j_hartling/paper_2025
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Holiday syncing :)

Browse Source
This commit is contained in:
j-hartling
2026-04-02 16:00:56 +02:00
parent 298969a067
commit 0b9264b1e1
14 changed files with 627 additions and 667 deletions
Download Patch File Download Diff File Expand all files Collapse all files

BIN
figures/fig_invariance_full.pdf
View File

Binary file not shown.

BIN
figures/fig_invariance_log_hp.pdf
View File

Binary file not shown.

204
python/fig_invariance_full.py
View File

@@ -4,9 +4,11 @@ import numpy as np
import matplotlib.pyplot as plt import matplotlib.pyplot as plt
from itertools import product from itertools import product
from thunderhopper.modeltools import load_data from thunderhopper.modeltools import load_data
from misc_functions import get_saturation
from color_functions import load_colors from color_functions import load_colors
from plot_functions import hide_axis, ylimits, xlabel, ylabel, title_subplot,\ from plot_functions import hide_axis, ylimits, xlabel, ylabel, title_subplot,\
plot_line, plot_barcode, strip_zeros, time_bar, super_xlabel plot_line, plot_barcode, strip_zeros, time_bar,\
letter_subplot, letter_subplots
from IPython import embed from IPython import embed
def plot_snippets(axes, time, snippets, ymin=None, ymax=None, **kwargs): def plot_snippets(axes, time, snippets, ymin=None, ymax=None, **kwargs):
@@ -20,52 +22,70 @@ def plot_bi_snippets(axes, time, snippets, **kwargs):
plot_barcode(ax, time, snippets[:, ..., i], **kwargs) plot_barcode(ax, time, snippets[:, ..., i], **kwargs)
return None return None
def plot_curves(ax, scales, measures, fill_kwargs={}, **kwargs):
if measures.ndim == 1:
ax.plot(scales, measures, **kwargs)[0]
return measures
median_measure = np.median(measures, axis=1)
spread_measure = [np.percentile(measures, 25, axis=1),
np.percentile(measures, 75, axis=1)]
ax.plot(scales, median_measure, **kwargs)[0]
ax.fill_between(scales, *spread_measure, **fill_kwargs)
return median_measure
def show_saturation(ax, scales, measures, high=0.95, **kwargs):
high_ind = get_saturation(measures, high=high)[1]
return ax.plot(scales[high_ind], 0, transform=ax.get_xaxis_transform(),
marker='o', ms=10, zorder=6, clip_on=False, **kwargs)
# GENERAL SETTINGS: # GENERAL SETTINGS:
target = 'Omocestus_rufipes' target = 'Omocestus_rufipes'
data_paths = glob.glob(f'../data/inv/full/{target}*.npz') data_paths = glob.glob(f'../data/inv/full/{target}*.npz')
stages = ['raw', 'filt', 'env', 'log', 'inv', 'conv', 'bi', 'feat'] ref_path = '../data/inv/full/ref_measures.npz'
save_path = '../figures/fig_invariance_full.pdf'
stages = ['filt', 'env', 'log', 'inv', 'conv', 'feat']
load_kwargs = dict( load_kwargs = dict(
files=stages, files=stages,
keywords=['scales', 'snip', 'measure'] keywords=['scales', 'snip', 'measure']
) )
save_path = '../figures/fig_invariance_full.pdf'
# GRAPH SETTINGS: # GRAPH SETTINGS:
fig_kwargs = dict( fig_kwargs = dict(
figsize=(32/2.54, 16/2.54), figsize=(32/2.54, 20/2.54),
) )
super_grid_kwargs = dict( super_grid_kwargs = dict(
nrows=1, nrows=2,
ncols=3, ncols=1,
wspace=0, wspace=0,
hspace=0, hspace=0,
left=0, left=0,
right=1, right=1,
bottom=0, bottom=0,
top=1 top=1,
height_ratios=[3, 2]
) )
subfig_specs = dict( subfig_specs = dict(
snip=(slice(None), slice(0, -1)), snip=(0, 0),
big=(slice(None), -1), big=(1, 0),
) )
snip_grid_kwargs = dict( snip_grid_kwargs = dict(
nrows=len(stages), nrows=len(stages),
ncols=None, ncols=None,
wspace=0.1, wspace=0.1,
hspace=0.4, hspace=0.4,
left=0.15, left=0.08,
right=0.95, right=0.95,
bottom=0.08, bottom=0.08,
top=0.95 top=0.95
) )
big_grid_kwargs = dict( big_grid_kwargs = dict(
nrows=1, nrows=1,
ncols=1, ncols=3,
wspace=0, wspace=0.2,
hspace=0, hspace=0,
left=0.2, left=snip_grid_kwargs['left'],
right=0.96, right=0.96,
bottom=0.08, bottom=0.2,
top=0.95 top=0.95
) )
@@ -79,9 +99,7 @@ fs = dict(
bar=16, bar=16,
) )
colors = load_colors('../data/stage_colors.npz') colors = load_colors('../data/stage_colors.npz')
colors['raw'] = "#000000"
lw = dict( lw = dict(
raw=0.25,
filt=0.25, filt=0.25,
env=0.25, env=0.25,
log=0.25, log=0.25,
@@ -92,25 +110,17 @@ lw = dict(
big=3 big=3
) )
xlabels = dict( xlabels = dict(
snip='time [s]',
big='scale $\\alpha$', big='scale $\\alpha$',
) )
ylabels = dict( ylabels = dict(
raw='$x$',
filt='$x_{\\text{filt}}$', filt='$x_{\\text{filt}}$',
env='$x_{\\text{env}}$', env='$x_{\\text{env}}$',
log='$x_{\\text{log}}$', log='$x_{\\text{db}}$',
inv='$x_{\\text{inv}}$', inv='$x_{\\text{inv}}$',
conv='$c_i$', conv='$c_i$',
bi='$b_i$', bi='$b_i$',
feat='$f_i$', feat='$f_i$',
big='norm. intensity measure' big=['intensity', 'rel. intensity', 'norm. intensity']
)
xlab_snip_kwargs = dict(
y=0,
fontsize=fs['lab_norm'],
ha='center',
va='bottom',
) )
xlab_big_kwargs = dict( xlab_big_kwargs = dict(
y=0, y=0,
@@ -126,18 +136,17 @@ ylab_snip_kwargs = dict(
va='center' va='center'
) )
ylab_big_kwargs = dict( ylab_big_kwargs = dict(
x=0, x=-0.12,
fontsize=fs['lab_norm'], fontsize=fs['lab_norm'],
ha='center', ha='center',
va='top', va='bottom',
) )
yloc = dict( yloc = dict(
raw=500, filt=3000,
filt=500, env=1000,
env=250, log=50,
log=25, inv=20,
inv=10, conv=2,
conv=1,
feat=1, feat=1,
) )
title_kwargs = dict( title_kwargs = dict(
@@ -148,20 +157,18 @@ title_kwargs = dict(
fontsize=fs['tit_norm'], fontsize=fs['tit_norm'],
) )
letter_snip_kwargs = dict( letter_snip_kwargs = dict(
x=0.02, x=0,
y=1, yref=0.5,
ha='left', ha='left',
va='top', va='center',
fontsize=fs['letter'], fontsize=fs['letter'],
fontweight='bold'
) )
letter_big_kwargs = dict( letter_big_kwargs = dict(
x=0, x=0,
y=1, y=1,
ha='left', ha='left',
va='top', va='bottom',
fontsize=fs['letter'], fontsize=fs['letter'],
fontweight='bold'
) )
bar_time = 5 bar_time = 5
bar_kwargs = dict( bar_kwargs = dict(
@@ -181,6 +188,8 @@ bar_kwargs = dict(
) )
) )
# PREPARATION:
ref_data = dict(np.load(ref_path))
# EXECUTION: # EXECUTION:
for data_path in data_paths: for data_path in data_paths:
@@ -188,7 +197,7 @@ for data_path in data_paths:
# Load invariance data: # Load invariance data:
data, config = load_data(data_path, **load_kwargs) data, config = load_data(data_path, **load_kwargs)
t_full = np.arange(data['snip_raw'].shape[0]) / config['rate'] t_full = np.arange(data['snip_filt'].shape[0]) / config['rate']
# Adjust grid parameters: # Adjust grid parameters:
snip_grid_kwargs['ncols'] = data['example_scales'].size snip_grid_kwargs['ncols'] = data['example_scales'].size
@@ -204,78 +213,91 @@ for data_path in data_paths:
for i, j in product(range(snip_grid.nrows), range(snip_grid.ncols)): for i, j in product(range(snip_grid.nrows), range(snip_grid.ncols)):
ax = snip_subfig.add_subplot(snip_grid[i, j]) ax = snip_subfig.add_subplot(snip_grid[i, j])
ax.set_xlim(t_full[0], t_full[-1]) ax.set_xlim(t_full[0], t_full[-1])
ax.yaxis.set_major_locator(plt.MultipleLocator(yloc[stages[i]]))
hide_axis(ax, 'bottom') hide_axis(ax, 'bottom')
if i == 0: if i == 0:
title = f'$\\alpha={strip_zeros(data["example_scales"][j])}$' title = title_subplot(ax, f'$\\alpha={strip_zeros(data["example_scales"][j])}$',
title_subplot(ax, title, ref=snip_subfig, **title_kwargs) ref=snip_subfig, **title_kwargs)
if j == 0: if j == 0:
ylabel(ax, ylabels[stages[i]], **ylab_snip_kwargs, transform=snip_subfig.transSubfigure) ylabel(ax, ylabels[stages[i]], **ylab_snip_kwargs, transform=snip_subfig.transSubfigure)
else: else:
hide_axis(ax, 'left') hide_axis(ax, 'left')
if stages[i] != 'bi':
ax.yaxis.set_major_locator(plt.MultipleLocator(yloc[stages[i]]))
snip_axes[i, j] = ax snip_axes[i, j] = ax
time_bar(snip_axes[-1, -1], **bar_kwargs) time_bar(snip_axes[-1, -1], **bar_kwargs)
letter_subplot(snip_subfig, 'a', ref=title, **letter_snip_kwargs)
# Prepare single analysis axis: # Prepare analysis axes:
big_subfig = fig.add_subfigure(super_grid[subfig_specs['big']]) big_subfig = fig.add_subfigure(super_grid[subfig_specs['big']])
big_grid = big_subfig.add_gridspec(**big_grid_kwargs) big_grid = big_subfig.add_gridspec(**big_grid_kwargs)
big_ax = big_subfig.add_subplot(big_grid[0, 0]) big_axes = np.zeros((big_grid.ncols,), dtype=object)
big_ax.set_xlim(data['scales'].min(), data['scales'].max()) for i in range(big_grid.ncols):
big_ax.set_xscale('symlog', linthresh=data['scales'][1], linscale=0.5) ax = big_subfig.add_subplot(big_grid[0, i])
big_ax.set_yscale('symlog', linthresh=0.01, linscale=0.1) ax.set_xlim(data['scales'][0], data['scales'][-1])
xlabel(big_ax, xlabels['big'], **xlab_big_kwargs, transform=big_subfig.transSubfigure) ax.set_xscale('symlog', linthresh=data['scales'][1], linscale=0.5)
ylabel(big_ax, ylabels['big'], **ylab_big_kwargs, transform=big_subfig.transSubfigure) ax.set_yscale('symlog', linthresh=0.01, linscale=0.1)
xlabel(ax, xlabels['big'], transform=big_subfig, **xlab_big_kwargs)
ylabel(ax, ylabels['big'][i], **ylab_big_kwargs)
big_axes[i] = ax
letter_subplots(big_axes, 'bc', **letter_big_kwargs)
# Plot raw snippets: # # Plot filtered snippets:
plot_snippets(snip_axes[0, :], t_full, data['snip_raw'], # plot_snippets(snip_axes[0, :], t_full, data['snip_filt'],
c=colors['raw'], lw=lw['raw']) # c=colors['filt'], lw=lw['filt'])
# Plot filtered snippets: # # Plot envelope snippets:
plot_snippets(snip_axes[1, :], t_full, data['snip_filt'], # plot_snippets(snip_axes[1, :], t_full, data['snip_env'],
c=colors['filt'], lw=lw['filt']) # ymin=0, c=colors['env'], lw=lw['env'])
# Plot envelope snippets: # # Plot logarithmic snippets:
plot_snippets(snip_axes[2, :], t_full, data['snip_env'], # plot_snippets(snip_axes[2, :], t_full, data['snip_log'],
ymin=0, c=colors['env'], lw=lw['env']) # c=colors['log'], lw=lw['log'])
# Plot logarithmic snippets: # # Plot invariant snippets:
plot_snippets(snip_axes[3, :], t_full, data['snip_log'], # plot_snippets(snip_axes[3, :], t_full, data['snip_inv'],
ymax=0, c=colors['log'], lw=lw['log']) # c=colors['inv'], lw=lw['inv'])
# Plot invariant snippets: # # Plot kernel response snippets:
plot_snippets(snip_axes[4, :], t_full, data['snip_inv'], # plot_snippets(snip_axes[4, :], t_full, data['snip_conv'],
c=colors['inv'], lw=lw['inv']) # c=colors['conv'], lw=lw['conv'])
# Plot kernel response snippets: # # Plot feature snippets:
plot_snippets(snip_axes[5, :], t_full, data['snip_conv'], # plot_snippets(snip_axes[5, :], t_full, data['snip_feat'],
c=colors['conv'], lw=lw['conv']) # ymin=0, ymax=1, c=colors['feat'], lw=lw['feat'])
# Plot binary snippets:
plot_bi_snippets(snip_axes[6, :], t_full, data['snip_bi'],
color=colors['bi'], lw=lw['bi'])
# Plot feature snippets:
plot_snippets(snip_axes[7, :], t_full, data['snip_feat'],
ymin=0, ymax=1, c=colors['feat'], lw=lw['feat'])
# Analysis results: # Analysis results:
scales_rel = data['scales'] - data['scales'][0]
scales_rel /= scales_rel[-1]
for stage in stages: for stage in stages:
key = f'measure_{stage}' measure = data[f'measure_{stage}']
if stage == 'bi':
continue
# Min-max normalization:
base_ind = np.argmin(data['scales'])
data[key] -= data[key][base_ind, ...]
data[key] /= data[key].max(axis=0)
# Condense measure: # Plot unmodified intensity measures:
if stage in ['conv', 'feat']: curve = plot_curves(big_axes[0], data['scales'], measure, c=colors[stage], lw=lw['big'],
data[key] = np.nanmedian(data[key], axis=1) fill_kwargs=dict(color=colors[stage], alpha=0.25))
if stage in ['log', 'inv', 'conv', 'feat']:
show_saturation(big_axes[0], data['scales'], curve, c=colors[stage])
# Plot measure over scales: # # Relate to pure-noise reference:
big_ax.plot(data['scales'], data[key], # norm_measure = measure / ref_data[stage]
c=colors[stage], lw=lw['big'])
# # Plot noise-related intensity measures:
# big_axes[1].plot(data['scales'], norm_measure, c=colors[stage], lw=lw['big'])
# Normalize measure to [0, 1]:
min_measure = measure.min(axis=0)
max_measure = measure.max(axis=0)
norm_measure = (measure - min_measure) / (max_measure - min_measure)
# Plot normalized intensity measures:
curve = plot_curves(big_axes[1], data['scales'], norm_measure, c=colors[stage], lw=lw['big'],
fill_kwargs=dict(color=colors[stage], alpha=0.25))
if stage in ['log', 'inv', 'conv', 'feat']:
show_saturation(big_axes[1], data['scales'], curve, c=colors[stage])
# # Plot over relative scales:
# plot_curves(big_axes[2], scales_rel, norm_measure, c=colors[stage], lw=lw['big'],
# fill_kwargs=dict(color=colors[stage], alpha=0.25))
# scales_rel = curve - curve.min()
# scales_rel /= scales_rel.max()
if save_path is not None: if save_path is not None:
fig.savefig(save_path) fig.savefig(save_path)

