Source code for tvb.contrib.scripts.utils.computations_utils

# -*- coding: utf-8 -*-
#  TheVirtualBrain-Contributors Package. This package holds simulator extensions.
#  See also
# (c) 2012-2022, Baycrest Centre for Geriatric Care ("Baycrest") and others
# This program is free software: you can redistribute it and/or modify it under the
# terms of the GNU General Public License as published by the Free Software Foundation,
# either version 3 of the License, or (at your option) any later version.
# This program is distributed in the hope that it will be useful, but WITHOUT ANY
# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
# PARTICULAR PURPOSE.  See the GNU General Public License for more details.
# You should have received a copy of the GNU General Public License along with this
# program.  If not, see <>.
# When using The Virtual Brain for scientific publications, please cite it as follows:
#   Paula Sanz Leon, Stuart A. Knock, M. Marmaduke Woodman, Lia Domide,
#   Jochen Mersmann, Anthony R. McIntosh, Viktor Jirsa (2013)
#       The Virtual Brain: a simulator of primate brain network dynamics.
#   Frontiers in Neuroinformatics (7:10. doi: 10.3389/fninf.2013.00010)

Some math tools

.. moduleauthor:: Dionysios Perdikis <Denis@tvb.invalid>

from itertools import product
import numpy as np
from tvb.contrib.scripts.utils.data_structures_utils import is_integer
from sklearn.cluster import AgglomerativeClustering
from tvb.basic.logger.builder import get_logger
from tvb.simulator.plot.config import FiguresConfig

logger = get_logger(__name__)

