotlib color Color of histogram. orientation : string (optional) Histogram orientation. Defaults to 'vertical'. label : string (optional) Label for x. Defaults to '' (no label). Returns ------- n : array The values of the histogram bins. bins : array The edges of the bins. patches : BarContainer Container of individual artists used to create the histogram. """ return ax.hist(x, bins=np.exp(np.linspace(np.log(bmin), np.log(bmax), n)), color=c, orientation=orientation, label=label) def plot_all(data, eod_p_times, eod_tr_times, fs, mean_eods)-
Quick way to view the output of extract_pulsefish in a single plot.
Parameters
data:array- Recording data.
eod_p_times:arrayofints- EOD peak indices.
eod_tr_times:arrayofints- EOD trough indices.
fs:float- Samplerate.
mean_eods:listofnumpy arrays- Mean EODs of each pulsefish found in the recording.
Expand source code
def plot_all(data, eod_p_times, eod_tr_times, fs, mean_eods): """Quick way to view the output of extract_pulsefish in a single plot. Parameters ---------- data: array Recording data. eod_p_times: array of ints EOD peak indices. eod_tr_times: array of ints EOD trough indices. fs: float Samplerate. mean_eods: list of numpy arrays Mean EODs of each pulsefish found in the recording. """ fig = plt.figure(figsize=(10, 5)) if len(eod_p_times) > 0: gs = gridspec.GridSpec(2, len(eod_p_times)) ax = fig.add_subplot(gs[0,:]) ax.plot(np.arange(len(data))/fs, data, c='k', alpha=0.3) for i, (pt, tt) in enumerate(zip(eod_p_times, eod_tr_times)): ax.plot(pt, data[(pt*fs).astype('int')], 'o', label=i+1, ms=10, c=cmap(i)) ax.plot(tt, data[(tt*fs).astype('int')], 'o', label=i+1, ms=10, c=cmap(i)) ax.set_xlabel('time [s]') ax.set_ylabel('amplitude [V]') for i, m in enumerate(mean_eods): ax = fig.add_subplot(gs[1,i]) ax.plot(1000*m[0], 1000*m[1], c='k') ax.fill_between(1000*m[0], 1000*(m[1]-m[2]), 1000*(m[1]+m[2]), color=cmap(i)) ax.set_xlabel('time [ms]') ax.set_ylabel('amplitude [mV]') else: plt.plot(np.arange(len(data))/fs, data, c='k', alpha=0.3) plt.tight_layout() def plot_clustering(samplerate, eod_widths, eod_hights, eod_shapes, disc_masks, merge_masks)-
Plot all clustering steps.
Plot clustering steps on width, height and shape. Then plot the remaining EODs after the EOD assessment step and the EODs after the merge step.
Parameters
samplerate:float- Samplerate of EOD snippets.
eod_widths:listofthree 1D numpy arrays- The first list entry gives the unique labels of all width clusters as a list of ints. The second list entry gives the width values for each EOD in samples as a
- 1D numpy array of ints.
- The third list entry gives the width labels for each EOD as a 1D numpy array of ints.
eod_hights:nested lists (2 layers)ofthree 1D numpy arrays- The first list entry gives the unique labels of all height clusters as a list of ints
- for each width cluster.
- The second list entry gives the height values for each EOD as a 1D numpy array
- of floats for each width cluster.
- The third list entry gives the height labels for each EOD as a 1D numpy array
- of ints for each width cluster.
eod_shapes:nested lists (3 layers)ofthree 1D numpy arrays- The first list entry gives the raw EOD snippets as a 2D numpy array for each
- height cluster in a width cluster.
- The second list entry gives the snippet PCA values for each EOD as a 2D numpy array
- of floats for each height cluster in a width cluster.
- The third list entry gives the shape labels for each EOD as a 1D numpy array of ints
- for each height cluster in a width cluster.
disc_masks:Nested lists (two layers)of1D numpy arrays- The masks of EODs that are discarded by the discarding step of the algorithm.
