rockpool.weights.reservoirweights

Utility functions for generating and manipulating networks

Functions

DiscretiseWeightMatrix(weights[, ...])

DiscretiseWeightMatrix - Discretise a weight matrix by strength

add_random_long_range(weights_res, ...[, ...])

combine_ff_rec_stack(weights_ff, weights_rec)

combine_ff_rec_stack - Combine a FFwd and Recurrent weight matrix into a single recurrent weight matrix :param weights_ff: MxN np.ndarray :param weights_rec: NxN np.ndarray

digital(shape[, connectivity, ratio_exc, ...])

digital - Create a weight matrix that conforms the specifications of the Dynapse Chip

dynapse_conform(shape[, connectivity, ...])

dynapse_conform - Create a weight matrix that conforms the specifications of the Dynapse Chip

gen_sparse_partitioned_network(partition_sizes)

Generate weight matrices that embody sparse networks with defined partition sizes

iaf_sparse_net([res_size, mean, std, density])

iaf_sparse_net - Return a random sparse reservoir, scaled for a standard IAF spiking network

in_res_digital(size[, input_density, ...])

in_res_digital - Create input weights and recurrent weights for reservoir, respecting dynapse specifications

in_res_dynapse(size[, input_density, ...])

in_res_dynapse - Create input weights and recurrent weights for reservoir, respecting dynapse specifications

in_res_dynapse_flex(size, size_in[, ...])

in_res_dynapse_flex - Like in_res_dynapse but number of input weights can be chosen

inp_to_rec([size_in, size_rec, ...])

inp_to_rec - Create an integer weight matrix that serves as input weights to the

one_dim_exc_res(size, n_neighbour[, ...])

one_dim_exc_res - Recurrent weight matrix where each neuron is connected to its n_neighbour nearest neighbours on a 1D grid.

partitioned_2d_reservoir([size_in, ...])

ring_reservoir([size_in, size_rec, ...])

rndm_ei_net(num_exc, num_inh[, ...])

rndm_ei_net - Generate a random nicely-tuned real-valued reservoir matrix :param num_exc: Number of excitatory neurons in the network :param num_inh: Number of inhibitory neurons in the network :param ratio_inh_exc: Factor relating total inhibitory and excitatory weight (w_inh = ratio_inh_exc * w_exc) default: 1 :param rndm_weight_fct: Function used to draw initial random weights.

rndm_sparse_ei_net(res_size[, connectivity, ...])

rndm_sparse_ei_net - Return a (sparse) matrix defining reservoir weights

two_dim_exc_res(size, n_neighbour, ...[, ...])

unit_lambda_net(res_size)

unit_lambda_net - Generate a network from Norm(0, sqrt(N))

wilson_cowan_net(num_nodes[, self_exc, ...])

wilson_cowan_net - FUNCTION Define a Wilson-Cowan network of oscillators

wipe_non_switiching_eigs(weights[, ...])

wipe_non_switiching_eigs - Eliminate eigenvectors that do not lead to a partition switching