Finished (:D) fig_invariance_thresh_lp_single.pdf.
Added/modified few plot functions. Cleaned up simulation/plotting scripts regarding Thresh-LP.
This commit is contained in:
j-hartling
@@ -1,86 +1,151 @@
|
||||
import glob
|
||||
import numpy as np
|
||||
import matplotlib.pyplot as plt
|
||||
from thunderhopper.modeltools import load_data, save_data
|
||||
from thunderhopper.filetools import crop_paths
|
||||
from thunderhopper.filetools import search_files, crop_paths
|
||||
from thunderhopper.filters import sosfilter
|
||||
from thunderhopper.filtertools import find_kern_specs
|
||||
from thunderhopper.model import convolve_kernels
|
||||
from IPython import embed
|
||||
|
||||
# GENERAL SETTINGS:
|
||||
target = 'Omocestus_rufipes'
|
||||
data_paths = glob.glob(f'../data/processed/{target}*.npz')
|
||||
target = ['Omocestus_rufipes', '*'][0]
|
||||
data_paths = search_files(target, dir='../data/processed/')
|
||||
noise_path = '../data/processed/white_noise_sd-1.npz'
|
||||
save_path = '../data/inv/thresh_lp/'
|
||||
|
||||
# ANALYSIS SETTINGS:
|
||||
add_noise = False
|
||||
thresh_percent = 90
|
||||
example_scales = np.array([0, 1, 10, 50])
|
||||
scales = np.geomspace(0.01, 100, 100)
|
||||
add_noise = True
|
||||
save_snippets = add_noise and True
|
||||
plot_results = False
|
||||
example_scales = np.array([0, 1, 10, 30, 100])
|
||||
scales = np.geomspace(0.01, 10000, 100)
|
||||
scales = np.unique(np.concatenate((scales, example_scales)))
|
||||
thresh_rel = np.array([0.5, 1, 3])
|
||||
kern_specs = np.array([
|
||||
[1, 0.008],
|
||||
[2, 0.004],
|
||||
[3, 0.002],
|
||||
])
|
||||
|
||||
# PREPARATION:
|
||||
pure_noise = np.load(noise_path)['inv']
|
||||
|
||||
# EXECUTION:
|
||||
for data_path, name in zip(data_paths, crop_paths(data_paths)):
|
||||
print(f'Processing {name}')
|
||||
save_name = save_path + name
|
||||
|
||||
# Get pure-song kernel responses:
|
||||
data, config = load_data(data_path, files='conv')
|
||||
song, rate = data['conv'], data['conv_rate']
|
||||
# Get adapted envelope (prior to convolution):
|
||||
data, config = load_data(data_path, files='inv')
|
||||
song, rate = data['inv'], data['inv_rate']
|
||||
|
||||
# Get song segment to be analyzed:
|
||||
time = np.arange(song.shape[0]) / rate
|
||||
start, end = data['songs_0'].ravel()
|
||||
segment = (time >= start) & (time <= end)
|
||||
|
||||
|
||||
# Reduce to kernel subset:
|
||||
kern_inds = find_kern_specs(config['k_specs'], kerns=kern_specs)
|
||||
config['kernels'] = config['kernels'][:, kern_inds]
|
||||
config['k_specs'] = config['k_specs'][kern_inds, :]
|
||||
config['k_props'] = [config['k_props'][i] for i in kern_inds]
|
||||
|
||||
# Normalize song component:
|
||||
song /= song[segment, :].std(axis=0)
|
||||
song /= song[segment].std()
|
||||
|
||||
if add_noise:
|
||||
# Get normalized noise:
|
||||
rng = np.random.default_rng()
|
||||
noise = rng.normal(size=(song.shape[0], 1))
|
||||
noise /= noise[segment].std()
|
||||
# Get normalized noise component:
|
||||
noise = pure_noise[:song.shape[0]]
|
||||
noise /= noise[segment].std()
|
||||
|
||||
# Prepare noise-bound threshold:
|
||||
threshold = np.percentile(noise, thresh_percent, axis=0)
|
||||
else:
|
||||
# Reuse threshold from previous noise run:
|
||||
threshold = np.load(save_name + '_noise.npz')['thresh']
|
||||
# Define kernel-specific threshold values based on pure-noise response SD:
|
||||
ref_conv = convolve_kernels(noise, config['kernels'], config['k_specs'])
|
||||
thresh_abs = ref_conv[segment, :].std(axis=0, keepdims=True) * thresh_rel[:, None]
|
||||
|
||||
# Prepare measure storage:
|
||||