- The masks are 1D boolean arrays where
- instances that are set to True are discarded by the algorithm. Discarding masks
- are saved in nested lists that represent the width and height clusters.
merge_masks:Nested lists (two layers)of2D numpy arrays- The masks of EODs that are discarded by the merging step of the algorithm.
The masks are 2D boolean arrays where for each sample point
ieithermerge_mask[i,0]ormerge_mask[i,1]is set to True. Here, merge_mask[:,0] represents the peak-centered clusters andmerge_mask[:,1]represents the trough-centered clusters. Merge masks are saved in nested lists that represent the width and height clusters.Expand source code
def plot_clustering(samplerate, eod_widths, eod_hights, eod_shapes, disc_masks, merge_masks): """Plot all clustering steps. Plot clustering steps on width, height and shape. Then plot the remaining EODs after the EOD assessment step and the EODs after the merge step. Parameters ---------- samplerate : float Samplerate of EOD snippets. eod_widths : list of three 1D numpy arrays The first list entry gives the unique labels of all width clusters as a list of ints. The second list entry gives the width values for each EOD in samples as a 1D numpy array of ints. The third list entry gives the width labels for each EOD as a 1D numpy array of ints. eod_hights : nested lists (2 layers) of three 1D numpy arrays The first list entry gives the unique labels of all height clusters as a list of ints for each width cluster. The second list entry gives the height values for each EOD as a 1D numpy array of floats for each width cluster. The third list entry gives the height labels for each EOD as a 1D numpy array of ints for each width cluster. eod_shapes : nested lists (3 layers) of three 1D numpy arrays The first list entry gives the raw EOD snippets as a 2D numpy array for each height cluster in a width cluster. The second list entry gives the snippet PCA values for each EOD as a 2D numpy array of floats for each height cluster in a width cluster. The third list entry gives the shape labels for each EOD as a 1D numpy array of ints for each height cluster in a width cluster. disc_masks : Nested lists (two layers) of 1D numpy arrays The masks of EODs that are discarded by the discarding step of the algorithm. The masks are 1D boolean arrays where instances that are set to True are discarded by the algorithm. Discarding masks are saved in nested lists that represent the width and height clusters. merge_masks : Nested lists (two layers) of 2D numpy arrays The masks of EODs that are discarded by the merging step of the algorithm. The masks are 2D boolean arrays where for each sample point `i` either `merge_mask[i,0]` or `merge_mask[i,1]` is set to True. Here, merge_mask[:,0] represents the peak-centered clusters and `merge_mask[:,1]` represents the trough-centered clusters. Merge masks are saved in nested lists that represent the width and height clusters. """ # create figure + transparant figure. fig = plt.figure(figsize=(12, 7)) transFigure = fig.transFigure.inverted() # set up the figure layout outer = gridspec.GridSpec(1, 5, width_ratios=[1, 1, 2, 1, 