268
python/fig_invariance_log-hp.py
View File

@@ -4,10 +4,11 @@ import matplotlib.pyplot as plt
from itertools import product from itertools import product
from thunderhopper.filetools import search_files from thunderhopper.filetools import search_files
from thunderhopper.modeltools import load_data from thunderhopper.modeltools import load_data
from misc_functions import shorten_species, get_kde, get_saturation
from color_functions import load_colors from color_functions import load_colors
from plot_functions import hide_axis, ylimits, xlabel, ylabel, hide_ticks,\ from plot_functions import hide_axis, ylimits, super_xlabel, ylabel, hide_ticks,\
plot_line, strip_zeros, time_bar, zoom_inset,\ plot_line, strip_zeros, time_bar, zoom_inset,\
letter_subplot, title_subplot letter_subplot, letter_subplots, title_subplot
from IPython import embed from IPython import embed
def add_snip_axes(fig, grid_kwargs): def add_snip_axes(fig, grid_kwargs):
@@ -26,39 +27,86 @@ def plot_snippets(axes, time, snippets, ymin=None, ymax=None, **kwargs):
handles.extend(plot_line(ax, time, snippet, ymin=ymin, ymax=ymax, **kwargs)) handles.extend(plot_line(ax, time, snippet, ymin=ymin, ymax=ymax, **kwargs))
return handles return handles
def plot_dist_shifted(ax, data, axis, pdf=None, sigma=0.1, which='x',
base=None, cap=None, add_pdf=False, shifted=False, **kwargs):
if pdf is None:
pdf, axis = get_kde(data, sigma, axis)
if base is None:
base = pdf.min()
if cap is None:
cap = pdf.max()
pdf = (pdf - pdf.min()) / (pdf.max() - pdf.min()) * (cap - base) + base
if which == 'x':
transform = ax.get_xaxis_transform()
elif which == 'y':
transform = ax.get_yaxis_transform()
else:
transform = ax.transData
rng = np.random.default_rng()
handles = []
for value in data:
ind = np.nonzero(axis == value)[0][0]
offset = base if not shifted else rng.uniform(base, pdf[ind])
variables = (offset, value) if which=='y' else (value, offset)
handles.extend(ax.plot(*variables, transform=transform, **kwargs))
if add_pdf:
variables = (pdf, axis) if which=='y' else (axis, pdf)
pdf_handle = ax.plot(*variables, transform=transform, c='k', lw=1)
return handles, pdf_handle
return handles
def zalpha(handles, background='w', down=1):
twins = []
for handle in handles:
twin = handle.copy()
twin.set(color=background, alpha=1)
twin.set_zorder(handle.get_zorder() - down)
twins.append(twin)
return twins
# GENERAL SETTINGS: # GENERAL SETTINGS:
target = 'Omocestus_rufipes' target = 'Omocestus_rufipes'
data_paths = search_files(target, excl='noise', dir='../data/inv/log_hp/') data_paths = search_files(target, excl='noise', dir='../data/inv/log_hp/')
species_paths = search_files('*', incl='noise', dir='../data/inv/log_hp/') ref_path = '../data/inv/log_hp/ref_measures.npz'
save_path = '../figures/fig_invariance_log_hp.pdf'
target_species = [
'Omocestus_rufipes',
'Chorthippus_biguttulus',
'Chorthippus_mollis',
'Chrysochraon_dispar',
'Gomphocerippus_rufus',
'Pseudochorthippus_parallelus',
]
stages = ['env', 'log', 'inv'] stages = ['env', 'log', 'inv']
load_kwargs = dict( load_kwargs = dict(
files=stages, files=stages,
keywords=['scales', 'snip', 'measure'] keywords=['scales', 'snip', 'measure']
) )
save_path = '../figures/fig_invariance_log_hp.pdf'
compute_ratios = True compute_ratios = True
show_diag = True show_diag = True
show_noise = True show_plateaus = True
# GRAPH SETTINGS: # GRAPH SETTINGS:
fig_kwargs = dict( fig_kwargs = dict(
figsize=(32/2.54, 16/2.54), figsize=(32/2.54, 32/2.54),
) )
super_grid_kwargs = dict( super_grid_kwargs = dict(
nrows=2, nrows=3,
ncols=3, ncols=1,
wspace=0, wspace=0,
hspace=0, hspace=0,
left=0, left=0,
right=1, right=1,
bottom=0, bottom=0,
top=1 top=1,
height_ratios=[1, 1, 1]
) )
subfig_specs = dict( subfig_specs = dict(
pure=(0, slice(0, -1)), pure=(0, slice(None)),
noise=(1, slice(0, -1)), noise=(1, slice(None)),
big=(slice(None), -1), big=(2, slice(None)),
) )
block_height = 0.8 block_height = 0.8
edge_padding = 0.08 edge_padding = 0.08
@@ -67,7 +115,7 @@ pure_grid_kwargs = dict(
ncols=None, ncols=None,
wspace=0.1, wspace=0.1,
hspace=0.15, hspace=0.15,
left=0.16, left=0.11,
right=0.95, right=0.95,
bottom=1 - block_height - edge_padding, bottom=1 - block_height - edge_padding,
top=1 - edge_padding, top=1 - edge_padding,
@@ -76,23 +124,23 @@ pure_grid_kwargs = dict(
noise_grid_kwargs = dict( noise_grid_kwargs = dict(
nrows=len(stages), nrows=len(stages),
ncols=None, ncols=None,
wspace=0.1, wspace=pure_grid_kwargs['wspace'],
hspace=0.15, hspace=pure_grid_kwargs['hspace'],
left=0.16, left=pure_grid_kwargs['left'],
right=0.95, right=pure_grid_kwargs['right'],
bottom=edge_padding, bottom=edge_padding,
top=edge_padding + block_height, top=edge_padding + block_height,
height_ratios=[1, 2, 1] height_ratios=[1, 2, 1]
) )
big_grid_kwargs = dict( big_grid_kwargs = dict(
nrows=2, nrows=1,
ncols=1, ncols=3,
wspace=0, wspace=0.3,
hspace=0.3, hspace=0,
left=0.19, left=pure_grid_kwargs['left'],
right=0.96, right=pure_grid_kwargs['right'],
bottom=0.09, bottom=0.05,
top=0.98 top=1
) )
anchor_kwargs = dict( anchor_kwargs = dict(
aspect='equal', aspect='equal',
@@ -110,8 +158,14 @@ fs = dict(
bar=16, bar=16,
) )
colors = load_colors('../data/stage_colors.npz') colors = load_colors('../data/stage_colors.npz')
lw_snippets = 1 species_colors = load_colors('../data/species_colors.npz')
lw_big = 3 noise_colors = [(0.5, 0.5, 0.5), (0.7, 0.7, 0.7)]
lw = dict(
snip=1,
big=4,
spec=2,
plateau=1,
)
xlabels = dict( xlabels = dict(
big='scale $\\alpha$', big='scale $\\alpha$',
) )
@@ -135,7 +189,7 @@ ylab_snip_kwargs = dict(
va='center', va='center',
) )
ylab_big_kwargs = dict( ylab_big_kwargs = dict(
x=0.05, x=0,
fontsize=fs['lab_tex'], fontsize=fs['lab_tex'],
ha='center', ha='center',
va='top', va='top',
@@ -160,10 +214,10 @@ letter_snip_kwargs = dict(
fontsize=fs['letter'], fontsize=fs['letter'],
) )
letter_big_kwargs = dict( letter_big_kwargs = dict(
x=0.05, x=0,
yref=letter_snip_kwargs['yref'], y=1,
ha='left', ha='left',
va='center', va='bottom',
fontsize=fs['letter'], fontsize=fs['letter'],
) )
zoom_inset_bounds = [0.1, 0.2, 0.8, 0.6] zoom_inset_bounds = [0.1, 0.2, 0.8, 0.6]
@@ -204,33 +258,77 @@ bar_kwargs = dict(
va='center', va='center',
) )
) )
leg_kwargs = dict(
ncols=2,
loc='upper right',
bbox_to_anchor=(0, 0.6, 1, 0.4),
frameon=False,
prop=dict(
size=12,
style='italic',
),
borderpad=0,
borderaxespad=0,
handlelength=1,
columnspacing=1,
)
diag_kwargs = dict( diag_kwargs = dict(
c=(0.75, 0.75, 0.75), c=(0.75, 0.75, 0.75),
lw=2, lw=2,
ls='--', ls='--',
zorder=1.9, zorder=1.9,
) )
noise_rel_thresh = 0.95 plateau_settings = dict(
noise_kwargs = dict( low=0.05,
fc=(0.9, 0.9, 0.9), high=0.95,
first=True,
last=True,
condense=None,
)
plateau_rect_kwargs = dict(
ec='none', ec='none',
lw=0, lw=0,
zorder=1.5, zorder=1.5,
) )
plateau_line_kwargs = dict(
lw=lw['plateau'],
ls='--',
zorder=1,
)
plateau_dot_kwargs = dict(
marker='o',
markersize=10,
markeredgecolor='k',
markeredgewidth=1,
# alpha=1,
zorder=6,
clip_on=False,
# base=0,
# cap=0.15,
# add_pdf=True,
)
kde_kwargs = dict(
sigma=0.1,
)
# PREPARATION: # PREPARATION:
if compute_ratios: if compute_ratios:
ref_data = load_data('../data/processed/white_noise_sd-1.npz', files=stages)[0] ref_measures = dict(np.load(ref_path))
ref_measures = {k: v.std() for k, v in ref_data.items() if not k.endswith('rate')}
species_measures = [] species_measures = {}
for species_path in species_paths: thresh_inds = np.zeros((len(target_species),), dtype=int)
species_data, _ = load_data(species_path, **load_kwargs) thresh_scales = np.zeros((len(target_species),), dtype=float)
species_measure = species_data['measure_inv'] for i, species in enumerate(target_species):
path = search_files(species, incl='noise', dir='../data/inv/log_hp/')[0]
species_data = load_data(path, **load_kwargs)[0]
measure = species_data['measure_inv']
scales = species_data['scales']
if compute_ratios: if compute_ratios:
species_measure /= ref_measures['inv'] measure /= ref_measures['inv']
species_measures.append(species_measure) species_measures[species] = measure
species_measures = np.array(species_measures).T thresh_inds[i] = get_saturation(measure, **plateau_settings)[1]
thresh_scales[i] = scales[thresh_inds[i]]
thresh_pdf, pdf_axis = get_kde(thresh_scales, axis=scales, **kde_kwargs)
# EXECUTION: # EXECUTION:
for data_path in data_paths: for data_path in data_paths:
@@ -273,7 +371,7 @@ for data_path in data_paths:
transform=noise_subfig.transSubfigure) transform=noise_subfig.transSubfigure)
for ax, scale in zip(noise_axes[0, :], noise_data['example_scales']): for ax, scale in zip(noise_axes[0, :], noise_data['example_scales']):