shape = (scales.size, song.shape[1])
|
||||
# measure_conv = np.zeros(shape, dtype=float)
|
||||
shape = (scales.size, kern_specs.shape[0], thresh_rel.size)
|
||||
measure_feat = np.zeros(shape, dtype=float)
|
||||
if save_snippets:
|
||||
# Prepare snippet storage:
|
||||
snip_inv = np.zeros((song.size, example_scales.size), dtype=float)
|
||||
shape = (song.size, kern_specs.shape[0], example_scales.size, thresh_rel.size)
|
||||
snip_conv = np.zeros(shape[:-1], dtype=float)
|
||||
snip_bi = np.zeros(shape, dtype=float)
|
||||
snip_feat = np.zeros(shape, dtype=float)
|
||||
|
||||
# Execute piecewise:
|
||||
for i, scale in enumerate(scales):
|
||||
print('Simulating scale', scale)
|
||||
|
||||
# Rescale song component:
|
||||
scaled_conv = song * scale
|
||||
scaled_song = song * scale
|
||||
if add_noise:
|
||||
# Add noise:
|
||||
scaled_conv += noise
|
||||
scaled_song += noise
|
||||
|
||||
# Process mixture:
|
||||
scaled_bi = (scaled_conv > threshold).astype(float)
|
||||
scaled_feat = sosfilter(scaled_bi, rate, config['feat_fcut'], 'lp',
|
||||
padtype='fixed', padlen=config['padlen'])
|
||||
scaled_conv = convolve_kernels(scaled_song, config['kernels'], config['k_specs'])
|
||||
|
||||
# Log threshold-independent snippet data:
|
||||
if save_snippets and scale in example_scales:
|
||||
save_ind = np.nonzero(example_scales == scale)[0][0]
|
||||
snip_inv[:, save_ind] = scaled_song
|
||||
snip_conv[:, :, save_ind] = scaled_conv
|
||||
|
||||
# Get intensity measure per stage:
|
||||
# measure_conv[i] = scaled_conv[segment, :].std(axis=0)
|
||||
measure_feat[i] = scaled_feat[segment, :].mean(axis=0)
|
||||
# Execute piecewise again:
|
||||
for j, thresholds in enumerate(thresh_abs):
|
||||
|
||||
# Process mixture further:
|
||||
scaled_bi = (scaled_conv > thresholds).astype(float)
|
||||
scaled_feat = sosfilter(scaled_bi, rate, config['feat_fcut'], 'lp',
|
||||
padtype='fixed', padlen=config['padlen'])
|
||||
|
||||
# Log threshold-dependent snippet data:
|
||||
if save_snippets and scale in example_scales:
|
||||
snip_bi[:, :, save_ind, j] = scaled_bi
|
||||
snip_feat[:, :, save_ind, j] = scaled_feat
|
||||
|
||||
# Log intensity measure:
|
||||
measure_feat[i, :, j] = scaled_feat[segment, :].mean(axis=0)
|
||||
|
||||
# Overview plot:
|
||||
if plot_results:
|
||||
fig, axes = plt.subplots(thresh_rel.size, kern_specs.shape[0],
|
||||
figsize=(16, 9), layout='constrained',
|
||||
sharex=True, sharey=True, squeeze=True)
|
||||
axes[0, 0].set_xscale('symlog', linthresh=scales[scales>0].min(),
|
||||
linscale=0.25)
|
||||
axes[0, 0].set_ylim(0, 1)
|
||||
for i, thresh in enumerate(thresh_rel):
|
||||
for j, kernel in enumerate(kern_specs):
|
||||
ax = axes[i, j]
|
||||
ax.plot(scales, measure_feat[:, j, i], 'k')
|
||||
if i == 0:
|
||||
ax.set_title(f'Kernel {kernel}')
|
||||
if j == 0:
|
||||
ax.set_ylabel(f'{thresh} * SD')
|
||||
plt.show()
|
||||
|
||||
# Save analysis results:
|
||||
if save_path is not None:
|
||||
data = dict(
|
||||
scales=scales,
|
||||
# measure_conv=measure_conv,
|
||||
example_scales=example_scales,
|
||||
measure_feat=measure_feat,
|
||||
thresh=threshold,
|
||||
thresh_perc=thresh_percent,
|
||||
thresh_rel=thresh_rel,
|
||||
thresh_abs=thresh_abs,
|
||||
)
|
||||
if save_snippets:
|
||||
data.update(dict(
|
||||
snip_inv=snip_inv,
|
||||
snip_conv=snip_conv,
|
||||
snip_bi=snip_bi,
|
||||
snip_feat=snip_feat,
|
||||
))
|
||||
if add_noise:
|
||||
save_name += '_noise'
|
||||
else:
|
||||
save_name += '_pure'
|
||||
save_data(save_name, data, config, overwrite=True)
|
||||
|
||||
print('Done.')
|
||||
embed()
|
||||
|
||||
Reference in New Issue
Block a user