2], left=0.05, right=0.95) # set titles for each clustering step titles = ['1. Widths', '2. Heights', '3. Shape', '4. Pulse EODs', '5. Merge'] for i, title in enumerate(titles): title_ax = gridspec.GridSpecFromSubplotSpec(1, 1, subplot_spec = outer[i]) ax = fig.add_subplot(title_ax[0]) ax.text(0, 110, title, ha='center', va='bottom', clip_on=False) ax.set_xlim(-100, 100) ax.set_ylim(-100, 100) ax.axis('off') # compute sizes for each axis w_size = 1 h_size = len(eod_hights[1]) shape_size = np.sum([len(sl) for sl in eod_shapes[0]]) # count required axes sized for the last two plot columns. disc_size = 0 merge_size= 0 for shapelabel, dmasks, mmasks in zip(eod_shapes[2], disc_masks, merge_masks): for sl, dm, mm in zip(shapelabel, dmasks, mmasks): uld1 = np.unique((sl[0]+1)*np.invert(dm[0])) uld2 = np.unique((sl[1]+1)*np.invert(dm[1])) disc_size = disc_size+len(uld1[uld1>0])+len(uld2[uld2>0]) uld1 = np.unique((sl[0]+1)*mm[0]) uld2 = np.unique((sl[1]+1)*mm[1]) merge_size = merge_size+len(uld1[uld1>0])+len(uld2[uld2>0]) # set counters to keep track of the plot axes disc_block = 0 merge_block = 0 shape_count = 0 # create all axes width_hist_ax = gridspec.GridSpecFromSubplotSpec(w_size, 1, subplot_spec = outer[0]) hight_hist_ax = gridspec.GridSpecFromSubplotSpec(h_size, 1, subplot_spec = outer[1]) shape_ax = gridspec.GridSpecFromSubplotSpec(shape_size, 1, subplot_spec = outer[2]) shape_windows = [gridspec.GridSpecFromSubplotSpec(2, 2, hspace=0.0, wspace=0.0, subplot_spec=shape_ax[i]) for i in range(shape_size)] EOD_delete_ax = gridspec.GridSpecFromSubplotSpec(disc_size, 1, subplot_spec=outer[3]) EOD_merge_ax = gridspec.GridSpecFromSubplotSpec(merge_size, 1, subplot_spec=outer[4]) # plot width labels histogram ax1 = fig.add_subplot(width_hist_ax[0]) # set axes features. ax1.set_xscale('log') ax1.spines['top'].set_visible(False) ax1.spines['right'].set_visible(False) ax1.spines['bottom'].set_visible(False) ax1.axes.xaxis.set_visible(False) ax1.set_yticklabels([]) # indices for plot colors (dark to light) colidxsw = -np.linspace(-1.25, -0.5, h_size) for i, (wl, colw, uhl, eod_h, eod_h_labs, w_snip, w_feat, w_lab, w_dm, w_mm) in enumerate(zip(eod_widths[0], colidxsw, eod_hights[0], eod_hights[1], eod_hights[2], eod_shapes[0], eod_shapes[1], eod_shapes[2], disc_masks, merge_masks)): # plot width hist hw, _, _ = ax1.hist(eod_widths[1][eod_widths[2]==wl], bins=np.linspace(np.min(eod_widths[1]), np.max(eod_widths[1]), 100), color=lighter(c_o, colw), orientation='horizontal') # set arrow when the last hist is plot so the size of the axes are known. if i == h_size-1: arrowed_spines(ax1, ms=20) # determine total size of the hight historgams now. my, b = np.histogram(eod_h, bins=np.exp(np.linspace(np.min(np.log(eod_h)), np.max(np.log(eod_h)), 100))) maxy = np.max(my) # set axes features for hight hist. ax2 = fig.add_subplot(hight_hist_ax[h_size-i-1]) ax2.set_xscale('log') ax2.spines['top'].set_visible(False) ax2.spines['right'].set_visible(False) ax2.spines['bottom'].set_visible(False) ax2.set_xlim(0.9, maxy) ax2.axes.xaxis.set_visible(False) ax2.set_yscale('log') ax2.yaxis.set_major_formatter(ticker.NullFormatter()) ax2.yaxis.set_minor_formatter(ticker.NullFormatter()) # define