noise_title = title_subplot(ax, f'$\\alpha={strip_zeros(scale)}$', **title_kwargs) noise_title = title_subplot(ax, f'$\\alpha={strip_zeros(scale)}$', **title_kwargs)
letter_subplot(noise_subfig, 'c', ref=noise_title, **letter_snip_kwargs) letter_subplot(noise_subfig, 'b', ref=noise_title, **letter_snip_kwargs)
noise_inset = noise_axes[0, 0].inset_axes(zoom_inset_bounds) noise_inset = noise_axes[0, 0].inset_axes(zoom_inset_bounds)
noise_inset.spines[:].set(visible=True, lw=zoom_kwargs['lw']) noise_inset.spines[:].set(visible=True, lw=zoom_kwargs['lw'])
noise_inset.tick_params(**inset_tick_kwargs) noise_inset.tick_params(**inset_tick_kwargs)
@@ -282,51 +380,49 @@ for data_path in data_paths:
# Prepare analysis axes: # Prepare analysis axes:
big_subfig = fig.add_subfigure(super_grid[subfig_specs['big']]) big_subfig = fig.add_subfigure(super_grid[subfig_specs['big']])
big_grid = big_subfig.add_gridspec(**big_grid_kwargs) big_grid = big_subfig.add_gridspec(**big_grid_kwargs)
big_axes = np.zeros((big_grid.nrows,), dtype=object) big_axes = np.zeros((big_grid.ncols,), dtype=object)
for i, scales in enumerate([pure_scales, noise_scales]): for i, scales in enumerate([pure_scales, noise_scales, noise_scales]):
ax = big_subfig.add_subplot(big_grid[i, 0]) ax = big_subfig.add_subplot(big_grid[0, i])
ax.set_xlim(scales[0], scales[-1]) ax.set_xlim(scales[0], scales[-1])
ax.set_ylim(scales[0], scales[-1]) ax.set_ylim(scales[0], scales[-1])
ax.set_xscale('symlog', linthresh=scales[1], linscale=0.5) ax.set_xscale('symlog', linthresh=scales[1], linscale=0.5)
ax.set_yscale('symlog', linthresh=scales[1], linscale=0.5) ax.set_yscale('symlog', linthresh=scales[1], linscale=0.5)
ax.set_aspect(**anchor_kwargs) ax.set_aspect(**anchor_kwargs)
ylabel(ax, ylabels['big'], transform=big_subfig.transSubfigure, **ylab_big_kwargs) if i > 0:
if i == 0: hide_ticks(ax, 'left')
hide_ticks(ax, 'bottom')
letter_subplot(big_subfig, 'b', ref=pure_title, **letter_big_kwargs)
else:
xlabel(ax, xlabels['big'], transform=big_subfig.transSubfigure, **xlab_big_kwargs)
letter_subplot(big_subfig, 'd', ref=noise_title, **letter_big_kwargs)
big_axes[i] = ax big_axes[i] = ax
ylabel(big_axes[0], ylabels['big'], transform=big_subfig.transSubfigure, **ylab_big_kwargs)
super_xlabel(xlabels['big'], big_subfig, big_axes[0], big_axes[-1], **xlab_big_kwargs)
letter_subplots(big_axes, 'cde', **letter_big_kwargs)
# Plot pure-song envelope snippets: # Plot pure-song envelope snippets:
handle = plot_snippets(pure_axes[0, :], t_full, pure_data['snip_env'], handle = plot_snippets(pure_axes[0, :], t_full, pure_data['snip_env'],
ymin=0, c=colors['env'], lw=lw_snippets)[0] ymin=0, c=colors['env'], lw=lw['snip'])[0]
zoom_inset(pure_axes[0, 0], pure_inset, handle, transform=pure_axes[0, 0].transAxes, **zoom_kwargs) zoom_inset(pure_axes[0, 0], pure_inset, handle, transform=pure_axes[0, 0].transAxes, **zoom_kwargs)
# Plot pure-song logarithmic snippets: # Plot pure-song logarithmic snippets:
plot_snippets(pure_axes[1, :], t_full, pure_data['snip_log'], plot_snippets(pure_axes[1, :], t_full, pure_data['snip_log'],
c=colors['log'], lw=lw_snippets) c=colors['log'], lw=lw['snip'])
# Plot pure-song invariant snippets: # Plot pure-song invariant snippets:
plot_snippets(pure_axes[2, :], t_full, pure_data['snip_inv'], plot_snippets(pure_axes[2, :], t_full, pure_data['snip_inv'],
c=colors['inv'], lw=lw_snippets) c=colors['inv'], lw=lw['snip'])
# Plot noise-song envelope snippets: # Plot noise-song envelope snippets:
ymin, ymax = pure_axes[0, 0].get_ylim() ymin, ymax = pure_axes[0, 0].get_ylim()
handle = plot_snippets(noise_axes[0, :], t_full, noise_data['snip_env'], handle = plot_snippets(noise_axes[0, :], t_full, noise_data['snip_env'],
ymin, ymax, c=colors['env'], lw=lw_snippets)[0] ymin, ymax, c=colors['env'], lw=lw['snip'])[0]
zoom_inset(noise_axes[0, 0], noise_inset, handle, transform=noise_axes[0, 0].transAxes, **zoom_kwargs) zoom_inset(noise_axes[0, 0], noise_inset, handle, transform=noise_axes[0, 0].transAxes, **zoom_kwargs)
# Plot noise-song logarithmic snippets: # Plot noise-song logarithmic snippets:
ymin, ymax = pure_axes[1, 0].get_ylim() ymin, ymax = pure_axes[1, 0].get_ylim()
plot_snippets(noise_axes[1, :], t_full, noise_data['snip_log'], plot_snippets(noise_axes[1, :], t_full, noise_data['snip_log'],
ymin, ymax, c=colors['log'], lw=lw_snippets) ymin, ymax, c=colors['log'], lw=lw['snip'])
# Plot noise-song invariant snippets: # Plot noise-song invariant snippets:
ymin, ymax = pure_axes[2, 0].get_ylim() ymin, ymax = pure_axes[2, 0].get_ylim()
plot_snippets(noise_axes[2, :], t_full, noise_data['snip_inv'], plot_snippets(noise_axes[2, :], t_full, noise_data['snip_inv'],
ymin, ymax, c=colors['inv'], lw=lw_snippets) ymin, ymax, c=colors['inv'], lw=lw['snip'])
# Indicate time scale: # Indicate time scale:
time_bar(noise_axes[-1, -1], **bar_kwargs) time_bar(noise_axes[-1, -1], **bar_kwargs)
@@ -342,34 +438,46 @@ for data_path in data_paths:
noise_data['measure_inv'] /= ref_measures['inv'] noise_data['measure_inv'] /= ref_measures['inv']
# Plot pure-song measures (ideal): # Plot pure-song measures (ideal):
big_axes[0].plot(pure_scales, pure_data['measure_env'], c=colors['env'], lw=lw_big) big_axes[0].plot(pure_scales, pure_data['measure_env'], c=colors['env'], lw=lw['big'])
big_axes[0].plot(pure_scales, pure_data['measure_log'], c=colors['log'], lw=lw_big) big_axes[0].plot(pure_scales, pure_data['measure_log'], c=colors['log'], lw=lw['big'])
big_axes[0].plot(pure_scales, pure_data['measure_inv'], c=colors['inv'], lw=lw_big) big_axes[0].plot(pure_scales, pure_data['measure_inv'], c=colors['inv'], lw=lw['big'])
# Plot noise-song measures (limited): # Plot noise-song measures (limited):
big_axes[1].plot(noise_scales, noise_data['measure_env'], c=colors['env'], lw=lw_big) big_axes[1].plot(noise_scales, noise_data['measure_env'], c=colors['env'], lw=lw['big'])
big_axes[1].plot(noise_scales, noise_data['measure_log'], c=colors['log'], lw=lw_big) big_axes[1].plot(noise_scales, noise_data['measure_log'], c=colors['log'], lw=lw['big'])
big_axes[1].plot(noise_scales, noise_data['measure_inv'], c=colors['inv'], lw=lw_big) big_axes[1].plot(noise_scales, noise_data['measure_inv'], c=colors['inv'], lw=lw['big'])
# Plot species measures:
big_axes[1].plot(noise_scales, species_measures, 'k', lw=lw_big, zorder=2.1)
if show_diag: if show_diag:
# Indicate diagonal: # Indicate diagonal:
big_axes[0].plot(pure_scales, pure_scales, **diag_kwargs) big_axes[0].plot(pure_scales, pure_scales, **diag_kwargs)
big_axes[1].plot(noise_scales, noise_scales, **diag_kwargs) big_axes[1].plot(noise_scales, noise_scales, **diag_kwargs)
if show_noise: if show_plateaus:
# Indicate noise floor: # Indicate low and high plateaus of noise invariance curve:
if compute_ratios: low_ind, high_ind = get_saturation(noise_data['measure_inv'], **plateau_settings)
span_measure = noise_data['measure_inv'][-1] - ref_measures['inv'] big_axes[1].axvspan(noise_scales[0], noise_scales[low_ind],
thresh_measure = ref_measures['inv'] + noise_rel_thresh * span_measure fc=noise_colors[0], **plateau_rect_kwargs)
else: big_axes[1].axvspan(noise_scales[low_ind], noise_scales[high_ind],
span_measure = noise_data['measure_inv'][-1] - noise_data['measure_inv'][0] fc=noise_colors[1], **plateau_rect_kwargs)
thresh_measure = noise_data['measure_inv'][0] + noise_rel_thresh * span_measure
thresh_ind = np.nonzero(noise_data['measure_inv'] < thresh_measure)[0][-1] # Plot species-specific noise-song measures:
thresh_scale = noise_scales[thresh_ind] for i, (species, measure) in enumerate(species_measures.items()):
big_axes[1].axvspan(noise_scales[0], thresh_scale, **noise_kwargs) color = species_colors[species]
ind, scale = thresh_inds[i], thresh_scales[i]
big_axes[2].plot(noise_scales, measure, label=shorten_species(species),
c=color, lw=lw['spec'])
big_axes[2].plot(scale, 0, c='w', alpha=1, zorder=5.5, **plateau_dot_kwargs,
transform=big_axes[2].get_xaxis_transform())
handle = big_axes[2].plot(scale, 0, c=color, alpha=0.5, **plateau_dot_kwargs,
transform=big_axes[2].get_xaxis_transform())
big_axes[2].vlines(scale, big_axes[2].get_ylim()[0], measure[ind],
color=color, **plateau_line_kwargs)
big_axes[2].legend(**leg_kwargs)
# handles = plot_dist_shifted(big_axes[2], species_threshs, axis=pdf_axis,
# pdf=thresh_pdf, **plateau_dot_kwargs)[0]
# [h.set_color(species_colors[s]) for h, s in zip(handles, target_species)]
if save_path is not None: if save_path is not None:
fig.savefig(save_path, bbox_inches='tight') fig.savefig(save_path, bbox_inches='tight')

