nonlinearbaseline2025/noisesplit.py

289 lines
11 KiB
Python

import numpy as np
import matplotlib.pyplot as plt
from pathlib import Path
from spectral import whitenoise
from plotstyle import plot_style, noise_files, plot_chi2
example_cell = ['2017-07-18-ai-invivo-1', 1]
model_cell = example_cell[0]
base_path = Path('data')
data_path = base_path / 'cells'
sims_path = base_path / 'simulations'
def plot_overn(ax, s, files, nmax=1e6):
ns = []
stats = []
for fname in files:
data = np.load(fname)
fcutoff = data['fcutoff']
n = data['nsegs']
if nmax is not None and n > nmax:
continue
alpha = data['contrast']
freqs = data['freqs']
pss = data['pss']
prss = data['prss']
chi2 = np.abs(prss)/0.5/(pss.reshape(1, -1)*pss.reshape(-1, 1))
chi2 *= 1e-4 # Hz/%^2
ns.append(n)
i0 = np.argmin(freqs < 0)
i1 = np.argmax(freqs > fcutoff)
if i1 == 0:
i1 = len(freqs)
chi2 = chi2[i0:i1, i0:i1]
stats.append(np.quantile(chi2, [0, 0.001, 0.05, 0.25, 0.5,
0.75, 0.95, 0.998, 1.0]))
ns = np.array(ns)
stats = np.array(stats)
indx = np.argsort(ns)
ns = ns[indx]
stats = stats[indx]
ax.set_visible(True)
ax.plot(ns, stats[:, 7], zorder=50, label='99.8\\%', **s.lsMax)
ax.fill_between(ns, stats[:, 2], stats[:, 6], fc='0.85', zorder=40, label='5--95\\%')
ax.fill_between(ns, stats[:, 3], stats[:, 5], fc='0.5', zorder=45, label='25-75\\%')
ax.plot(ns, stats[:, 4], zorder=50, label='median', **s.lsMedian)
#ax.plot(ns, stats[:, 8], '0.0')
ax.set_xlim(1e2, nmax)
ax.set_xscale('log')
ax.set_yscale('log')
ax.set_yticks_log(numticks=5)
if nmax > 1e6:
ax.set_ylim(3e-1, 5e3)
ax.set_minor_yticks_log(numticks=5)
ax.set_xticks_log(numticks=4)
ax.set_minor_xticks_log(numticks=8)
else:
ax.set_ylim(1e-1, 2e3)
#ax.set_minor_yticks_log(numticks=5)
ax.set_minor_yticks_off()
ax.set_xticks_log(numticks=6)
#ax.set_minor_xticks_log(numticks=6)
ax.set_xlabel('segments')
ax.set_ylabel(r'$|\chi_2|$', r'Hz/\%$^2$')
def plot_chi2_data(ax, s, cell_name, run):
vmax = 15
data_file = data_path / f'{cell_name}-baseline.npz'
data = np.load(data_file)
eodf = float(data['eodf'])
ratebase = float(data['ratebase/Hz'])
cvbase = float(data['cvbase'])
data_file = data_path / f'{cell_name}-spectral-100-s{run:02d}.npz'
data = np.load(data_file)
nsegs = data['nsegs']
fcutoff = data['fcutoff']
nfft = data['nfft']
deltat = data['deltat']
alpha = data['contrast']
freqs = data['freqs']
pss = data['pss']
prss = data['prss']
chi2 = np.abs(prss)*0.5/(pss.reshape(1, -1)*pss.reshape(-1, 1))
print(f'Measured cell {"-".join(data_file.name.split("-")[:-3])} at {100*alpha:4.1f}% contrast:')
print(f' r={ratebase:3.0f}Hz, CV={cvbase:4.2f}, dt={1000*deltat:4.2f}ms, nfft={nfft}, win={1000*deltat*nfft:6.1f}ms, nsegs={nsegs}')
print()
ax.text(1, 1.1, f'$N={nsegs}$',
ha='right', transform=ax.transAxes)
plot_chi2(ax, s, freqs, chi2, fcutoff, ratebase, vmax)
return alpha, ratebase, eodf
def plot_chi2_contrast(ax1, ax2, s, files, nums, nsmall, nlarge, rate):
vmax = {0.05: {nsmall: 15, nlarge: 1.2},
0.01: {nsmall: 400, nlarge: 4}}
