# -*- coding: utf-8 -*-
#
#
# TheVirtualBrain-Framework Package. This package holds all Data Management, and
# Web-UI helpful to run brain-simulations. To use it, you also need to download
# TheVirtualBrain-Scientific Package (for simulators). See content of the
# documentation-folder for more details. See also http://www.thevirtualbrain.org
#
# (c) 2012-2024, 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 <http://www.gnu.org/licenses/>.
#
#
# CITATION:
# When using The Virtual Brain for scientific publications, please cite it as explained here:
# https://www.thevirtualbrain.org/tvb/zwei/neuroscience-publications
#
#
"""
.. moduleauthor:: Ionel Ortelecan <ionel.ortelecan@codemart.ro>
.. moduleauthor:: Mihai Andrei <mihai.andrei@codemart.ro>
"""
import numpy
import six
from tvb.basic.logger.builder import get_logger
# To plot nullclines we need a function that computes contours of a scalar field.
# We use the internal matplotlib one.
# Now that newer matplotlib versions have changed this internal API it would be a
# good idea to use a proper library for this.
try:
from matplotlib import _contour
# old matplotlib
def nullcline(x, y, z):
c = _contour.QuadContourGenerator(x, y, z, None, True, 0)
segments = c.create_contour(0.0)
return segments[0]
except ImportError:
import matplotlib.pyplot as plt
# new matplotlib, version 3.8.3
[docs]
def nullcline(x, y, z):
c = plt.contour(x, y, z, levels=[0.0])
segments = c.collections[0].get_paths()
return segments
# how much courser is the grid used to show the vectors
GRID_SUBSAMPLE = 2
# Set the resolution of the phase-plane and sample trajectories.
NUMBEROFGRIDPOINTS = GRID_SUBSAMPLE * 42
class _PhaseSpace(object):
"""
Dimensionality independent code
"""
def __init__(self, model, integrator):
self.log = get_logger(self.__class__.__module__)
self.model = model
self.integrator = integrator
def _compute_trajectories(self, states, n_steps):
"""
A vectorized method of computing a number of trajectories in parallel.
Returns a collection of nvar-dimensional trajectories.
"""
scheme = self.integrator.scheme
trajs = numpy.zeros((n_steps + 1, self.model.nvar, len(states), self.model.number_of_modes))
# reshape to what dfun expects: from n, sv to sv, n, mode
states = numpy.tile(states.T[:, :, numpy.newaxis], self.model.number_of_modes)
trajs[0, :] = states
no_coupling = numpy.zeros((self.model.nvar, states.shape[1], self.model.number_of_modes))
# grow trajectories step by step
for step in range(n_steps):
states = scheme(states, self.model.dfun, no_coupling, 0.0, 0.0)
trajs[step + 1, :] = states
if numpy.isnan(trajs).any():
self.log.warning("NaN in trajectories")
return trajs
def get_axes_ranges(self, sv):
lo, hi = self.model.state_variable_range[sv]
default_range = hi - lo
min_val = lo - 4.0 * default_range
max_val = hi + 4.0 * default_range
return min_val, max_val, lo, hi
[docs]
class PhasePlane(_PhaseSpace):
"""
Responsible with computing phase space slices and trajectories.
A collection of math-y utilities it is view independent (holds no state related to views).
"""
@staticmethod
def _create_mesh_jitter():
shape = NUMBEROFGRIDPOINTS, NUMBEROFGRIDPOINTS
d = 1.0 / (4 * NUMBEROFGRIDPOINTS)
return numpy.random.normal(0, d, shape), numpy.random.normal(0, d, shape)
@staticmethod
def _get_mesh_grid(x_range, y_range, noise=None):
"""
Generate the phase-plane gridding based on the given
state-variable indices and their range values.
"""
xlo, xhi = x_range
ylo, yhi = y_range
xg = numpy.mgrid[xlo:xhi:(NUMBEROFGRIDPOINTS * 1j)]
yg = numpy.mgrid[ylo:yhi:(NUMBEROFGRIDPOINTS * 1j)]
xgr, ygr = numpy.meshgrid(xg, yg)
if noise:
# add scaled noise
xgr += noise[0] * (xhi - xlo)
ygr += noise[1] * (yhi - ylo)
return xgr, ygr
def _calc_phase_plane(self, state, svx_ind, svy_ind, xg, yg):
"""
Computes a 2d axis aligned rectangle of the vector field returning a u, v vector field.
The slice passes through the `state` point and varies along the axes given by svx_ind and svy_ind.
The last 2 parameters specify the mesh in the varying directions. To be computed by _get_mesh_grid
Vectorized function, it evaluates all grid points at once as if they were connectivity nodes.