407
python/fig_invariance_log-hp_backup.py
View File

@@ -1,407 +0,0 @@
import plotstyle_plt
import numpy as np
import matplotlib.pyplot as plt
from itertools import product
from thunderhopper.filetools import search_files
from thunderhopper.modeltools import load_data
from misc_functions import shorten_species, get_saturation
from color_functions import load_colors
from plot_functions import hide_axis, ylimits, super_xlabel, ylabel, hide_ticks,\
plot_line, strip_zeros, time_bar, zoom_inset,\
letter_subplot, letter_subplots, title_subplot
from IPython import embed
def add_snip_axes(fig, grid_kwargs):
grid = fig.add_gridspec(**grid_kwargs)
axes = np.zeros((grid.nrows, grid.ncols), dtype=object)
for i, j in product(range(grid.nrows), range(grid.ncols)):
axes[i, j] = fig.add_subplot(grid[i, j])
[hide_axis(ax, 'left') for ax in axes[:, 1:].flatten()]
[hide_axis(ax, 'bottom') for ax in axes.flatten()]
return axes
def plot_snippets(axes, time, snippets, ymin=None, ymax=None, **kwargs):
ymin, ymax = ylimits(snippets, minval=ymin, maxval=ymax, pad=0.05)
handles = []
for ax, snippet in zip(axes, snippets.T):
handles.extend(plot_line(ax, time, snippet, ymin=ymin, ymax=ymax, **kwargs))
return handles
# GENERAL SETTINGS:
target = 'Omocestus_rufipes'
data_paths = search_files(target, excl='noise', dir='../data/inv/log_hp/')
target_species = [
'Omocestus_rufipes',
'Chorthippus_biguttulus',
'Chorthippus_mollis',
'Chrysochraon_dispar',
'Gomphocerippus_rufus',
'Pseudochorthippus_parallelus',
]
stages = ['env', 'log', 'inv']
load_kwargs = dict(
files=stages,
keywords=['scales', 'snip', 'measure']
)
save_path = '../figures/fig_invariance_log_hp.pdf'
compute_ratios = True
show_diag = True
show_plateaus = True
# GRAPH SETTINGS:
fig_kwargs = dict(
figsize=(32/2.54, 32/2.54),
)
# snip_rows = 1
# big_rows = 1
super_grid_kwargs = dict(
nrows=3,
ncols=1,
wspace=0,
hspace=0,
left=0,
right=1,
bottom=0,
top=1,
height_ratios=[1, 1, 1]
)
subfig_specs = dict(
pure=(0, slice(None)),
noise=(1, slice(None)),
big=(2, slice(None)),
)
block_height = 0.8
edge_padding = 0.08
pure_grid_kwargs = dict(
nrows=len(stages),
ncols=None,
wspace=0.1,
hspace=0.15,
left=0.11,
right=0.95,
bottom=1 - block_height - edge_padding,
top=1 - edge_padding,
height_ratios=[1, 2, 1]
)
noise_grid_kwargs = dict(
nrows=len(stages),
ncols=None,
wspace=pure_grid_kwargs['wspace'],
hspace=pure_grid_kwargs['hspace'],
left=pure_grid_kwargs['left'],
right=pure_grid_kwargs['right'],
bottom=edge_padding,
top=edge_padding + block_height,
height_ratios=[1, 2, 1]
)
big_grid_kwargs = dict(
nrows=1,
ncols=3,
wspace=0.3,
hspace=0,
left=pure_grid_kwargs['left'],
right=pure_grid_kwargs['right'],
bottom=0.05,
top=1
)
anchor_kwargs = dict(
aspect='equal',
adjustable='box',
anchor=(0.5, 0.5)
)
# PLOT SETTINGS:
fs = dict(
lab_norm=16,
lab_tex=20,
letter=22,
tit_norm=16,
tit_tex=20,
bar=16,
)
colors = load_colors('../data/stage_colors.npz')
species_colors = load_colors('../data/species_colors.npz')
noise_colors = [(0.5, 0.5, 0.5), (0.7, 0.7, 0.7)]
lw_snippets = 1
lw_big = 3
xlabels = dict(
big='scale $\\alpha$',
)
ylabels = dict(
env='$x_{\\text{env}}$',
log='$x_{\\text{dB}}$',
inv='$x_{\\text{adapt}}$',
big='$\\sigma_{\\alpha}\\,/\\,\\sigma_{\\eta}$',
)
xlab_big_kwargs = dict(
y=0,
fontsize=fs['lab_norm'],
ha='center',
va='bottom',
)
ylab_snip_kwargs = dict(
x=0,
fontsize=fs['lab_tex'],
rotation=0,
ha='left',
va='center',
)
ylab_big_kwargs = dict(
x=0,
fontsize=fs['lab_tex'],
ha='center',
va='top',
)
yloc = dict(
env=1000,
log=40,
inv=20
)
title_kwargs = dict(
x=0.5,
y=1,
ha='center',
va='bottom',
fontsize=fs['tit_norm'],
)
letter_snip_kwargs = dict(
x=0,
yref=0.5,
ha='left',
va='center',
fontsize=fs['letter'],
)
letter_big_kwargs = dict(
x=0,
y=1,
ha='left',
va='bottom',
fontsize=fs['letter'],
)
zoom_inset_bounds = [0.1, 0.2, 0.8, 0.6]
zoom_kwargs = dict(
x0=0.45,
x1=0.55,
y0=0,
y1=0.0006,
low_left=True,
low_right=True,
ec='k',
lw=1,
alpha=1,
)
inset_tick_kwargs = dict(
axis='y',
length=3,
pad=1,
left=False,
labelleft=False,
right=True,
labelright=True,
)
bar_time = 5
bar_kwargs = dict(
dur=bar_time,
y0=-0.25,
y1=-0.1,
xshift=1,
color='k',
lw=0,
clip_on=False,
text_pos=(-0.1, 0.5),
text_str=f'${bar_time}\\,\\text{{s}}$',
text_kwargs=dict(
fontsize=fs['bar'],
ha='right',
va='center',
)
)
leg_kwargs = dict(
ncols=2,
loc='upper right',
bbox_to_anchor=(0, 0.6, 1, 0.4),
frameon=False,
prop=dict(
size=12,
style='italic',
),
borderpad=0,
borderaxespad=0,
handlelength=1,
columnspacing=1,
)
diag_kwargs = dict(
c=(0.75, 0.75, 0.75),
lw=2,
ls='--',
zorder=1.9,
)
plateau_settings = dict(
low=0.05,
high=0.95,
first=True,
last=True,
condense=None,
)
plateau_kwargs = dict(
ec='none',
lw=0,
zorder=1.5,
)
# PREPARATION:
if compute_ratios:
ref_data = load_data('../data/processed/white_noise_sd-1.npz', files=stages)[0]
ref_measures = {k: v.std() for k, v in ref_data.items() if not k.endswith('rate')}
species_measures = {}
for species in target_species:
path = search_files(species, incl='noise', dir='../data/inv/log_hp/')[0]
measure = load_data(path, **load_kwargs)[0]['measure_inv']
if compute_ratios:
measure /= ref_measures['inv']
species_measures[species] = measure
# EXECUTION:
for data_path in data_paths:
print(f'Processing {data_path}')
# Load invariance data:
pure_data, config = load_data(data_path, **load_kwargs)
noise_data, _ = load_data(data_path.replace('.npz', '_noise.npz'), **load_kwargs)
pure_scales, noise_scales = pure_data['scales'], noise_data['scales']
t_full = np.arange(pure_data['snip_env'].shape[0]) / config['env_rate']
# Prepare overall graph:
fig = plt.figure(**fig_kwargs)
super_grid = fig.add_gridspec(**super_grid_kwargs)
fig.canvas.draw()
# Prepare pure-song snippet axes:
pure_grid_kwargs['ncols'] = pure_data['example_scales'].size
pure_subfig = fig.add_subfigure(super_grid[subfig_specs['pure']])
pure_axes = add_snip_axes(pure_subfig, pure_grid_kwargs)
for ax, stage in zip(pure_axes[:, 0], stages):
ax.yaxis.set_major_locator(plt.MultipleLocator(yloc[stage]))
ylabel(ax, ylabels[stage], **ylab_snip_kwargs,
transform=pure_subfig.transSubfigure)
for ax, scale in zip(pure_axes[0, :], pure_data['example_scales']):
pure_title = title_subplot(ax, f'$\\alpha={strip_zeros(scale)}$', **title_kwargs)
letter_subplot(pure_subfig, 'a', ref=pure_title, **letter_snip_kwargs)
pure_inset = pure_axes[0, 0].inset_axes(zoom_inset_bounds)
pure_inset.spines[:].set(visible=True, lw=zoom_kwargs['lw'])
pure_inset.tick_params(**inset_tick_kwargs)
hide_ticks(pure_inset, 'bottom', ticks=False)
# Prepare noise-song snippet axes:
noise_grid_kwargs['ncols'] = noise_data['example_scales'].size
noise_subfig = fig.add_subfigure(super_grid[subfig_specs['noise']])
noise_axes = add_snip_axes(noise_subfig, noise_grid_kwargs)
for ax, stage in zip(noise_axes[:, 0], stages):
ax.yaxis.set_major_locator(plt.MultipleLocator(yloc[stage]))
ylabel(ax, ylabels[stage], **ylab_snip_kwargs,
transform=noise_subfig.transSubfigure)
for ax, scale in zip(noise_axes[0, :], noise_data['example_scales']):
noise_title = title_subplot(ax, f'$\\alpha={strip_zeros(scale)}$', **title_kwargs)
letter_subplot(noise_subfig, 'b', ref=noise_title, **letter_snip_kwargs)
noise_inset = noise_axes[0, 0].inset_axes(zoom_inset_bounds)
noise_inset.spines[:].set(visible=True, lw=zoom_kwargs['lw'])
noise_inset.tick_params(**inset_tick_kwargs)
hide_ticks(noise_inset, 'bottom', ticks=False)
# Prepare analysis axes:
big_subfig = fig.add_subfigure(super_grid[subfig_specs['big']])
big_grid = big_subfig.add_gridspec(**big_grid_kwargs)
big_axes = np.zeros((big_grid.ncols,), dtype=object)
for i, scales in enumerate([pure_scales, noise_scales, noise_scales]):
ax = big_subfig.add_subplot(big_grid[0, i])
ax.set_xlim(scales[0], scales[-1])
ax.set_ylim(scales[0], scales[-1])
ax.set_xscale('symlog', linthresh=scales[1], linscale=0.5)
ax.set_yscale('symlog', linthresh=scales[1], linscale=0.5)
ax.set_aspect(**anchor_kwargs)
big_axes[i] = ax
ylabel(big_axes[0], ylabels['big'], transform=big_subfig.transSubfigure, **ylab_big_kwargs)
super_xlabel(xlabels['big'], big_subfig, big_axes[0], big_axes[-1], **xlab_big_kwargs)
letter_subplots(big_axes, 'cde', **letter_big_kwargs)
# Plot pure-song envelope snippets:
handle = plot_snippets(pure_axes[0, :], t_full, pure_data['snip_env'],
ymin=0, c=colors['env'], lw=lw_snippets)[0]
zoom_inset(pure_axes[0, 0], pure_inset, handle, transform=pure_axes[0, 0].transAxes, **zoom_kwargs)
# Plot pure-song logarithmic snippets:
plot_snippets(pure_axes[1, :], t_full, pure_data['snip_log'],
c=colors['log'], lw=lw_snippets)
# Plot pure-song invariant snippets:
plot_snippets(pure_axes[2, :], t_full, pure_data['snip_inv'],
c=colors['inv'], lw=lw_snippets)
# Plot noise-song envelope snippets:
ymin, ymax = pure_axes[0, 0].get_ylim()
handle = plot_snippets(noise_axes[0, :], t_full, noise_data['snip_env'],
ymin, ymax, c=colors['env'], lw=lw_snippets)[0]
zoom_inset(noise_axes[0, 0], noise_inset, handle, transform=noise_axes[0, 0].transAxes, **zoom_kwargs)
# Plot noise-song logarithmic snippets:
ymin, ymax = pure_axes[1, 0].get_ylim()
plot_snippets(noise_axes[1, :], t_full, noise_data['snip_log'],
ymin, ymax, c=colors['log'], lw=lw_snippets)
# Plot noise-song invariant snippets:
ymin, ymax = pure_axes[2, 0].get_ylim()
plot_snippets(noise_axes[2, :], t_full, noise_data['snip_inv'],
ymin, ymax, c=colors['inv'], lw=lw_snippets)
# Indicate time scale:
time_bar(noise_axes[-1, -1], **bar_kwargs)
if compute_ratios:
# Relate pure-song measures to zero scale:
pure_data['measure_env'] /= ref_measures['env']
pure_data['measure_log'] /= ref_measures['log']
pure_data['measure_inv'] /= ref_measures['inv']
# Relate noise-song measures to zero scale:
noise_data['measure_env'] /= ref_measures['env']
noise_data['measure_log'] /= ref_measures['log']
noise_data['measure_inv'] /= ref_measures['inv']
# Plot pure-song measures (ideal):
big_axes[0].plot(pure_scales, pure_data['measure_env'], c=colors['env'], lw=lw_big)
big_axes[0].plot(pure_scales, pure_data['measure_log'], c=colors['log'], lw=lw_big)
big_axes[0].plot(pure_scales, pure_data['measure_inv'], c=colors['inv'], lw=lw_big)
# Plot noise-song measures (limited):
big_axes[1].plot(noise_scales, noise_data['measure_env'], c=colors['env'], lw=lw_big)
big_axes[1].plot(noise_scales, noise_data['measure_log'], c=colors['log'], lw=lw_big)
big_axes[1].plot(noise_scales, noise_data['measure_inv'], c=colors['inv'], lw=lw_big)
if show_diag:
# Indicate diagonal:
big_axes[0].plot(pure_scales, pure_scales, **diag_kwargs)
big_axes[1].plot(noise_scales, noise_scales, **diag_kwargs)
if show_plateaus:
# Indicate low and high plateaus of noise invariance curve:
low_ind, high_ind = get_saturation(noise_data['measure_inv'], **plateau_settings)
big_axes[1].axvspan(noise_scales[0], noise_scales[low_ind],
fc=noise_colors[0], **plateau_kwargs)
big_axes[1].axvspan(noise_scales[low_ind], noise_scales[high_ind],
fc=noise_colors[1], **plateau_kwargs)
# Plot species-specific noise-song measures:
for species, measure in species_measures.items():
label = shorten_species(species)
big_axes[2].plot(noise_scales, measure, label=label,
c=species_colors[species], lw=lw_big)
big_axes[2].legend(**leg_kwargs)
if save_path is not None:
fig.savefig(save_path, bbox_inches='tight')
plt.show()
print('Done.')
embed()