for ax, n in zip([ax1, ax2], [nsmall, nlarge]):
i = nums.index(n)
data = np.load(files[i])
nsegs = data['nsegs']
fcutoff = float(data['fcutoff'])
alpha = float(data['contrast'])
freqs = data['freqs']
pss = data['pss']
prss = data['prss']
if nsegs < 10000:
ax.text(1, 1.1, f'$N={nsegs}$',
ha='right', transform=ax.transAxes)
else:
ax.text(1, 1.1, f'$N=10^{{{np.log10(nsegs):.0f}}}$',
ha='right', transform=ax.transAxes)
chi2 = np.abs(prss)*0.5/(pss.reshape(1, -1)*pss.reshape(-1, 1))
cax = plot_chi2(ax, s, freqs, chi2, fcutoff, rate, vmax[alpha][n])
cax.set_ylabel('')
print(f'Modeled cell {"-".join(files[i].name.split("-")[:-4])} at {100*alpha:4.1f}% contrast: noise_frac={1:3.1f}, nsegs={n}')
print()
def plot_chi2_split(ax1, ax2, s, files, nums, nsmall, nlarge, rate):
vmax = {nsmall: 4, nlarge: 1.2}
for ax, n in zip([ax1, ax2], [nsmall, nlarge]):
i = nums.index(n)
data = np.load(files[i])
nsegs = data['nsegs']
fcutoff = float(data['fcutoff'])
alpha = float(data['contrast'])
noise_frac = data['noise_frac']
freqs = data['freqs']
pss = data['pss']
prss = data['prss']
chi2 = np.abs(prss)*0.5/(pss.reshape(1, -1)*pss.reshape(-1, 1))
if nsegs < 10000:
ax.text(1, 1.1, f'$N={nsegs}$',
ha='right', transform=ax.transAxes)
else:
ax.text(1, 1.1, f'$N=10^{{{np.log10(nsegs):.0f}}}$',
ha='right', transform=ax.transAxes)
cax = plot_chi2(ax, s, freqs, chi2, fcutoff, rate, vmax[n])
cax.set_ylabel('')
print(f'Modeled cell {"-".join(files[i].name.split("-")[:-3])} at {100*alpha:4.1f}% contrast: noise_frac={noise_frac:3.1f}, nsegs={n}')
print()
return alpha, noise_frac
def plot_ram(ax, contrast, eodf, wtime, wnoise):
tmax = 50
am = 1 + contrast*wnoise
eod = np.sin(2*np.pi*eodf*wtime)*am
ax.show_spines('l')
ax.plot(1e3*wtime, eod, clip_on=False, **s.lsEOD)
ax.plot(1e3*wtime, +am, clip_on=False, **s.lsAM)
ax.plot(1e3*wtime, -am, clip_on=False, **s.lsAM)
ax.set_xlim(0, tmax)
ax.set_ylim(-1.3, 1.3)
ax.set_yticks_delta(1)
ax.set_ylabel('EOD')
ax.text(1, 1, f'RAM ($c={100*contrast:.0f}$\\,\\%)', ha='right',
transform=ax.transAxes, color=s.lsAM['color'])
def plot_noise_split(ax, contrast, noise_contrast, noise_frac,
wtime, wnoise):
axr, axs, axn = ax.subplots(3, 1, hspace=0.2)
cmax = 26
cdelta = 20
tmax = 50
axr.show_spines('l')
axr.axhline(0, **s.lsGrid)
axr.plot(1e3*wtime, 100*contrast*wnoise, clip_on=False, **s.lsAM)
axr.set_xlim(0, tmax)
axr.set_ylim(-cmax, cmax)
axr.set_yticks_delta(cdelta)
axr.set_ylabel('\\%')
axr.text(1, 1, f'RAM ($c={100*contrast:.0f}$\\,\\%)', ha='right',
transform=axr.transAxes, color=s.lsAM['color'])
axs.show_spines('l')
axs.axhline(0, **s.lsGrid)
axs.plot(1e3*wtime, 100*noise_contrast*wnoise, clip_on=False, **s.lsAMsplit)
axs.set_xlim(0, tmax)
axs.set_ylim(-cmax, cmax)
axs.set_yticks_delta(cdelta)
axs.set_ylabel('\\%')
if noise_contrast > 0:
axs.text(1, 1, f'$s_{{\\xi}}(t)$ ($c={100*noise_contrast:.0f}$\\,\\%)',
ha='right', transform=axs.transAxes,
color=s.lsAMsplit['color'])
ntime = np.linspace(0, 1e-3*tmax, 800)
rng = np.random.default_rng(45432)