[docs]def weighted_vector_sum(weights, vectors, normalize=True): if isinstance(vectors, np.ndarray): vectors = list(vectors.T) if normalize: weights /= np.sum(weights) vector_sum = weights[0] * vectors[0] for iv in range(1, len(weights)): vector_sum += weights[iv] * vectors[iv] return np.array(vector_sum)
[docs]def normalize_weights(weights, percentile, remove_diagonal=True, ceil=1.0): # Create the normalized connectivity weights: if len(weights) > 0: normalized_w = np.array(weights) if remove_diagonal: # Remove diagonal elements n_regions = normalized_w.shape[0] normalized_w *= (1.0 - np.eye(n_regions)) # Normalize with the 95th percentile normalized_w = np.array(normalized_w / np.percentile(normalized_w, percentile)) if ceil: if ceil is True: ceil = 1.0 normalized_w[normalized_w > ceil] = ceil return normalized_w else: return np.array([])
[docs]def compute_in_degree(weights): return np.expand_dims(np.sum(weights, axis=1), 1).T
[docs]def compute_gain_matrix(locations1, locations2, normalize=100.0, ceil=False): n1 = locations1.shape[0] n2 = locations2.shape[0] projection = np.zeros((n1, n2)) dist = np.zeros((n1, n2)) for i1, i2 in product(range(n1), range(n2)): dist[i1, i2] = np.abs(np.sum((locations1[i1, :] - locations2[i2, :]) ** 2)) projection[i1, i2] = 1 / dist[i1, i2] if normalize: projection /= np.percentile(projection, normalize) if ceil: if ceil is True: ceil = 1.0 projection[projection > ceil] = ceil return projection
[docs]def get_greater_values_array_inds(values, n_vals=1): return np.argsort(values)[::-1][:n_vals]
[docs]def select_greater_values_array_inds(values, threshold=None, percentile=None, nvals=None, verbose=False): if threshold is None and percentile is not None: threshold = np.percentile(values, percentile) if threshold is not None: return np.where(values > threshold)[0] else: if is_integer(nvals): return get_greater_values_array_inds(values, nvals) if verbose: logger.warning("Switching to curve elbow point method since threshold=" + str(threshold)) elbow_point = curve_elbow_point(values) return get_greater_values_array_inds(values, elbow_point)
[docs]def select_greater_values_2Darray_inds(values, threshold=None, percentile=None, nvals=None, verbose=False): return np.unravel_index( select_greater_values_array_inds(values.flatten(), threshold, percentile, nvals, verbose), values.shape)
[docs]def select_by_hierarchical_group_metric_clustering(distance, disconnectivity=np.array([]), metric=None, n_groups=10, members_per_group=1): if disconnectivity.shape == distance.shape: distance += disconnectivity * distance.max() n_groups = np.minimum(np.maximum(n_groups, 3), n_groups // members_per_group) clustering = AgglomerativeClustering(n_groups, affinity="precomputed", linkage="average") clusters_labels = clustering.fit_predict(distance) selection = [] for cluster_id in range(len(np.unique(clusters_labels))): # For each cluster, select the first... cluster_inds = np.where(clusters_labels == cluster_id)[0] # ... at least members_per_group elements... n_select = np.minimum(members_per_group, len(cluster_inds)) if metric is not None and len(metric) == distance.shape[0]: # ...optionally according to some metric inds_select = np.argsort(metric[cluster_inds])[-n_select:] else: # ...otherwise, randomly inds_select = range(n_select) selection.append(cluster_inds[inds_select]) return np.unique(np.hstack(selection)).tolist()
[docs]def curve_elbow_point(vals, interactive): # Solution found in # vals = np.array(vals).flatten() if np.any(vals[0:-1] - vals[1:] < 0): vals = np.sort(vals) vals = vals[::-1] cumsum_vals = np.cumsum(vals) grad = np.gradient(np.gradient(np.gradient(cumsum_vals))) elbow = np.argmax(grad) # alternatively: # dif = np.diff(np.diff(np.diff(cumsum_vals))) # elbow = np.argmax(dif) + 2 if interactive: import matplotlib matplotlib.use(FiguresConfig.MATPLOTLIB_BACKEND) from matplotlib import pyplot pyplot.ion() fig, ax = pyplot.subplots() xdata = range(len(vals)) lines = [] lines.append(ax.plot(xdata, cumsum_vals, 'g*', picker=None, label="values' cumulative sum")[0]) lines.append(ax.plot(xdata, vals, 'bo', picker=None, label="values in descending order")[0]) lines.append(ax.plot(elbow, vals[elbow], "rd", label="suggested elbow point (maximum of third central difference)")[0]) lines.append(ax.plot(elbow, cumsum_vals[elbow], "rd")[0]) pyplot.legend(handles=lines[:2]) class MyClickableLines(object): def __init__(self, figure_no, axe, lines_list): self.x = None # self.y = None = axe title = "Mouse lef-click please to select the elbow point..." + \ "\n...or click ENTER to continue accepting our automatic choice in red..." self.lines = lines_list self.fig = figure_no def event_loop(self): self.fig.canvas.mpl_connect('button_press_event', self.onclick) self.fig.canvas.mpl_connect('key_press_event', self.onkey) self.fig.canvas.draw_idle() self.fig.canvas.start_event_loop(timeout=-1) return def onkey(self, event): if event.key == "enter": self.fig.canvas.stop_event_loop() return def onclick(self, event): if event.inaxes != self.lines[0].axes: return dist = np.sqrt((self.lines[0].get_xdata() - event.xdata) ** 2.0) # + (self.lines[0].get_ydata() - event.ydata) ** 2.) self.x = np.argmin(dist) self.fig.canvas.stop_event_loop() return click_point = MyClickableLines(fig, ax, lines) click_point.event_loop() if click_point.x is not None: elbow = click_point.x"\nmanual selection: " + str(elbow)) else:"\nautomatic selection: " + str(elbow)) return elbow else: return elbow
[docs]def spikes_events_to_time_index(spike_time, time): if spike_time < time[0] or spike_time > time[-1]: return None return np.argmin(np.abs(time - spike_time))
[docs]def compute_spikes_counts(spikes_times, time): spikes_counts = np.zeros(time.shape) for spike_time in spikes_times: ind = spikes_events_to_time_index(spike_time, time) if ind is not None: spikes_counts[ind] += 1 return spikes_counts
[docs]def spikes_rate_convolution(spike, spikes_kernel): if (spike != 0).any(): if len(spikes_kernel) > 1: return np.convolve(spike, spikes_kernel, mode="same") else: return spike * spikes_kernel else: return np.zeros(spike.shape)