93
python/fig_invariance_thresh-lp_single.py
View File

@@ -5,6 +5,7 @@ from itertools import product
from thunderhopper.filetools import search_files from thunderhopper.filetools import search_files
from thunderhopper.modeltools import load_data from thunderhopper.modeltools import load_data
from thunderhopper.filtertools import find_kern_specs from thunderhopper.filtertools import find_kern_specs
from misc_functions import get_saturation
from color_functions import load_colors, shade_colors from color_functions import load_colors, shade_colors
from plot_functions import shift_subplot, hide_axis, ylimits, xlabel, ylabel,\ from plot_functions import shift_subplot, hide_axis, ylimits, xlabel, ylabel,\
super_ylabel, plot_line, plot_barcode, strip_zeros,\ super_ylabel, plot_line, plot_barcode, strip_zeros,\
@@ -64,7 +65,7 @@ load_kwargs = dict(
files=stages, files=stages,
keywords=['scales', 'snip', 'measure', 'thresh'] keywords=['scales', 'snip', 'measure', 'thresh']
) )
save_path = '../figures/fig_invariance_thresh_lp_single.pdf' save_path = None#'../figures/fig_invariance_thresh_lp_single.pdf'
exclude_zero = True exclude_zero = True
# GRAPH SETTINGS: # GRAPH SETTINGS:
@@ -79,7 +80,7 @@ super_grid_kwargs = dict(
left=0, left=0,
right=1, right=1,
bottom=0, bottom=0,
top=1 top=1,
) )
input_rows = 1 input_rows = 1
snip_rows = 2 snip_rows = 2
@@ -111,10 +112,10 @@ input_grid_kwargs = dict(
top=0.75, top=0.75,
) )
big_grid_kwargs = dict( big_grid_kwargs = dict(
nrows=1, nrows=2,
ncols=1, ncols=1,
wspace=0, wspace=0,
hspace=0, hspace=0.3,
left=0.17, left=0.17,
right=0.96, right=0.96,
bottom=0.1, bottom=0.1,
@@ -141,7 +142,8 @@ lw = dict(
big=4, big=4,
) )
xlabels = dict( xlabels = dict(
big='scale $\\alpha$', alpha='scale $\\alpha$',
sigma='$\\sigma_{\\text{adapt}}$',
) )
ylabels = dict( ylabels = dict(
inv='$x_{\\text{adapt}}$', inv='$x_{\\text{adapt}}$',
@@ -150,11 +152,17 @@ ylabels = dict(
feat='$f_i$', feat='$f_i$',
big='$\\mu_f$', big='$\\mu_f$',
) )
xlab_big_kwargs = dict( xlab_alpha_kwargs = dict(
y=0, y=-0.15,
fontsize=fs['lab_norm'], fontsize=fs['lab_norm'],
ha='center', ha='center',
va='bottom', va='top',
)
xlab_sigma_kwargs = dict(
y=-0.12,
fontsize=fs['lab_tex'],
ha=xlab_alpha_kwargs['ha'],
va=xlab_alpha_kwargs['va'],
) )
ylab_snip_kwargs = dict( ylab_snip_kwargs = dict(
x=0.08, x=0.08,
@@ -178,7 +186,7 @@ ylab_big_kwargs = dict(
ypad = dict( ypad = dict(
inv=0.05, inv=0.05,
conv=0.05, conv=0.05,
big=0.075 big=0.1
) )
yloc = dict( yloc = dict(
inv=(2, 200), inv=(2, 200),
@@ -242,6 +250,13 @@ leg_kwargs = dict(
handlelength=1.5, handlelength=1.5,
columnspacing=1, columnspacing=1,
) )
plateau_settings = dict(
low=0.05,
high=0.95,
first=True,
last=True,
condense=None,
)
kern_specs = np.array([ kern_specs = np.array([
[1, 0.008], [1, 0.008],
[2, 0.004], [2, 0.004],
@@ -281,6 +296,7 @@ for data_path in data_paths:
# Reduce to nonzero scales: # Reduce to nonzero scales:
nonzero_inds = scales > 0 nonzero_inds = scales > 0
scales = scales[nonzero_inds] scales = scales[nonzero_inds]
noise_data['measure_inv'] = noise_data['measure_inv'][nonzero_inds]
noise_data['measure_feat'] = noise_data['measure_feat'][nonzero_inds, :] noise_data['measure_feat'] = noise_data['measure_feat'][nonzero_inds, :]
pure_data['measure_feat'] = pure_data['measure_feat'][nonzero_inds, :] pure_data['measure_feat'] = pure_data['measure_feat'][nonzero_inds, :]
@@ -293,7 +309,7 @@ for data_path in data_paths:
) )
# Adjust grid parameters to loaded data: # Adjust grid parameters to loaded data:
super_grid_kwargs['nrows'] = snip_rows * thresh_rel.size + 1 super_grid_kwargs['nrows'] = snip_rows * thresh_rel.size + input_rows
input_grid_kwargs['ncols'] = plot_scales.size input_grid_kwargs['ncols'] = plot_scales.size
snip_grid_kwargs['ncols'] = plot_scales.size snip_grid_kwargs['ncols'] = plot_scales.size
@@ -325,8 +341,6 @@ for data_path in data_paths:
ax1.yaxis.set_major_locator(plt.MultipleLocator(yloc[stage][0])) ax1.yaxis.set_major_locator(plt.MultipleLocator(yloc[stage][0]))
ax2.yaxis.set_major_locator(plt.MultipleLocator(yloc[stage][1])) ax2.yaxis.set_major_locator(plt.MultipleLocator(yloc[stage][1]))
ylabel(ax1, ylabels[stage], transform=snip_subfig.transSubfigure, **ylab_snip_kwargs) ylabel(ax1, ylabels[stage], transform=snip_subfig.transSubfigure, **ylab_snip_kwargs)
# for ax, scale in zip(axes[0, :], plot_scales):
# title_subplot(ax, f'$\\alpha={strip_zeros(scale)}$', ref=snip_subfig, **title_kwargs)
if i == thresh_rel.size - 1: if i == thresh_rel.size - 1:
axes[-1, -1].set_xlim(t_full[0], t_full[-1]) axes[-1, -1].set_xlim(t_full[0], t_full[-1])
time_bar(axes[-1, -1], **bar_kwargs) time_bar(axes[-1, -1], **bar_kwargs)
@@ -334,17 +348,27 @@ for data_path in data_paths:
snip_axes.append(axes) snip_axes.append(axes)
letter_subplots(snip_subfigs, 'bcd', **letter_snip_kwargs) letter_subplots(snip_subfigs, 'bcd', **letter_snip_kwargs)
# Prepare analysis axis: # Prepare analysis axes:
big_subfig = fig.add_subfigure(super_grid[subfig_specs['big']]) big_subfig = fig.add_subfigure(super_grid[subfig_specs['big']])
big_grid = big_subfig.add_gridspec(**big_grid_kwargs) big_grid = big_subfig.add_gridspec(**big_grid_kwargs)
big_ax = big_subfig.add_subplot(big_grid[0, 0])
big_ax.set_xlim(scales[0], scales[-1]) alpha_ax = big_subfig.add_subplot(big_grid[0, 0])
big_ax.set_xscale('symlog', linthresh=scales[scales > 0][0], linscale=0.5) alpha_ax.set_xlim(scales[0], scales[-1])
ylimits(noise_data['measure_feat'], big_ax, minval=0, pad=ypad['big']) alpha_ax.set_xscale('symlog', linthresh=scales[scales > 0][0], linscale=0.5)
big_ax.yaxis.set_major_locator(plt.MultipleLocator(yloc['big'])) ylimits(pure_data['measure_feat'], alpha_ax, minval=0, pad=ypad['big'])
xlabel(big_ax, xlabels['big'], transform=big_subfig.transSubfigure, **xlab_big_kwargs) alpha_ax.yaxis.set_major_locator(plt.MultipleLocator(yloc['big']))
ylabel(big_ax, ylabels['big'], transform=big_subfig.transSubfigure, **ylab_big_kwargs) xlabel(alpha_ax, xlabels['alpha'], **xlab_alpha_kwargs)
letter_subplot(big_subfig, 'e', **letter_big_kwargs, ref=input_subfig) ylabel(alpha_ax, ylabels['big'], transform=big_subfig.transSubfigure, **ylab_big_kwargs)
sigma_ax = big_subfig.add_subplot(big_grid[1, 0])
sigma_ax.set_xlim(noise_data['measure_inv'].min(), noise_data['measure_inv'].max())
# sigma_ax.set_xscale('log')
sigma_ax.set_xlim(scales[0], scales[-1])
sigma_ax.set_xscale('symlog', linthresh=scales[scales > 0][0], linscale=0.5)
ylimits(pure_data['measure_feat'], sigma_ax, minval=0, pad=ypad['big'])
sigma_ax.yaxis.set_major_locator(plt.MultipleLocator(yloc['big']))
xlabel(sigma_ax, xlabels['sigma'], **xlab_sigma_kwargs)
ylabel(sigma_ax, ylabels['big'], transform=big_subfig.transSubfigure, **ylab_big_kwargs)