colors for plots colidxsh = -np.linspace(-1.25, -0.5, len(uhl)) for n, (hl, hcol, snippets, features, labels, dmasks, mmasks) in enumerate(zip(uhl, colidxsh, w_snip, w_feat, w_lab, w_dm, w_mm)): hh, _, _ = loghist(ax2, eod_h[eod_h_labs==hl], np.min(eod_h), np.max(eod_h), 100, lighter(c_g, hcol), orientation='horizontal') # set arrow spines only on last plot if n == len(uhl)-1: arrowed_spines(ax2, ms=10) # plot line from the width histogram to the height histogram. if n == 0: coord1 = transFigure.transform(ax1.transData.transform([np.median(hw[hw!=0]), np.median(eod_widths[1][eod_widths[2]==wl])])) coord2 = transFigure.transform(ax2.transData.transform([0.9, np.mean(eod_h)])) line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]), transform=fig.transFigure, color='grey', linewidth=0.5) fig.lines.append(line) # compute sizes of the eod_discarding and merge steps s1 = np.unique((labels[0]+1)*(~dmasks[0])) s2 = np.unique((labels[1]+1)*(~dmasks[1])) disc_block = disc_block + len(s1[s1>0]) + len(s2[s2>0]) s1 = np.unique((labels[0]+1)*(mmasks[0])) s2 = np.unique((labels[1]+1)*(mmasks[1])) merge_block = merge_block + len(s1[s1>0]) + len(s2[s2>0]) axs = [] disc_count = 0 merge_count = 0 # now plot the clusters for peak and trough centerings for pt, cmap_pt in zip([0, 1], cmap_pts): ax3 = fig.add_subplot(shape_windows[shape_size-1-shape_count][pt,0]) ax4 = fig.add_subplot(shape_windows[shape_size-1-shape_count][pt,1]) # remove axes ax3.axes.xaxis.set_visible(False) ax4.axes.yaxis.set_visible(False) ax3.axes.yaxis.set_visible(False) ax4.axes.xaxis.set_visible(False) # set color indices colidxss = -np.linspace(-1.25, -0.5, len(np.unique(labels[pt][labels[pt]>=0]))) j=0 for c in np.unique(labels[pt]): if c<0: # plot noise features + snippets ax3.plot(features[pt][labels[pt]==c,0], features[pt][labels[pt]==c,1], '.', color='lightgrey', label='-1', rasterized=True) ax4.plot(snippets[pt][labels[pt]==c].T, linewidth=0.1, color='lightgrey', label='-1', rasterized=True) else: # plot cluster features and snippets ax3.plot(features[pt][labels[pt]==c,0], features[pt][labels[pt]==c,1], '.', color=lighter(cmap_pt, colidxss[j]), label=c, rasterized=True) ax4.plot(snippets[pt][labels[pt]==c].T, linewidth=0.1, color=lighter(cmap_pt, colidxss[j]), label=c, rasterized=True) # check if the current cluster is an EOD, if yes, plot it. if np.sum(dmasks[pt][labels[pt]==c]) == 0: ax = fig.add_subplot(EOD_delete_ax[disc_size-disc_block+disc_count]) ax.axis('off') # plot mean EOD snippet ax.plot(np.mean(snippets[pt][labels[pt]==c], axis=0), color=lighter(cmap_pt, colidxss[j])) disc_count = disc_count + 1 # match colors and draw line.. coord1 = transFigure.transform(ax4.transData.transform([ax4.get_xlim()[1], ax4.get_ylim()[0] + 0.5*(ax4.get_ylim()[1]-ax4.get_ylim()[0])])) coord2 = transFigure.transform(ax.transData.transform([ax.get_xlim()[0],ax.get_ylim()[0] + 0.5*(ax.get_ylim()[1]-ax.get_ylim()[0])])) line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]), transform=fig.transFigure, color='grey', linewidth=0.5) fig.lines.append(line) axs.append(ax) # check if the current EOD survives the merge step # if so, plot it. if np.sum(mmasks[pt, labels[pt]==c])>0: ax = fig.add_subplot(EOD_merge_ax[merge_size-merge_block+merge_count]) ax.axis('off') ax.plot(np.mean(snippets[pt][labels[pt]==c], axis=0), color=lighter(cmap_pt, colidxss[j])) merge_count = merge_count + 1 j=j+1 if pt==0: # draw line from hight cluster to EOD shape clusters. coord1 = transFigure.transform(ax2.transData.transform([np.median(hh[hh!