nnoise = rng.normal(size=len(ntime))
axn.show_spines('l')
axn.axhline(0, **s.lsGrid)
axn.plot(1e3*ntime, noise_frac*nnoise, clip_on=False, **s.lsNoise)
axn.set_ylim(-2, 2)
axn.set_xlim(0, tmax)
axn.set_yticks_delta(5)
axn.set_yticks_blank()
#axn.set_xticks_delta(25)
#axn.set_xlabel('Time', 'ms')
y = 0.8 if noise_frac < 1 else 1.2
axn.text(1, y, f'Intrinsic noise (${100*noise_frac:.0f}$\\,\\%)',
ha='right', transform=axn.transAxes, color=s.lsNoise['color'])
if noise_frac < 1:
axn.xscalebar(1, -0.1, 10, 'ms', ha='right')
return axr
if __name__ == '__main__':
nsmall = 100
nlarge = 1000000
contrast = 0.01
wdt = 0.0001
wnoise = whitenoise(0, 300, wdt, 0.05, rng=np.random.default_rng(51234))
wtime = np.arange(len(wnoise))*wdt
s = plot_style()
fig, axs = plt.subplots(4, 4, cmsize=(s.plot_width, 0.85*s.plot_width),
width_ratios=[1, 0, 1, 1, 0.2, 0.85])
fig.subplots_adjust(leftm=8, rightm=1.5, topm=3.5, bottomm=4,
wspace=0.25, hspace=0.6)
axs[0, 2].set_visible(False)
axs[0, 3].set_visible(False)
xt = -2.25
yt = 1.25
# data:
axss = axs[0]
axss[1].text(xt, yt, 'P-unit data', fontsize='large',
transform=axss[1].transAxes, color=s.punit_color1)
data_contrast, ratebase, eodf = plot_chi2_data(axss[1], s, example_cell[0],
example_cell[1])
plot_ram(axss[0], data_contrast, eodf, wtime, wnoise)
axss[1].text(xt + 0.9, yt, f'$r={ratebase:.0f}$\\,Hz',
transform=axss[1].transAxes, fontsize='large')
# model 5%:
axss = axs[1]
files, nums = noise_files(sims_path, model_cell, data_contrast)
axss[1].text(xt, yt, 'P-unit model', fontsize='large',
transform=axs[1, 1].transAxes, color=s.model_color1)
plot_chi2_contrast(axss[1], axss[2], s, files, nums, nsmall, nlarge,
ratebase)
axr1 = plot_noise_split(axss[0], data_contrast, 0, 1, wtime, wnoise)
plot_overn(axss[3], s, files, nmax=1e6)
axss[3].legend(loc='lower center', bbox_to_anchor=(0.5, 1.2),
markerfirst=False, title='$|\\chi_2|$ percentiles')
# model 1%:
axss = axs[2]
files, nums = noise_files(sims_path, model_cell, contrast)
plot_chi2_contrast(axss[1], axss[2], s, files, nums, nsmall, nlarge,
ratebase)
axr2 = plot_noise_split(axss[0], contrast, 0, 1, wtime, wnoise)
plot_overn(axss[3], s, files, nmax=1e6)
# model noise split:
axss = axs[3]
files, nums = noise_files(sims_path, model_cell)
axss[1].text(xt, yt, 'P-unit model', fontsize='large',
transform=axss[1].transAxes, color=s.model_color1)
axss[1].text(xt + 0.9, yt, f'(noise split)', fontsize='large',
transform=axss[1].transAxes)
noise_contrast, noise_frac = plot_chi2_split(axss[1], axss[2], s,
files, nums, nsmall, nlarge,
ratebase)
axr3 = plot_noise_split(axss[0], 0, noise_contrast, noise_frac,
wtime, wnoise)
plot_overn(axss[3], s, files, nmax=1e6)
fig.common_xticks(axs[:, 1])
fig.common_xticks(axs[:, 2])
fig.common_xticks(axs[:, 3])
fig.common_yticks(axs[1, 1:3])
fig.common_yticks(axs[2, 1:3])
fig.common_yticks(axs[3, 1:3])
fig.tag([axs[0, :2],
[axr1] + axs[1, 1:].tolist(),
[axr2] + axs[2, 1:].tolist(),
[axr3] + axs[3, 1:].tolist()],
xoffs=[-4.5, 1, 1, -4.5], yoffs=2)
fig.savefig()