# Plot intensity-adapted snippets: # Plot intensity-adapted snippets:
plot_snippets(input_axes, t_full, noise_data['snip_inv'], plot_snippets(input_axes, t_full, noise_data['snip_inv'],
@@ -375,18 +399,25 @@ for data_path in data_paths:
ymin=0, ymax=1, c=shaded['feat'][i], lw=lw['feat']) ymin=0, ymax=1, c=shaded['feat'][i], lw=lw['feat'])
[set_clip_box(h[0], ax, bounds=[[0, -0.05], [1, 1.05]]) for h, ax in zip(handles, axes[2, :])] [set_clip_box(h[0], ax, bounds=[[0, -0.05], [1, 1.05]]) for h, ax in zip(handles, axes[2, :])]
# Plot pure-song analysis results: # Get threshold-specific saturation:
handles = big_ax.plot(scales, pure_data['measure_feat'], lw=lw['big'], ls='dotted') for i in range(thresh_rel.size):
[h.set_color(c) for h, c in zip(handles, shaded['feat'])] ind = get_saturation(noise_data['measure_feat'][:, i], **plateau_settings)[1]
# Plot noise-song analysis results: # Plot analysis results:
handles = big_ax.plot(scales, noise_data['measure_feat'], lw=lw['big']) for ax, x in zip([alpha_ax, sigma_ax], [scales, noise_data['measure_inv']]):
[h.set_color(c) for h, c in zip(handles, shaded['feat'])] # Plot pure-song analysis results:
handles = ax.plot(x, pure_data['measure_feat'], lw=lw['big'], ls='dotted')
[h.set_color(c) for h, c in zip(handles, shaded['feat'])]
# Add proxy legend: # Plot noise-song analysis results:
h1 = big_ax.plot([], [], c='k', lw=lw['big'], label='$\\alpha\\cdot s(t) + \\eta(t)$')[0] handles = ax.plot(x, noise_data['measure_feat'], lw=lw['big'])
h2 = big_ax.plot([], [], c='k', lw=lw['big'], ls='dotted', label='$\\alpha\\cdot s(t)$')[0] [h.set_color(c) for h, c in zip(handles, shaded['feat'])]
big_ax.legend(handles=[h1, h2], **leg_kwargs)
# Add proxy legend:
if ax == alpha_ax:
h1 = ax.plot([], [], c='k', lw=lw['big'], label='$\\alpha\\cdot s(t) + \\eta(t)$')[0]
h2 = ax.plot([], [], c='k', lw=lw['big'], ls='dotted', label='$\\alpha\\cdot s(t)$')[0]
ax.legend(handles=[h1, h2], **leg_kwargs)
if save_path is not None: if save_path is not None:
fig.savefig(save_path) fig.savefig(save_path)

131
python/fig_temp.py Normal file
View File

@@ -0,0 +1,131 @@
import glob
import numpy as np
import matplotlib.pyplot as plt
from thunderhopper.modeltools import load_data
from thunderhopper.filtertools import find_kern_specs
from misc_functions import unsort_unique
from color_functions import sample_cmap
from IPython import embed
# Settings:
target = 'Omocestus_rufipes'
data_path = glob.glob(f'../data/inv/full/{target}*.npz')[0]
stages = ['conv', 'feat']
load_kwargs = dict(
files=stages,
keywords=['scales', 'measure']
)
# Subset settings:
all_types = np.array([1, -1, 2, -2, 3, -3, 4, -4, 5, -5,
6, -6, 7, -7, 8, -8, 9, -9, 10, -10]).astype(float)
all_sigmas = np.array([0.001, 0.002, 0.004, 0.008, 0.016, 0.032])
kerns = None
types = np.array([1, -1, 2, -2, 3, -3, 4, -4, 5, -5,
6, -6, 7, -7, 8, -8, 9, -9, 10, -10])
sigmas = np.array([0.008, 0.016])
# Plot settings:
line_kwargs = dict(
linewidth=2,
)
median_kwargs = dict(
linewidth=4,
c='k',
linestyle='--'
)
# Load invariance data:
data, config = load_data(data_path, **load_kwargs)
scales = data['scales']
# Reduce to kernel subset:
if any(var is not None for var in [kerns, types, sigmas]):
subset_inds = find_kern_specs(config['k_specs'], kerns, types, sigmas)
data['measure_conv'] = data['measure_conv'][:, subset_inds]
data['measure_feat'] = data['measure_feat'][:, subset_inds]
config['kernels'] = config['kernels'][:, subset_inds]
config['k_specs'] = config['k_specs'][subset_inds, :]
kern_types = unsort_unique(config['k_specs'][:, 0])
kern_sigmas = unsort_unique(config['k_specs'][:, 1])
# Prepare colors:
type_colors = {t: c for t, c in zip(all_types, sample_cmap('turbo', all_types.size))}
sigma_colors = {s: c for s, c in zip(all_sigmas, sample_cmap('turbo', all_sigmas.size))}
# Prepare graph:
fig, axes = plt.subplots(2, 4, figsize=(16, 16), layout='constrained', sharex=True)
axes[0, 0].set_xlim(scales[0], scales[-1])
axes[0, 0].set_xscale('log')
axes[0, 0].set_ylabel('conv')
axes[1, 0].set_ylabel('feat')
# Condense across kernels:
median_conv = np.median(data['measure_conv'], axis=1)
median_feat = np.median(data['measure_feat'], axis=1)
# Coded by type:
leg_handles = []
for kern_type in kern_types:
color = type_colors[kern_type]
inds = find_kern_specs(config['k_specs'], types=kern_type)
leg_handles.append(axes[0, 0].plot(scales, data['measure_conv'][:, inds],
c=color, label=f'{kern_type}', **line_kwargs)[0])
axes[0, 0].plot(scales, median_conv, **median_kwargs)
axes[1, 0].plot(scales, data['measure_feat'][:, inds], c=color, **line_kwargs)
axes[1, 0].plot(scales, median_feat, **median_kwargs)
axes[0, 0].legend(handles=leg_handles, loc='upper left')
axes[0, 0].set_title('Coded by type')
# Coded by sigma:
leg_handles = []
for kern_sigma in kern_sigmas:
color = sigma_colors[kern_sigma]
inds = find_kern_specs(config['k_specs'], sigmas=kern_sigma)
leg_handles.append(axes[0, 1].plot(scales, data['measure_conv'][:, inds],
c=color, label=f'{kern_sigma}', **line_kwargs)[0])
axes[0, 1].plot(scales, median_conv, **median_kwargs)
axes[1, 1].plot(scales, data['measure_feat'][:, inds], c=color, **line_kwargs)
axes[1, 1].plot(scales, median_feat, **median_kwargs)
axes[0, 1].legend(handles=leg_handles, loc='upper left')
axes[0, 1].set_title('Coded by sigma')
# Normalize measures:
data['measure_conv'] -= data['measure_conv'].min(axis=0)
data['measure_conv'] /= data['measure_conv'].max(axis=0)
data['measure_feat'] -= data['measure_feat'].min(axis=0)
data['measure_feat'] /= data['measure_feat'].max(axis=0)
# Condense across kernels:
median_conv = np.median(data['measure_conv'], axis=1)
median_feat = np.median(data['measure_feat'], axis=1)
# Coded by type:
leg_handles = []
for kern_type in kern_types:
color = type_colors[kern_type]
inds = find_kern_specs(config['k_specs'], types=kern_type)
leg_handles.append(axes[0, 2].plot(scales, data['measure_conv'][:, inds],
c=color, label=f'{kern_type}', **line_kwargs)[0])
axes[0, 2].plot(scales, median_conv, **median_kwargs)
axes[1, 2].plot(scales, data['measure_feat'][:, inds], c=color, **line_kwargs)
axes[1, 2].plot(scales, median_feat, **median_kwargs)
axes[0, 2].legend(handles=leg_handles, loc='upper left')
axes[0, 2].set_title('Coded by type')
# Coded by sigma:
leg_handles = []
for kern_sigma in kern_sigmas:
color = sigma_colors[kern_sigma]
inds = find_kern_specs(config['k_specs'], sigmas=kern_sigma)
leg_handles.append(axes[0, 3].plot(scales, data['measure_conv'][:, inds],
c=color, label=f'{kern_sigma}', **line_kwargs)[0])
axes[0, 3].plot(scales, median_conv, **median_kwargs)
axes[1, 3].plot(scales, data['measure_feat'][:, inds], c=color, **line_kwargs)
axes[1, 3].plot(scales, median_feat, **median_kwargs)
axes[0, 3].legend(handles=leg_handles, loc='upper left')
axes[0, 3].set_title('Coded by sigma')
plt.show()
embed()

