echotorch.tools package

Submodules

echotorch.utils.error_measures module

echotorch.utils.error_measures.cumperplexity(output_probs, targets, log=False)

Cumulative perplexity :param output_probs: :param targets: :param log: :return:

echotorch.utils.error_measures.generalized_squared_cosine(Sa, Ua, Sb, Ub)

Generalized square cosine :param Sa: :param Ua: :param Sb: :param Ub: :return:

echotorch.utils.error_measures.mse(outputs, targets)

Mean square error :param outputs: Module’s outputs :param targets: Target signal to be learned :return: Mean square deviation

echotorch.utils.error_measures.nmse(outputs, targets)

Normalized mean square error :param outputs: Module’s output :param targets: Target signal to be learned :return: Normalized mean square deviation

echotorch.utils.error_measures.nrmse(outputs, targets)

Normalized root-mean square error :param outputs: Module’s outputs :param targets: Target signal to be learned :return: Normalized root-mean square deviation

echotorch.utils.error_measures.perplexity(output_probs, targets, log=False)

Perplexity :param output_probs: Output probabilities for each word/tokens (length x n_tokens) :param targets: Real word index :return: Perplexity

echotorch.utils.error_measures.rmse(outputs, targets)

Root-mean square error :param outputs: Module’s outputs :param targets: Target signal to be learned :return: Root-mean square deviation

echotorch.utils.utility_functions module

echotorch.utils.utility_functions.align_pattern(interpolation_rate, truth_pattern, generated_pattern)

Align pattern :param interpolation_rate: :param truth_pattern: :param generated_pattern: :return:

echotorch.utils.utility_functions.average_prob(tensor, dim=0)

Average probabilities through time :param tensor: :param dim: :return:

echotorch.utils.utility_functions.compute_correlation_matrix(states)

Compute correlation matrix :param states: :return:

echotorch.utils.utility_functions.compute_similarity_matrix(svd_list)

Compute similarity matrix :param svd_list: :return:

echotorch.utils.utility_functions.compute_singular_values(stats)

Compute singular values :param states: :return:

echotorch.utils.utility_functions.deep_spectral_radius(m, leaky_rate)

Compute spectral radius of a square 2-D tensor for stacked-ESN :param m: squared 2D tensor :param leaky_rate: Layer’s leaky rate :return:

echotorch.utils.utility_functions.find_phase_shift(p, y, interpolation_rate, error_measure=<function nrmse>)

Find phase shift :param s1: :param s2: :param window_size: :return:

echotorch.utils.utility_functions.max_average_through_time(tensor, dim=0)

Max average through time :param tensor: :param dim: Time dimension :return:

echotorch.utils.utility_functions.normalize(tensor, dim=1)

Normalize a tensor on a single dimension :param t: :return:

echotorch.utils.utility_functions.spectral_radius(m)

Compute spectral radius of a square 2-D tensor :param m: squared 2D tensor :return:

Module contents

echotorch.utils.align_pattern(interpolation_rate, truth_pattern, generated_pattern)

Align pattern :param interpolation_rate: :param truth_pattern: :param generated_pattern: :return:

echotorch.utils.compute_correlation_matrix(states)

Compute correlation matrix :param states: :return:

echotorch.utils.nrmse(outputs, targets)

Normalized root-mean square error :param outputs: Module’s outputs :param targets: Target signal to be learned :return: Normalized root-mean square deviation

echotorch.utils.nmse(outputs, targets)

Normalized mean square error :param outputs: Module’s output :param targets: Target signal to be learned :return: Normalized mean square deviation

echotorch.utils.rmse(outputs, targets)

Root-mean square error :param outputs: Module’s outputs :param targets: Target signal to be learned :return: Root-mean square deviation

echotorch.utils.mse(outputs, targets)

Mean square error :param outputs: Module’s outputs :param targets: Target signal to be learned :return: Mean square deviation

echotorch.utils.perplexity(output_probs, targets, log=False)

Perplexity :param output_probs: Output probabilities for each word/tokens (length x n_tokens) :param targets: Real word index :return: Perplexity

echotorch.utils.cumperplexity(output_probs, targets, log=False)

Cumulative perplexity :param output_probs: :param targets: :param log: :return:

echotorch.utils.spectral_radius(m)

Compute spectral radius of a square 2-D tensor :param m: squared 2D tensor :return:

echotorch.utils.deep_spectral_radius(m, leaky_rate)

Compute spectral radius of a square 2-D tensor for stacked-ESN :param m: squared 2D tensor :param leaky_rate: Layer’s leaky rate :return:

echotorch.utils.normalize(tensor, dim=1)

Normalize a tensor on a single dimension :param t: :return:

echotorch.utils.average_prob(tensor, dim=0)

Average probabilities through time :param tensor: :param dim: :return:

echotorch.utils.max_average_through_time(tensor, dim=0)

Max average through time :param tensor: :param dim: Time dimension :return:

echotorch.utils.show_3d_timeseries(ts, title)

Show 3D timeseries :param axis: :param title: :return:

echotorch.utils.show_2d_timeseries(ts, title)

Show 2D timeseries :param ts: :param title: :return:

echotorch.utils.show_1d_timeseries(ts, title, xmin, xmax, ymin, ymax, start=0, timesteps=-1)

Show 1D time series :param ts: :param title: :return:

echotorch.utils.neurons_activities_1d(stats, neurons, title, colors, xmin, xmax, ymin, ymax, timesteps=-1, start=0)

Display neurons activities :param stats: :param neurons: :return:

echotorch.utils.neurons_activities_2d(stats, neurons, title, colors, timesteps=-1, start=0)

Display neurons activities on a 2D plot :param stats: :param neurons: :param title: :param timesteps: :param start: :return:

echotorch.utils.neurons_activities_3d(stats, neurons, title, timesteps=-1, start=0)

Display neurons activities on a 3D plot :param stats: :param neurons: :param title: :param timesteps: :param start: :return:

echotorch.utils.plot_singular_values(stats, title, xmin, xmax, ymin, ymax, log=False)

Plot singular values :param stats: :param title: :param timestep: :param start: :return:

echotorch.utils.compute_singular_values(stats)

Compute singular values :param states: :return:

echotorch.utils.generalized_squared_cosine(Sa, Ua, Sb, Ub)

Generalized square cosine :param Sa: :param Ua: :param Sb: :param Ub: :return:

echotorch.utils.compute_similarity_matrix(svd_list)

Compute similarity matrix :param svd_list: :return:

echotorch.utils.show_similarity_matrix(sim_matrix, title, column_labels=None, row_labels=None)

Show similarity matrix :param sim_matrix: :return:

echotorch.utils.show_conceptors_similarity_matrix(conceptors, title)

Show conceptors similarity matrix :param conceptors: :param title: :return:

echotorch.utils.show_sv_for_increasing_aperture(conceptor, factor, title)

Show singular values for increasing aperture :param conceptors: :param factor: :param title: :return:

echotorch.utils.find_phase_shift(p, y, interpolation_rate, error_measure=<function nrmse>)

Find phase shift :param s1: :param s2: :param window_size: :return: