Added loads of units in nearly all graphs.
Overhauled fig_invariance_full.pdf. Added some legends, somewhere.
This commit is contained in:
j-hartling
294
python/fig_invariance_full-short.py
Normal file
294
python/fig_invariance_full-short.py
Normal file
@@ -0,0 +1,294 @@
|
||||
import plotstyle_plt
|
||||
import string
|
||||
import numpy as np
|
||||
import matplotlib.pyplot as plt
|
||||
from thunderhopper.filetools import search_files
|
||||
from thunderhopper.modeltools import load_data
|
||||
from thunderhopper.filtertools import find_kern_specs
|
||||
from color_functions import load_colors
|
||||
from plot_functions import hide_ticks, ylabel, super_xlabel, letter_subplots,\
|
||||
ylimits, title_subplot
|
||||
from misc_functions import exclude_zero_scale, reduce_kernel_set
|
||||
from IPython import embed
|
||||
|
||||
# GENERAL SETTINGS:
|
||||
target_species = [
|
||||
'Chorthippus_biguttulus',
|
||||
'Chorthippus_mollis',
|
||||
'Chrysochraon_dispar',
|
||||
'Euchorthippus_declivus',
|
||||
'Gomphocerippus_rufus',
|
||||
'Omocestus_rufipes',
|
||||
'Pseudochorthippus_parallelus',
|
||||
][5]
|
||||
modes = [
|
||||
'unnormed',
|
||||
'norm-base',
|
||||
'norm-min',
|
||||
'norm-max',
|
||||
]
|
||||
full_folder = '../data/inv/full/condensed/'
|
||||
short_folder = '../data/inv/short/condensed/'
|
||||
save_path = '../figures/fig_invariance_full_short.pdf'
|
||||
load_kwargs = dict(
|
||||
files=['scales', 'mean_feat', 'sd_feat'],
|
||||
keywords=['thresh'],
|
||||
)
|
||||
|
||||
# ANALYSIS SETTINGS:
|
||||
exclude_zero = True
|
||||
scale_subset_kwargs = dict(
|
||||
combis=[['mean', 'sd'], ['feat']],
|
||||
)
|
||||
kern_subset_kwargs = dict(
|
||||
combis=[['mean', 'sd'], ['feat']],
|
||||
keys=['thresh_abs'],
|
||||
)
|
||||
thresh_rel = np.array([0, 0.5, 1, 1.5, 2, 2.5, 3])
|
||||
percentiles = np.array([
|
||||
[25, 75],
|
||||
# [0, 100],
|
||||
])
|
||||
|
||||
# SUBSET SETTINGS:
|
||||
types = np.array([1, 2, 3])
|
||||
# types = [1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 7, -7, 8, -8, 9, -9, 10, -10]
|
||||
sigmas = np.array([0.001, 0.002, 0.004, 0.008, 0.016, 0.032])
|
||||
# sigmas = [0.001, 0.002, 0.004, 0.008, 0.016, 0.032]
|
||||
kernels = None
|
||||
reduce_kernels = any(var is not None for var in [kernels, types, sigmas])
|
||||
|
||||
# GRAPH SETTINGS:
|
||||
fig_kwargs = dict(
|
||||
figsize=(32/2.54, 32/2.54),
|
||||
)
|
||||
super_grid_kwargs = dict(
|
||||
nrows=1,
|
||||
ncols=2,
|
||||
wspace=0,
|
||||
hspace=0,
|
||||
left=0,
|
||||
right=1,
|
||||
bottom=0,
|
||||
top=1,
|
||||
)
|
||||
subfig_specs = dict(
|
||||
full=(0, 0),
|
||||
short=(0, 1),
|
||||
)
|
||||
col_width = 0.85
|
||||
col_rest = 1 - col_width
|
||||
full_grid_kwargs = dict(
|
||||
nrows=len(modes),
|
||||
ncols=1,
|
||||
wspace=0,
|
||||
hspace=0.1,
|
||||
left=col_rest,
|
||||
right=1,
|
||||
bottom=0.05,
|
||||
top=0.9
|
||||
)
|
||||
short_grid_kwargs = dict(
|
||||
nrows=len(modes),
|
||||
ncols=1,
|
||||
wspace=full_grid_kwargs['wspace'],
|
||||
hspace=full_grid_kwargs['hspace'],
|
||||
left=col_rest / 2,
|
||||
right=1 - col_rest / 2,
|
||||
bottom=full_grid_kwargs['bottom'],
|
||||
top=full_grid_kwargs['top']
|
||||
)
|
||||
|
||||
# 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')
|
||||
feat_colors = load_colors('../data/feat_colors_all.npz')