=0]), np.median(eod_h[eod_h_labs==hl])])) coord2 = transFigure.transform(ax3.transData.transform([ax3.get_xlim()[0], ax3.get_ylim()[0]])) line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]), transform=fig.transFigure, color='grey', linewidth=0.5) fig.lines.append(line) shape_count = shape_count + 1 if len(axs)>0: # plot lines that indicate the merged clusters. coord1 = transFigure.transform(axs[0].transData.transform([axs[0].get_xlim()[1]+0.1*(axs[0].get_xlim()[1]-axs[0].get_xlim()[0]), axs[0].get_ylim()[1]-0.25*(axs[0].get_ylim()[1]-axs[0].get_ylim()[0])])) coord2 = transFigure.transform(axs[-1].transData.transform([axs[-1].get_xlim()[1]+0.1*(axs[-1].get_xlim()[1]-axs[-1].get_xlim()[0]), axs[-1].get_ylim()[0]+0.25*(axs[-1].get_ylim()[1]-axs[-1].get_ylim()[0])])) line = Line2D((coord1[0], coord2[0]), (coord1[1], coord2[1]), transform=fig.transFigure, color='grey', linewidth=1) fig.lines.append(line) def plot_bgm(x, means, variances, weights, use_log, labels, labels_am, xlab)-
Plot a BGM clustering step either on EOD width or height.
Parameters
x:1D numpy arrayoffloats- BGM input values.
means:listoffloats- BGM Gaussian means
variances:listoffloats- BGM Gaussian variances.
weights:listoffloats- BGM Gaussian weights.
use_log:boolean- True if the z-scored logarithm of the data was used as BGM input.
labels:1D numpy arrayofints- Labels defined by BGM model (before merging based on merge factor).
labels_am:1D numpy arrayofints- Labels defined by BGM model (after merging based on merge factor).
xlab:string- Label for plot (defines the units of the BGM data).
Expand source code
def plot_bgm(x, means, variances, weights, use_log, labels, labels_am, xlab): """Plot a BGM clustering step either on EOD width or height. Parameters ---------- x : 1D numpy array of floats BGM input values. means : list of floats BGM Gaussian means variances : list of floats BGM Gaussian variances. weights : list of floats BGM Gaussian weights. use_log : boolean True if the z-scored logarithm of the data was used as BGM input. labels : 1D numpy array of ints Labels defined by BGM model (before merging based on merge factor). labels_am : 1D numpy array of ints Labels defined by BGM model (after merging based on merge factor). xlab : string Label for plot (defines the units of the BGM data). """ if 'width' in xlab: ccol = c_o elif 'height' in xlab: ccol = c_g else: ccol = 'b' # get the transform that was used as BGM input if use_log: x_transform = stats.zscore(np.log(x)) xplot = np.exp(np.linspace(np.log(np.min(x)), np.log(np.max(x)), 1000)) else: x_transform = stats.zscore(x) xplot = np.linspace(np.min(x), np.max(x), 1000) # compute the x values and gaussians x2 = np.linspace(np.min(x_transform), np.max(x_transform), 