"""
state_variables = numpy.tile(state, (NUMBEROFGRIDPOINTS ** 2, 1))
for mode_idx in range(self.model.number_of_modes):
state_variables[svx_ind, :, mode_idx] = xg.flat
state_variables[svy_ind, :, mode_idx] = yg.flat
no_coupling = numpy.zeros((self.model.nvar, state_variables.shape[1], self.model.number_of_modes))
d_grid = self.model.dfun(state_variables, no_coupling)
flat_uv_grid = d_grid[[svx_ind, svy_ind], :, :] # subset of the state variables to be displayed
u, v = flat_uv_grid.reshape((2, NUMBEROFGRIDPOINTS, NUMBEROFGRIDPOINTS, self.model.number_of_modes))
if numpy.isnan(u).any() or numpy.isnan(v).any():
self.log.error("NaN")
return u, v
[docs]
class PhasePlaneD3(PhasePlane):
"""
Provides data for a d3 client
"""
def __init__(self, model, integrator):
PhasePlane.__init__(self, model, integrator)
self.mode = 0
self.svx_ind = 0 # x-axis: 1st state variable
if self.model.nvar > 1:
self.svy_ind = 1 # y-axis: 2nd state variable
else:
self.svy_ind = 0
self.x_range = None
self.y_range = None
# Set up a vector containing the default state-variable values
svr = self.model.state_variable_range
sv_mean = numpy.array([svr[key].mean() for key in self.model.state_variables])
sv_mean = sv_mean.reshape((self.model.nvar, 1, 1))
self.default_sv = sv_mean.repeat(self.model.number_of_modes, axis=2)
self.update_integrator_clamping()
self._jitter = None # self._create_mesh_jitter()
[docs]
def update_integrator_clamping(self):
clamped_sv_indices = [i for i in range(self.model.nvar) if i not in [self.svx_ind, self.svy_ind]]
if clamped_sv_indices:
self.integrator.clamped_state_variable_indices = numpy.array(clamped_sv_indices)
self.integrator.clamped_state_variable_values = self.default_sv[
self.integrator.clamped_state_variable_indices]
else:
self.integrator.clamped_state_variable_indices = None
self.integrator.clamped_state_variable_values = None
[docs]
def update_axis(self, mode, svx, svy, x_range, y_range, state_vars):
self.mode = int(mode)
self.svx_ind = self.model.state_variables.index(svx)
self.svy_ind = self.model.state_variables.index(svy)
self.x_range = x_range
self.y_range = y_range
for name, val in six.iteritems(state_vars):
k = self.model.state_variables.index(name)
self.default_sv[k] = val
self.update_integrator_clamping()
[docs]
def compute_phase_plane(self):
"""
:return: A json representation of the phase plane.
"""
x, y = self._get_mesh_grid(self.x_range, self.y_range, noise=self._jitter)
u, v = self._calc_phase_plane(self.default_sv, self.svx_ind, self.svy_ind, x, y)
u = u[..., self.mode] # project on active mode
v = v[..., self.mode]
xnull = [{'path': segment.vertices.tolist(), 'nullcline_index': 0} for segment in nullcline(x, y, u)]
ynull = [{'path': segment.vertices.tolist(), 'nullcline_index': 1} for segment in nullcline(x, y, v)]
# a courser mesh for the arrows
xsmall = x[::GRID_SUBSAMPLE, ::GRID_SUBSAMPLE]
ysmall = y[::GRID_SUBSAMPLE, ::GRID_SUBSAMPLE]
usmall = u[::GRID_SUBSAMPLE, ::GRID_SUBSAMPLE]
vsmall = v[::GRID_SUBSAMPLE, ::GRID_SUBSAMPLE]
d = numpy.dstack((xsmall, ysmall, usmall, vsmall))
d = d.reshape(((NUMBEROFGRIDPOINTS // GRID_SUBSAMPLE) ** 2, 4)).tolist()
return {'plane': d, 'nullclines': xnull + ynull}
def _state_dict_to_array(self, state):
arr = numpy.zeros(len(self.model.state_variables))
for svn, v in six.iteritems(state):
svn_idx = self.model.state_variables.index(svn)
arr[svn_idx] = v
return arr
[docs]
def trajectories(self, starting_points, n_steps=512):
"""
:param starting_points: A list of starting points represented as dicts of state_var_name to value
:return: a tuple of trajectories and signals
"""
starting_points = numpy.array([self._state_dict_to_array(s) for s in starting_points])
traj = self._compute_trajectories(starting_points, n_steps) # point_on_traj_idx, sv_idx, traj_idx, mode
# reshape it and project it on the plane defined by the current axis state vars
traj = traj.transpose(2, 0, 1, 3) # traj_idx, point, sv_idx, mode
trajectory = traj[:, :, [self.svx_ind, self.svy_ind], self.mode] # traj_idx, points, x, y
# signals for last trajectory
signal_x = numpy.arange(n_steps + 1) * self.integrator.dt
signals = [list(zip(signal_x, traj[-1, :, i, self.mode].tolist())) for i in [self.svx_ind, self.svy_ind]]
return trajectory.tolist(), signals
[docs]
class PhaseLineD3(_PhaseSpace):
def __init__(self, model, integrator):
_PhaseSpace.__init__(self, model, integrator)
self.mode = 0
def _grid(self):
svr = self.model.state_variable_range
xlo, xhi = svr[self.model.state_variables[0]]
return numpy.linspace(xlo, xhi, NUMBEROFGRIDPOINTS)
[docs]
def compute_phase_plane(self):
xg = self._grid()
# dfun modifies state in place, so we need to copy xg
state = xg.reshape((1, NUMBEROFGRIDPOINTS, 1)).copy() # will broadcast to modes
no_coupling = numpy.zeros((self.model.nvar, state.shape[1], self.model.number_of_modes))
u = self.model.dfun(state, no_coupling)
u = u[0, :, self.mode]
d = numpy.vstack((xg, u)).T
if numpy.isnan(d).any():
self.log.error("NaN")
# find zeroes. This method is not exact
zero_crossings = numpy.where(numpy.diff(numpy.sign(u)))[0]
zero_crossings = xg[zero_crossings]
return {'signal': d.tolist(), 'zeroes': zero_crossings.tolist()}
[docs]
def update_axis(self, mode, svx, x_range):
self.mode = int(mode)
svr = self.model.state_variable_range
svr[svx][:] = x_range
[docs]
def phase_space_d3(model, integrator):
"""
:return: A phase plane or a phase line depending on the dimensionality of the model
"""
if model.nvar == 1:
return PhaseLineD3(model, integrator)
else:
return PhasePlaneD3(model, integrator)