29
python/misc_functions.py
View File

@@ -1,9 +1,20 @@
import numpy as np import numpy as np
from scipy.stats import gaussian_kde
def shorten_species(name): def shorten_species(name):
genus, species = name.split('_') genus, species = name.split('_')
return genus[0] + '. ' + species return genus[0] + '. ' + species
def unsort_unique(array):
values, inds = np.unique(array, return_index=True)
return values[np.argsort(inds)]
def get_kde(data, sigma, axis=None, n=1000, pad=10):
if axis is None:
axis = np.linspace(data.min() - pad * sigma, data.max() + pad * sigma, n)
pdf = gaussian_kde(data, sigma)(axis)
return pdf, axis
def get_saturation(sigmoid, low=0.05, high=0.95, first=True, last=True, def get_saturation(sigmoid, low=0.05, high=0.95, first=True, last=True,
condense=None): condense=None):
if condense == 'norm' and sigmoid.ndim == 2: if condense == 'norm' and sigmoid.ndim == 2:
@@ -16,17 +27,17 @@ def get_saturation(sigmoid, low=0.05, high=0.95, first=True, last=True,
low_value = min_value + low * span low_value = min_value + low * span
high_value = min_value + high * span high_value = min_value + high * span
low_mask = sigmoid >= low_value low_mask = sigmoid <= low_value
high_mask = sigmoid >= high_value high_mask = sigmoid <= high_value
if sigmoid.ndim == 1: if sigmoid.ndim == 1:
low_ind = np.nonzero(low_mask)[0][0] low_ind = np.nonzero(low_mask)[0][-1]
high_ind = np.nonzero(high_mask)[0][0] high_ind = np.nonzero(high_mask)[0][-1]
elif condense == 'all': elif condense == 'all':
low_ind = np.nonzero(low_mask.all(axis=1))[0][0] low_ind = np.nonzero(low_mask.all(axis=1))[0][-1]
high_ind = np.nonzero(high_mask.all(axis=1))[0][0] high_ind = np.nonzero(high_mask.all(axis=1))[0][-1]
else: else:
low_ind, high_ind = [], [] low_ind, high_ind = [], []
for i in range(sigmoid.shape[1]): for i in range(sigmoid.shape[1]):
low_ind.append(np.nonzero(low_mask[:, i])[0][0]) low_ind.append(np.nonzero(low_mask[:, i])[0][-1])
high_ind.append(np.nonzero(high_mask[:, i])[0][0]) high_ind.append(np.nonzero(high_mask[:, i])[0][-1])
return low_ind, high_ind return low_ind, high_ind

67
python/save_inv_data_full.py
View File

@@ -4,19 +4,23 @@ import matplotlib.pyplot as plt
from thunderhopper.modeltools import load_data, save_data from thunderhopper.modeltools import load_data, save_data
from thunderhopper.filetools import crop_paths from thunderhopper.filetools import crop_paths
from thunderhopper.filtertools import find_kern_specs from thunderhopper.filtertools import find_kern_specs
from thunderhopper.model import process_signal from thunderhopper.model import process_signal, convolve_kernels
from IPython import embed from IPython import embed
# GENERAL SETTINGS: # GENERAL SETTINGS:
target = 'Omocestus_rufipes' target = 'Omocestus_rufipes'
data_paths = glob.glob(f'../data/processed/{target}*.npz') data_paths = glob.glob(f'../data/processed/{target}*.npz')
stages = ['raw', 'filt', 'env', 'log', 'inv', 'conv', 'bi', 'feat'] noise_path = '../data/processed/white_noise_sd-1.npz'
stages = ['filt', 'env', 'log', 'inv', 'conv', 'feat']
save_path = '../data/inv/full/' save_path = '../data/inv/full/'
# ANALYSIS SETTINGS: # ANALYSIS SETTINGS:
example_scales = np.array([0, 1, 10, 50]) example_scales = np.array([0.1, 1, 10, 30, 100, 300])
scales = np.geomspace(0.01, 100, 100) scales = np.geomspace(0.01, 10000, 100)
scales = np.unique(np.concatenate((scales, example_scales))) scales = np.unique(np.concatenate((scales, example_scales)))
thresh_rel = 3
# SUBSET SETTINGS:
kernels = np.array([ kernels = np.array([
[1, 0.002], [1, 0.002],
[-1, 0.002], [-1, 0.002],
@@ -29,20 +33,29 @@ kernels = None
types = None#np.array([-1]) types = None#np.array([-1])
sigmas = None#np.array([0.001, 0.002, 0.004, 0.008, 0.016, 0.032]) sigmas = None#np.array([0.001, 0.002, 0.004, 0.008, 0.016, 0.032])
# PREPARATION:
noise_data = np.load(noise_path)
pure_noise = noise_data['raw']
# EXECUTION: # EXECUTION:
for data_path, name in zip(data_paths, crop_paths(data_paths)): for data_path, name in zip(data_paths, crop_paths(data_paths)):
print(f'Processing {name}') print(f'Processing {name}')
# Get song recording: # Get song recording (prior to anything):
data, config = load_data(data_path, files='raw') data, config = load_data(data_path, files='raw')
song, rate = data['raw'], config['rate'] song, rate = data['raw'], config['rate']
if thresh_rel is not None:
# Get noise-bound kernel-specific thresholds:
config['feat_thresh'] = noise_data['conv'].std(axis=0) * thresh_rel
# Reduce to kernel subset: # Reduce to kernel subset:
kern_inds = find_kern_specs(config['k_specs'], kernels, types, sigmas) if any(var is not None for var in [kernels, types, sigmas]):
config['kernels'] = config['kernels'][:, kern_inds] kern_inds = find_kern_specs(config['k_specs'], kernels, types, sigmas)
config['k_specs'] = config['k_specs'][kern_inds, :] config['kernels'] = config['kernels'][:, kern_inds]
config['k_props'] = [config['k_props'][i] for i in kern_inds] config['k_specs'] = config['k_specs'][kern_inds, :]
config['feat_thresh'] = config['feat_thresh'][kern_inds] config['k_props'] = [config['k_props'][i] for i in kern_inds]
config['feat_thresh'] = config['feat_thresh'][kern_inds]
# Get song segment to be analyzed: # Get song segment to be analyzed:
time = np.arange(song.shape[0]) / rate time = np.arange(song.shape[0]) / rate
@@ -52,22 +65,19 @@ for data_path, name in zip(data_paths, crop_paths(data_paths)):
# Normalize song component: # Normalize song component:
song /= song[segment].std(axis=0) song /= song[segment].std(axis=0)
# Get normalized noise: # Get normalized noise component:
rng = np.random.default_rng() noise = pure_noise[:song.shape[0]]
noise = rng.normal(size=song.shape[0])
noise /= noise[segment].std() noise /= noise[segment].std()
# Prepare snippet storage: # Prepare snippet storage:
shape_low = (song.shape[0], example_scales.size) shape_low = (song.shape[0], example_scales.size)
shape_high = (song.shape[0], config['k_specs'].shape[0], example_scales.size) shape_high = (song.shape[0], config['k_specs'].shape[0], example_scales.size)
snippets = dict( snippets = dict(
snip_raw=np.zeros(shape_low, dtype=float),
snip_filt=np.zeros(shape_low, dtype=float), snip_filt=np.zeros(shape_low, dtype=float),
snip_env=np.zeros(shape_low, dtype=float), snip_env=np.zeros(shape_low, dtype=float),
snip_log=np.zeros(shape_low, dtype=float), snip_log=np.zeros(shape_low, dtype=float),
snip_inv=np.zeros(shape_low, dtype=float), snip_inv=np.zeros(shape_low, dtype=float),
snip_conv=np.zeros(shape_high, dtype=float), snip_conv=np.zeros(shape_high, dtype=float),
snip_bi=np.zeros(shape_high, dtype=float),
snip_feat=np.zeros(shape_high, dtype=float) snip_feat=np.zeros(shape_high, dtype=float)
) )
@@ -75,7 +85,6 @@ for data_path, name in zip(data_paths, crop_paths(data_paths)):
shape_low = (scales.size,) shape_low = (scales.size,)
shape_high = (scales.size, config['k_specs'].shape[0]) shape_high = (scales.size, config['k_specs'].shape[0])
measures = dict( measures = dict(
measure_raw=np.zeros(shape_low, dtype=float),
measure_filt=np.zeros(shape_low, dtype=float), measure_filt=np.zeros(shape_low, dtype=float),
measure_env=np.zeros(shape_low, dtype=float), measure_env=np.zeros(shape_low, dtype=float),
measure_log=np.zeros(shape_low, dtype=float), measure_log=np.zeros(shape_low, dtype=float),
@@ -96,36 +105,18 @@ for data_path, name in zip(data_paths, crop_paths(data_paths)):
signal=scaled, rate=rate) signal=scaled, rate=rate)
# Store results: # Store results:
for stage in stages: for stage in stages:
key = f'measure_{stage}' mkey, skey = f'measure_{stage}', f'snip_{stage}'
# Log snippet data: # Log snippet data:
if scale in example_scales: if scale in example_scales:
scale_ind = np.nonzero(example_scales == scale)[0][0] scale_ind = np.nonzero(example_scales == scale)[0][0]
snippets[f'snip_{stage}'][:, ..., scale_ind] = signals[stage] snippets[skey][:, ..., scale_ind] = signals[stage]
# Log intensity measure per stage (excluding binary): # Log intensity measure per stage (excluding binary):
if stage in ['raw', 'filt', 'env', 'log', 'inv', 'conv']: if stage in ['raw', 'filt', 'env', 'log', 'inv', 'conv']:
measures[key][i] = signals[stage][segment, ...].std(axis=0) measures[mkey][i] = signals[stage][segment, ...].std(axis=0)
elif stage == 'feat': elif stage == 'feat':
measures[key][i] = signals[stage][segment, :].mean(axis=0) measures[mkey][i] = signals[stage][segment, :].mean(axis=0)
# thresh_y = np.percentile(measures['measure_feat'], 99, axis=0)
# kern_types = np.unique()
# thresh_x = np.zeros(thresh_y.shape, dtype=float)
# for i, thresh in enumerate(thresh_y):
# if thresh < 0.1:
# thresh_x[i] = scales[-1]
# continue
# mask = (measures['measure_feat'][:, i] < thresh)
# thresh_x[i] = scales[np.nonzero(mask)[0][-1]]
# inds = np.argsort(thresh_x)
# print(config['k_specs'][inds, :])
# fig, axes = plt.subplots(1, 2)
# axes[0].plot(snippets['snip_feat'][:, inds, -1])
# axes[1].plot(scales, measures['measure_feat'][:, inds])
# plt.show()
# embed()
# Save analysis results: # Save analysis results:
if save_path is not None: if save_path is not None:

11
python/save_inv_data_log-hp.py
View File

@@ -7,15 +7,19 @@ from IPython import embed
# GENERAL SETTINGS: # GENERAL SETTINGS:
target = ['Omocestus_rufipes', '*'][0] target = ['Omocestus_rufipes', '*'][0]
data_paths = search_files(target, excl='noise', dir='../data/processed/') data_paths = search_files(target, excl='noise', dir='../data/processed/')
noise_path = '../data/processed/white_noise_sd-1.npz'
save_path = '../data/inv/log_hp/' save_path = '../data/inv/log_hp/'
# ANALYSIS SETTINGS: # ANALYSIS SETTINGS:
add_noise = False add_noise = target == '*' or False
save_snippets = target == 'Omocestus_rufipes' save_snippets = target == 'Omocestus_rufipes'
example_scales = np.array([0.1, 1, 10, 30, 100, 300]) example_scales = np.array([0.1, 1, 10, 30, 100, 300])
scales = np.geomspace(0.1, 10000, 500) scales = np.geomspace(0.1, 10000, 500)
scales = np.unique(np.concatenate((scales, example_scales))) scales = np.unique(np.concatenate((scales, example_scales)))
# PREPARATION:
pure_noise = np.load(noise_path)['filt']
# EXECUTION: # EXECUTION:
for data_path, name in zip(data_paths, crop_paths(data_paths)): for data_path, name in zip(data_paths, crop_paths(data_paths)):
print(f'Processing {name}') print(f'Processing {name}')
@@ -36,9 +40,8 @@ for data_path, name in zip(data_paths, crop_paths(data_paths)):
mix = song[:, None] * scales[None, :] mix = song[:, None] * scales[None, :]
if add_noise: if add_noise:
# Add normalized envelopenoise: # Add normalized noise component:
rng = np.random.default_rng() noise = pure_noise[:song.shape[0]]
noise = rng.normal(scale=1, size=song.shape)
noise /= noise[segment].std() noise /= noise[segment].std()
mix += noise[:, None] mix += noise[:, None]

12
python/save_inv_data_thresh-lp.py
View File

@@ -8,13 +8,13 @@ from thunderhopper.model import convolve_kernels
from IPython import embed from IPython import embed
# GENERAL SETTINGS: # GENERAL SETTINGS:
target = ['Omocestus_rufipes', '*'][1] target = ['Omocestus_rufipes', '*'][0]
data_paths = search_files(target, excl='noise', dir='../data/processed/') data_paths = search_files(target, excl='noise', dir='../data/processed/')
noise_path = '../data/processed/white_noise_sd-1.npz' noise_path = '../data/processed/white_noise_sd-1.npz'
save_path = '../data/inv/thresh_lp/' save_path = '../data/inv/thresh_lp/'
# ANALYSIS SETTINGS: # ANALYSIS SETTINGS:
add_noise = True add_noise = False
save_snippets = add_noise and (target == 'Omocestus_rufipes') save_snippets = add_noise and (target == 'Omocestus_rufipes')
plot_results = False plot_results = False
example_scales = np.array([0, 1, 10, 30, 100]) example_scales = np.array([0, 1, 10, 30, 100])
@@ -62,8 +62,8 @@ for data_path, name in zip(data_paths, crop_paths(data_paths)):
thresh_abs = ref_conv[segment, :].std(axis=0, keepdims=True) * thresh_rel[:, None] thresh_abs = ref_conv[segment, :].std(axis=0, keepdims=True) * thresh_rel[:, None]
# Prepare measure storage: # Prepare measure storage:
shape = (scales.size, kern_specs.shape[0], thresh_rel.size) measure_inv = np.zeros((scales.size,), dtype=float)
measure_feat = np.zeros(shape, dtype=float) measure_feat = np.zeros((scales.size, kern_specs.shape[0], thresh_rel.size), dtype=float)
if save_snippets: if save_snippets:
# Prepare snippet storage: # Prepare snippet storage:
snip_inv = np.zeros((song.size, example_scales.size), dtype=float) snip_inv = np.zeros((song.size, example_scales.size), dtype=float)
@@ -81,6 +81,9 @@ for data_path, name in zip(data_paths, crop_paths(data_paths)):
if add_noise: if add_noise:
# Add noise: # Add noise:
scaled_song += noise scaled_song += noise
# Log input intensity measure:
measure_inv[i] = scaled_song[segment].std()
# Process mixture: # Process mixture:
scaled_conv = convolve_kernels(scaled_song, config['kernels'], config['k_specs']) scaled_conv = convolve_kernels(scaled_song, config['kernels'], config['k_specs'])
@@ -130,6 +133,7 @@ for data_path, name in zip(data_paths, crop_paths(data_paths)):
data = dict( data = dict(
scales=scales, scales=scales,
example_scales=example_scales, example_scales=example_scales,
measure_inv=measure_inv,
measure_feat=measure_feat, measure_feat=measure_feat,
thresh_rel=thresh_rel, thresh_rel=thresh_rel,
thresh_abs=thresh_abs, thresh_abs=thresh_abs,

4
python/save_noise_data.py
View File

@@ -7,7 +7,7 @@ from IPython import embed
# General: # General:
save_path = '../data/processed/white_noise' save_path = '../data/processed/white_noise'
stages = ['filt', 'env', 'log', 'inv', 'conv', 'bi', 'feat'] stages = ['raw', 'filt', 'env', 'log', 'inv', 'conv', 'bi', 'feat']
sds = [1] sds = [1]
dur = 60 dur = 60
@@ -23,9 +23,9 @@ types = [1, -1, 2, -2, 3, -3, 4, -4, 5, -5,
6, -6, 7, -7, 8, -8, 9, -9, 10, -10] 6, -6, 7, -7, 8, -8, 9, -9, 10, -10]
config = configuration(env_rate, feat_rate, types=types, sigmas=sigmas) config = configuration(env_rate, feat_rate, types=types, sigmas=sigmas)
config.update({ config.update({
'bp_fcut': None,
'rate_ratio': None, 'rate_ratio': None,
'env_fcut': 250, 'env_fcut': 250,
'db_ref': 1,
'inv_fcut': 5, 'inv_fcut': 5,
'feat_thresh': np.load('../data/kernel_thresholds.npy') * 0.2, 'feat_thresh': np.load('../data/kernel_thresholds.npy') * 0.2,
'feat_fcut': 0.5, 'feat_fcut': 0.5,

65
python/save_ref_measures.py Normal file
View File

@@ -0,0 +1,65 @@
import numpy as np
from thunderhopper.filters import decibel, sosfilter
from thunderhopper.model import convolve_kernels, process_signal
from thunderhopper.modeltools import load_data
from IPython import embed
## SETTINGS:
# General:
mode = ['log_hp', 'thresh_lp', 'full'][2]
noise_path = '../data/processed/white_noise_sd-1.npz'
save_path = '../data/inv/'
pad = np.array([0.1, 0.9])
stages = dict(
log_hp=['filt', 'env', 'log', 'inv'],
thresh_lp=['inv', 'conv', 'feat'],
full=['raw', 'filt', 'env', 'log', 'inv', 'conv', 'feat']
)[mode]
# PROCESSING:
print(f'Fetching references for {mode} invariance...')
# Load pure-noise starter representation:
noise_data, config = load_data(noise_path, stages[0])
starter = noise_data[stages[0]]
# Prepare buffered measurement segment:
pad = (pad * starter.shape[0]).astype(int)
segment = np.arange(starter.shape[0])[pad[0]:pad[1]]
# Normalize starter:
starter /= starter[segment].std()
# Run pipeline:
if mode == 'log_hp':
data = {'filt': starter}
data['env'] = sosfilter(np.abs(data['filt']), config['rate'], config['env_fcut'],
'lp', padtype='even', padlen=config['padlen'])
data['log'] = decibel(data['env'], ref=1)
data['inv'] = sosfilter(data['log'], config['env_rate'], config['inv_fcut'],
'hp', padtype='constant', padlen=config['padlen'])
elif mode == 'thresh_lp':
data = {'inv': starter}
data['conv'] = convolve_kernels(data['inv'], config['kernels'], config['k_specs'])
data['feat'] = sosfilter((data['conv'] > config['feat_thresh']).astype(float),
config['env_rate'], config['feat_fcut'], 'lp',
padtype='fixed', padlen=config['padlen'])
elif mode == 'full':
data = process_signal(config, stages, signal=starter, rate=config['rate'])[0]
# Get measures:
measures = {}
for stage in stages:
if stage == 'feat':
measures[stage] = data[stage][segment, :].mean(axis=0)
else:
measures[stage] = data[stage][segment, ...].std(axis=0)
# Save results:
np.savez(save_path + f'{mode}/ref_measures.npz', **measures)
print('Done.')
embed()

3
python/save_snippet_data.py
View File

@@ -16,7 +16,7 @@ if False:
# Interactivity: # Interactivity:
reload_saved = False reload_saved = False
gui = False gui = True
# Processing: # Processing:
env_rate = 44100.0 env_rate = 44100.0
@@ -29,6 +29,7 @@ config.update({
'channel': 0, 'channel': 0,
'rate_ratio': None, 'rate_ratio': None,
'env_fcut': 250, 'env_fcut': 250,
'db_ref': 1,
'inv_fcut': 5, 'inv_fcut': 5,
'feat_thresh': np.load('../data/kernel_thresholds.npy') * 0.2, 'feat_thresh': np.load('../data/kernel_thresholds.npy') * 0.2,
'feat_fcut': 0.5, 'feat_fcut': 0.5,