|
||||
lw = dict(
|
||||
feat=3,
|
||||
plateau=1.5
|
||||
)
|
||||
line_kwargs = dict(
|
||||
lw=lw['feat']
|
||||
)
|
||||
fill_kwargs = dict(
|
||||
alpha=0.15
|
||||
)
|
||||
xlabels = dict(
|
||||
super='scale $\\alpha$',
|
||||
)
|
||||
ylabels = {
|
||||
'unnormed': '$\\mu_{f_i}$',
|
||||
'norm-base': '$\\mu_{f_i}\\,/\\,\\mu_{f_i}\\,[\\,\\eta\\,]$',
|
||||
'norm-min': '$\\mu_{f_i}\\,/\\,\\min\\,[\\,\\mu_{f_i}\\,]$',
|
||||
'norm-max': '$\\mu_{f_i}\\,/\\,\\max\\,[\\,\\mu_{f_i}\\,]$'
|
||||
}
|
||||
xlab_kwargs = dict(
|
||||
y=0,
|
||||
fontsize=fs['lab_norm'],
|
||||
ha='center',
|
||||
va='bottom',
|
||||
)
|
||||
ylab_kwargs = dict(
|
||||
x=0,
|
||||
fontsize=fs['lab_tex'],
|
||||
ha='center',
|
||||
va='top',
|
||||
)
|
||||
ylims = {
|
||||
'unnormed': [0, 1],
|
||||
'norm-base': [0, None],
|
||||
'norm-min': [0, None],
|
||||
'norm-max': [0, 1]
|
||||
}
|
||||
yloc = {
|
||||
'unnormed': 0.5,
|
||||
'norm-base': 0.5,
|
||||
'norm-min': 0.5,
|
||||
'norm-max': 0.5
|
||||
}
|
||||
title_kwargs = dict(
|
||||
x=0.5,
|
||||
y=1,
|
||||
ha='center',
|
||||
va='bottom',
|
||||
fontsize=fs['tit_norm'],
|
||||
)
|
||||
titles = dict(
|
||||
full='Including $x_\\text{dB}$',
|
||||
short='Excluding $x_\\text{dB}$',
|
||||
)
|
||||
letter_kwargs = dict(
|
||||
xref=0.01,
|
||||
y=1,
|
||||
ha='left',
|
||||
va='center',
|
||||
fontsize=fs['letter'],
|
||||
)
|
||||
letters = dict(
|
||||
full=string.ascii_lowercase[0::2],
|
||||
short=string.ascii_lowercase[1::2],
|
||||
)
|
||||
plateau_settings = dict(
|
||||
low=0.05,
|
||||
high=0.95,
|
||||
first=True,
|
||||
last=True,
|
||||
condense=None,
|
||||
)
|
||||
plateau_line_kwargs = dict(
|
||||
lw=lw['plateau'],
|
||||
ls='--',
|
||||
zorder=1,
|
||||
)
|
||||
plateau_dot_kwargs = dict(
|
||||
marker='o',
|
||||
markersize=8,
|
||||
markeredgewidth=1,
|
||||
clip_on=False,
|
||||
)
|
||||
|
||||
# EXECUTION:
|
||||
|
||||
# Prepare overall graph:
|
||||
fig = plt.figure(**fig_kwargs)
|
||||
super_grid = fig.add_gridspec(**super_grid_kwargs)
|
||||
|
||||
# Prepare full analysis axes:
|
||||
full_subfig = fig.add_subfigure(super_grid[*subfig_specs['full']])
|
||||
full_grid = full_subfig.add_gridspec(**full_grid_kwargs)
|
||||
full_axes = np.zeros((len(modes),), dtype=object)
|
||||
for i, mode in enumerate(modes):
|
||||
full_axes[i] = full_subfig.add_subplot(full_grid[i, 0])
|
||||
# full_axes[i].yaxis.set_major_locator(plt.MultipleLocator(yloc[mode]))
|
||||
full_axes[i].set_xscale('symlog', linthresh=0.01, linscale=0.5)
|
||||
ylabel(full_axes[i], ylabels[mode], transform=full_subfig.transSubfigure, **ylab_kwargs)
|
||||
if i == 0:
|
||||
title_subplot(full_axes[i], titles['full'], **title_kwargs)
|
||||
if i < full_grid_kwargs['nrows'] - 1:
|
||||
hide_ticks(full_axes[i], 'bottom')
|
||||
letter_subplots(full_axes, letters['full'], ref=full_subfig, **letter_kwargs)
|
||||
|
||||
# Prepare short analysis axes:
|
||||
short_subfig = fig.add_subfigure(super_grid[*subfig_specs['short']])
|
||||
short_grid = short_subfig.add_gridspec(**short_grid_kwargs)
|
||||
short_axes = np.zeros((len(modes),), dtype=object)
|
||||
for i, mode in enumerate(modes):
|
||||
short_axes[i] = short_subfig.add_subplot(short_grid[i, 0])
|
||||
# short_axes[i].yaxis.set_major_locator(plt.MultipleLocator(yloc[mode]))
|
