simrun ❭ modular_reduced_model_inference ❭ strategy ❭ Strategy_spatiotemporalRaisedCosine

Strategy_spatiotemporalRaisedCosine

class simrun.modular_reduced_model_inference.strategy.Strategy_spatiotemporalRaisedCosine(name, RaisedCosineBasis_spatial, RaisedCosineBasis_temporal)

Spatiotemporal raised cosine strategy.

Uses the :py:class:RaisedCosineBasis to create a set of basis functions.

Attention

The input data must contain the following keys: - spatiotemporalSa: The spatiotemporal synaptic activation patterns of shape (n_spatial_bins, n_temporal_bins, n_trials). - st: The spike times. - y: The labels. - ISI: The inter-spike intervals.

Parameters:

  • name (str) – The name of the strategy.

  • RaisedCosineBasis_spatial (RaisedCosineBasis) – The spatial basis functions \(\mathbf{g}(z)\).

  • RaisedCosineBasis_temporal (RaisedCosineBasis) – The temporal basis \(\mathbf{f}(t)\).

Attributes:

RaisedCosineBasis_spatial

The spatial basis functions \(\mathbf{g}(z)\).

Type:

RaisedCosineBasis

RaisedCosineBasis_temporal

The temporal basis functions \(\mathbf{f}(t)\).

Type:

RaisedCosineBasis

base_vectors_arrays_dict

The basis vectors for each group. basis vectors are of shape (n_trials, N_{tau}, N_{z}) These basis vectors are used for the optimizer, and are already multiplied with the data. Do not confuse them with the basis vectors of RaisedCosineBasis_spatial and RaisedCosineBasis_temporal, as the latter are not multiplied with the synapse activaiton data.

Type:

dict

groups

The list of groups. Usually simply ['EXC', 'INH'].

Type:

list

len_z

The length of the spatial domain i.e. the amount of spatial basis vectors.

Type:

int

len_t

The length of the temporal domain i.e. the amount of temporal basis vectors.

Type:

int

len_trials

The number of trials.

Type:

int

convert_x

The conversion function to convert the 1D learnable weight vector \(\mathbf{x}\) into a structured dictionary.

Type:

callable

_get_score

The cost function to provide to the optimizer.

Type:

callable

Methods:

_setup()

Compute the strategy’s basis vectors and set up the objective function.

compute_basis()

Compute the basis vectors for the dataset.

_get_x0()

Get an initial guess for the learnable weights \(\mathbf{x}\) and \(\mathbf{y}\) of the basis functions \(\mathbf{f}(\tau)\) and \(\mathbf{g}(z)\).

_convert_x_static(groups, len_z, x)

static Convert the input array \(\mathbf{x}\) into a dictionary of basis vectors.

_get_score_static(convert_x, base_vectors_arrays_dict, x)

static Calculate the weighted net input \(WNI(t)\) for the given weights \(\mathbf{x}\).

normalize(x, flipkey)

Normalize the kernel basis functions such that sum of all absolute values of all kernels is 1.

get_color_by_group(group)

Map groups to a color.

visualize(optimizer_output, only_successful, normalize)

Plot the basis functions.