1000) gaussians = [] gmax = 0 for i, (w, m, std) in enumerate(zip(weights, means, variances)): gaus = np.sqrt(w*stats.norm.pdf(x2, m, np.sqrt(std))) gaussians.append(gaus) gmax = max(np.max(gaus), gmax) # compute classes defined by gaussian intersections classes = np.argmax(np.vstack(gaussians), axis=0) # find the minimum of any gaussian that is within its class gmin = 100 for i, c in enumerate(np.unique(classes)): gmin=min(gmin, np.min(gaussians[c][classes==c])) # set up the figure fig, ax1 = plt.subplots(figsize=(8, 4.8)) fig_ysize = 4 ax2 = ax1.twinx() ax1.spines['top'].set_visible(False) ax2.spines['top'].set_visible(False) ax1.set_xlabel('x [a.u.]') ax1.set_ylabel('#') ax2.set_ylabel('Likelihood') ax2.set_yscale('log') ax1.set_yscale('log') if use_log: ax1.set_xscale('log') ax1.set_xlabel(xlab) # define colors for plotting gaussians colidxs = -np.linspace(-1.25, -0.5, len(np.unique(classes))) # plot the gaussians for i, c in enumerate(np.unique(classes)): ax2.plot(xplot, gaussians[c], c=lighter(c_grey, colidxs[i]), linewidth=2, label=r'$N(\mu_%i, \sigma_%i)$'%(c, c)) # plot intersection lines ax2.vlines(xplot[1:][np.diff(classes)!=0], 0, gmax/gmin, color='k', linewidth=2, linestyle='--') ax2.set_ylim(gmin, np.max(np.vstack(gaussians))*1.1) # plot data distributions and classes colidxs = -np.linspace(-1.25, -0.5, len(np.unique(labels))) for i, l in enumerate(np.unique(labels)): if use_log: h, binn, _ = loghist(ax1, x[labels==l], np.min(x), np.max(x), 100, lighter(ccol, colidxs[i]), label=r'$x_%i$'%l) else: h, binn, _ = ax1.hist(x[labels==l], bins=np.linspace(np.min(x), np.max(x), 100), color=lighter(ccol, colidxs[i]), label=r'$x_%i$'%l) # annotate merged clusters for l in np.unique(labels_am): maps = np.unique(labels[labels_am==l]) if len(maps) > 1: x1 = x[labels==maps[0]] x2 = x[labels==maps[1]] print(np.median(x1)) print(np.median(x2)) print(gmax) ax2.plot([np.median(x1), np.median(x2)], [1.2*gmax, 1.2*gmax], c='k', clip_on=False) ax2.plot([np.median(x1), np.median(x1)], [1.1*gmax, 1.2*gmax], c='k', clip_on=False) ax2.plot([np.median(x2), np.median(x2)], [1.1*gmax, 1.2*gmax], c='k', clip_on=False) ax2.annotate(r'$\frac{|{\tilde{x}_%i-\tilde{x}_%i}|}{max(\tilde{x}_%i, \tilde{x}_%i)} < \epsilon$' % (maps[0], maps[1], maps[0], maps[1]), [np.median(x1)*1.1, gmax*1.2], xytext=(10, 10), textcoords='offset points', fontsize=12, annotation_clip=False, ha='center') # add legends and plot. ax2.legend(loc='lower left', frameon=False, bbox_to_anchor=(-0.05, 1.3), ncol=len(np.unique(classes))) ax1.legend(loc='upper left', frameon=False, bbox_to_anchor=(-0.05, 1.3), ncol=len(np.unique(labels))) plt.tight_layout() def plot_feature_extraction(raw_snippets, normalized_snippets, features, labels, dt, pt)-
Plot clustering step on EOD shape.
Parameters
raw_snippets:2D numpy array- Raw EOD snippets.
normalized_snippets:2D numpy array- Normalized EOD snippets.
features:2D numpy array- PCA values for each normalized EOD snippet.
labels:1D numpy arrayofints- Cluster labels.
dt:float- Sample interval of snippets.
pt:int- Set to 0 for peak-centered EODs and set to 1 for trough-centered EODs.