||||
short_axes[i].set_xscale('symlog', linthresh=0.01, linscale=0.5)
|
||||
hide_ticks(short_axes[i], 'left')
|
||||
if i == 0:
|
||||
title_subplot(short_axes[i], titles['short'], **title_kwargs)
|
||||
if i < short_grid_kwargs['nrows'] - 1:
|
||||
hide_ticks(short_axes[i], 'bottom')
|
||||
letter_subplots(short_axes, letters['short'], ref=short_subfig, **letter_kwargs)
|
||||
|
||||
super_xlabel(xlabels['super'], fig, full_axes[-1], short_axes[-1],
|
||||
left_fig=full_subfig, right_fig=short_subfig, **xlab_kwargs)
|
||||
|
||||
# Run through normalization modes:
|
||||
for mode, full_ax, short_ax in zip(modes, full_axes, short_axes):
|
||||
|
||||
# Load invariance data:
|
||||
full_path = search_files(target_species, incl=mode, dir=full_folder)[0]
|
||||
short_path = search_files(target_species, incl=mode, dir=short_folder)[0]
|
||||
full_data, config = load_data(full_path, **load_kwargs)
|
||||
short_data, _ = load_data(short_path, **load_kwargs)
|
||||
|
||||
# Reduce datasets:
|
||||
if reduce_kernels:
|
||||
kern_inds = find_kern_specs(config['k_specs'], kernels, types, sigmas)
|
||||
full_data = reduce_kernel_set(full_data, kern_inds, **kern_subset_kwargs)
|
||||
short_data = reduce_kernel_set(short_data, kern_inds, **kern_subset_kwargs)
|
||||
config['k_specs'] = config['k_specs'][kern_inds, :]
|
||||
config['kernels'] = config['kernels'][:, kern_inds]
|
||||
if exclude_zero:
|
||||
full_data = exclude_zero_scale(full_data, **scale_subset_kwargs)
|
||||
short_data = exclude_zero_scale(short_data, **scale_subset_kwargs)
|
||||
|
||||
# Average over recordings:
|
||||
full_measure = full_data['mean_feat'].mean(axis=-1)
|
||||
short_measure = short_data['mean_feat'].mean(axis=-1)
|
||||
|
||||
# Condense over kernels:
|
||||
full_median = np.nanmedian(full_measure, axis=1)
|
||||
full_spread = np.nanpercentile(full_measure, percentiles, axis=1)
|
||||
short_median = np.nanmedian(short_measure, axis=1)
|
||||
short_spread = np.nanpercentile(short_measure, percentiles, axis=1)
|
||||
|
||||
# Determine shared ylims:
|
||||
if None in ylims[mode]:
|
||||
min_val, max_val = ylims[mode]
|
||||
full_limits = ylimits(full_median, minval=min_val, maxval=max_val)
|
||||
short_limits = ylimits(short_median, minval=min_val, maxval=max_val)
|
||||
ylims[mode] = [min(full_limits[0], short_limits[0]),
|
||||
max(full_limits[1], short_limits[1])]
|
||||
if np.inf in ylims[mode]:
|
||||
embed()
|
||||
|
||||
# Plot full analysis results:
|
||||
for i, thresh in enumerate(thresh_rel):
|
||||
full_ax.plot(full_data['scales'], full_median[:, i], lw=lw['feat'])
|
||||
for spread in full_spread[:, :, :, i]:
|
||||
full_ax.fill_between(full_data['scales'], *spread, **fill_kwargs)
|
||||
full_ax.set_xlim(full_data['scales'][0], full_data['scales'][-1])
|
||||
full_ax.set_ylim(ylims[mode])
|
||||
|
||||
# Plot short analysis results:
|
||||
for i, thresh in enumerate(thresh_rel):
|
||||
short_ax.plot(short_data['scales'], short_median[:, i], lw=lw['feat'])
|
||||
for spread in short_spread[:, :, :, i]:
|
||||
short_ax.fill_between(short_data['scales'], *spread, **fill_kwargs)
|
||||
short_ax.set_xlim(short_data['scales'][0], short_data['scales'][-1])
|
||||
short_ax.set_ylim(ylims[mode])
|
||||
|
||||
plt.show()
|
||||
|
||||
|
||||
Reference in New Issue
Block a user