n_smooth_samples = np.floor(len(rates_all) / d_idx).astype(int)
sm_rates = np.zeros((n_smooth_samples, nCells_tot))
for i in range(n_smooth_samples):
si = i * d_idx
ei = (i + 1) * d_idx
sm_rates[i] = np.mean(rates_all[si:ei], axis=0)
results = {'session': session, 'h0': [], 'h1': [], 'h2': []}
# if greater than 10 cells, dim reduce to 10 dims using Isomap
fit_dim = 10
dr_method = 'iso'
n_neighbors = 5
dim_red_params = {'n_neighbors': n_neighbors, 'target_dim': fit_dim}
if nCells_tot > 10:
rates = run_dim_red(sm_rates, params=dim_red_params, method=dr_method)
else:
rates = sm_rates
# threshold out outlier points with low neighborhood density
if thrsh:
# a) find number of neighbors of each point within radius of 1st percentile of all
# pairwise dist.
dist = pdist(rates, 'euclidean')
rad = np.percentile(dist, 1)
neigh = neighbors.NearestNeighbors()
neigh.fit(rates)
num_nbrs = np.array(
map(len,
neigh.radius_neighbors(X=rates, radius=rad,
return_distance=False)))
rate_params = {'dt': dt_kernel, 'sigma': sigma}
method = 'iso'
n_neighbors = 5
dim_red_params = {'n_neighbors': n_neighbors, 'target_dim': target_dim}
to_plot = True
session_rates = spike_counts(session,
rate_params,
count_type='rate',
anat_region='ADn')
t0 = time.time()
if condition == 'solo':
counts, tmp_angles = session_rates.get_spike_matrix(state)
sel_counts = counts[:desired_nSamples]
proj = run_dim_red(sel_counts, params=dim_red_params, method=method)
to_save = {
'seed': sd,
state: proj,
'meas_angles': tmp_angles[:desired_nSamples]
}
fname = '%s_%s_kern_%dms_sigma_%dms_binsep_%s_embed_%s_%ddims_%dneighbors_%d.p' % (
session, area, sigma * 1000, dt_kernel * 1000, state, method,
target_dim, n_neighbors, sd)
elif condition == 'joint':
counts1, _ = session_rates.get_spike_matrix(state)
counts2, _ = session_rates.get_spike_matrix(state2)
print('Counts for each ', len(counts1), len(counts2))
nSamples = min(len(counts1), len(counts2), desired_nSamples)
print('nSamples = ', nSamples)
sel1 = counts1[:nSamples]