Expand source code
def plot_feature_extraction(raw_snippets, normalized_snippets, features, labels, dt, pt): """Plot clustering step on EOD shape. Parameters ---------- raw_snippets : 2D numpy array Raw EOD snippets. normalized_snippets : 2D numpy array Normalized EOD snippets. features : 2D numpy array PCA values for each normalized EOD snippet. labels : 1D numpy array of ints Cluster labels. dt : float Sample interval of snippets. pt : int Set to 0 for peak-centered EODs and set to 1 for trough-centered EODs. """ ccol = cmap_pts[pt] # set up the figure layout fig = plt.figure(figsize=(((2+0.2)*4.8), 4.8)) outer = gridspec.GridSpec(1, 2, wspace=0.2, hspace=0) x = np.arange(-dt*1000*raw_snippets.shape[1]/2, dt*1000*raw_snippets.shape[1]/2, dt*1000) snip_ax = gridspec.GridSpecFromSubplotSpec(2, 1, subplot_spec = outer[0], hspace=0.35) pc_ax = gridspec.GridSpecFromSubplotSpec(features.shape[1]-1, features.shape[1]-1, subplot_spec = outer[1], hspace=0, wspace=0) # 3 plots: raw snippets, normalized, pcs. ax_raw_snip = fig.add_subplot(snip_ax[0]) ax_normalized_snip = fig.add_subplot(snip_ax[1]) colidxs = -np.linspace(-1.25, -0.5, len(np.unique(labels[labels>=0]))) j=0 for c in np.unique(labels): if c<0: color='lightgrey' else: color = lighter(ccol, colidxs[j]) j=j+1 ax_raw_snip.plot(x, raw_snippets[labels==c].T, color=color, label='-1', rasterized=True, alpha=0.25) ax_normalized_snip.plot(x, normalized_snippets[labels==c].T, color=color, alpha=0.25) ax_raw_snip.spines['top'].set_visible(False) ax_raw_snip.spines['right'].set_visible(False) ax_raw_snip.get_xaxis().set_ticklabels([]) ax_raw_snip.set_title('Raw snippets') ax_raw_snip.set_ylabel('Amplitude [a.u.]') ax_normalized_snip.spines['top'].set_visible(False) ax_normalized_snip.spines['right'].set_visible(False) ax_normalized_snip.set_title('Normalized snippets') ax_normalized_snip.set_ylabel('Amplitude [a.u.]') ax_normalized_snip.set_xlabel('Time [ms]') ax_raw_snip.axis('off') ax_normalized_snip.axis('off') ax_overlay = fig.add_subplot(pc_ax[:,:]) ax_overlay.set_title('Features') ax_overlay.axis('off') for n in range(features.shape[1]): for m in range(n): ax = fig.add_subplot(pc_ax[n-1,m]) ax.scatter(features[labels==c,m], features[labels==c,n], marker='.', color=color, alpha=0.25) ax.set_xlim(np.min(features), np.max(features)) ax.set_ylim(np.min(features), np.max(features)) ax.get_xaxis().set_ticklabels([]) ax.get_yaxis().set_ticklabels([]) ax.get_xaxis().set_ticks([]) ax.get_yaxis().set_ticks([]) if m==0: ax.set_ylabel('PC %i'%(n+1)) if n==features.shape[1]-1: ax.set_xlabel('PC %i'%(m+1)) ax = fig.add_subplot(pc_ax[0,features.shape[1]-2]) ax.set_xlim(np.min(features), np.max(features)) ax.set_ylim(np.min(features), np.max(features)) size = max(1, int(np.ceil(-np.log10(np.max(features)-np.min(features))))) wbar = np.floor((np.max(features)-np.min(features))*10**size)/10**size # should be smaller than the actual thing! so like x% of it? xscalebar(ax, 0, 0, wbar, wformat='%%.%if'%size) yscalebar(ax, 0, 0, wbar, hformat='%%.%if'%size) ax.axis('off') def plot_moving_fish(ws, dts, clusterss, ts, fishcounts, T, ignore_stepss)-
Plot moving fish detection step.
Parameters
ws:listoffloats- Median width for each width cluster that the moving fish algorithm is computed on
- (in seconds).
dts:listoffloats- Sliding window size (in seconds) for each width cluster.
clusterss:listof1D numpy int arrays- Cluster labels for each EOD cluster in a width cluster.
ts:listof1D numpy float arrays- EOD emission times for each EOD in a width cluster.
fishcounts:listoflists- Sliding window timepoints and fishcounts for each width cluster.
T:float- Lenght of analyzed recording in seconds.
ignore_stepss:listof1D int arrays- Mask for fishcounts that were ignored (ignored if True) in the moving_fish analysis.
Expand source code
def plot_moving_fish(ws, dts, clusterss, ts, fishcounts, T, ignore_stepss): """Plot moving fish detection step. Parameters ---------- ws : list of floats Median width for each width cluster that the moving fish algorithm is computed on (in seconds). dts : list of floats Sliding window size (in seconds) for each width cluster. clusterss : list of 1D numpy int arrays Cluster labels for each EOD cluster in a width cluster. ts : list of 1D numpy float arrays EOD emission times for each EOD in a width cluster. fishcounts : list of lists Sliding window timepoints and fishcounts for each width cluster. T : float Lenght of analyzed recording in seconds. ignore_stepss : list of 1D int arrays Mask for fishcounts that were ignored (ignored if True) in the moving_fish analysis. """ fig = plt.figure() # create gridspec outer = gridspec.GridSpec(len(ws), 1) for i, (w, dt, clusters, t, fishcount, ignore_steps) in enumerate(zip(ws, dts, clusterss, ts, fishcounts, ignore_stepss)): gs = gridspec.GridSpecFromSubplotSpec(2, 1, subplot_spec = outer[i]) # axis for clusters ax1 = fig.add_subplot(gs[0]) # axis for fishcount ax2 = fig.add_subplot(gs[1]) # plot clusters as eventplot for cnum, c in enumerate(np.unique(clusters[clusters>=0])): ax1.eventplot(t[clusters==c], lineoffsets=cnum, linelengths=0.5, color=cmap(i)) cnum = cnum + 1 # Plot the sliding window rect=Rectangle((0, -0.5), dt, cnum, linewidth=1, linestyle='--', edgecolor='k', facecolor='none', clip_on=False) ax1.add_patch(rect) ax1.arrow(dt+0.1, -0.5, 0.5, 0, head_width=0.1, head_length=0.1, facecolor='k', edgecolor='k') # plot parameters ax1.set_title(r'$\tilde{w}_%i = %.3f ms$'%(i, 1000*w)) ax1.set_ylabel('cluster #') ax1.set_yticks(range(0, cnum)) ax1.set_xlabel('time') ax1.set_xlim(0, T) ax1.axes.xaxis.set_visible(False) ax1.spines['bottom'].set_visible(False) ax1.spines['top'].set_visible(False) ax1.spines['right'].set_visible(False) ax1.spines['left'].set_visible(False) # plot for fishcount x = fishcount[0] y = fishcount[1] ax2 = fig.add_subplot(gs[1]) ax2.spines['top'].set_visible(False) ax2.spines['right'].set_visible(False) ax2.spines['bottom'].set_visible(False) ax2.axes.xaxis.set_visible(False) yplot = np.copy(y) ax2.plot(x+dt/2, yplot, linestyle='-', marker='.', c=cmap(i), alpha=0.25) yplot[ignore_steps.astype(bool)] = np.NaN ax2.plot(x+dt/2, yplot, linestyle='-', marker='.', c=cmap(i)) ax2.set_ylabel('Fish count') ax2.set_yticks(range(int(np.min(y)), 1+int(np.max(y)))) ax2.set_xlim(0, T) if i < len(ws)-1: ax2.axes.xaxis.set_visible(False) else: ax2.axes.xaxis.set_visible(False) xscalebar(ax2, 1, 0, 1, wunit='s', ha='right') con = ConnectionPatch([0, -0.5], [dt/2, y[0]], "data", "data", axesA=ax1, axesB=ax2, color='k') ax2.add_artist(con) con = ConnectionPatch([dt, -0.5], [dt/2, y[0]], "data", "data", axesA=ax1, axesB=ax2, color='k') ax2.add_artist(con) plt.xlim(0, T)