tfConstrainedGauss.solve_me package

Submodules

tfConstrainedGauss.solve_me.model_me module

class tfConstrainedGauss.solve_me.model_me.LayerPrecToCovMat(*args, **kwargs)

Bases: tensorflow.python.keras.engine.base_layer.Layer

Layer that converts precision to covariance matrix

__init__(n: int, non_zero_idx_pairs: List[Tuple[int, int]], non_zero_vals: numpy.array, **kwargs)

Constructor

Args:: n (int): Size of matirx non_zero_idx_pairs (List[Tuple[int,int]]): List of lower triangular index pairs for non-zero elements in precision matrix non_zero_vals (np.array): Non-zero values corresponding to non_zero_idx_pairs

call(inputs)

Construct the full precision matrix from the non-zero values and invert it to return the covariance matrix

Args:: inputs ([type]): Anything; used only to determine batch size: size should be batch_size x whatever
Returns:: [type]: Covariance matrix: batch_size x n x n

classmethod constructDiag(n: int, non_zero_idx_pairs: List[Tuple[int, int]], init_diag_val: float = 1.0, **kwargs)

Constructor with initial precision matrix that is diagonsl

Args:: n (int): Size of matirx non_zero_idx_pairs (List[Tuple[int,int]]): List of lower triangular index pairs for non-zero elements in precision matrix init_diag_val (float, optional): Initial diagonal value of precision matrix. Defaults to 1.0.
Returns:: LayerPrecToCovMat: layer

classmethod from_config(config)

get_config()

Returns the config of the layer.

A layer config is a Python dictionary (serializable) containing the configuration of a layer. The same layer can be reinstantiated later (without its trained weights) from this configuration.

The config of a layer does not include connectivity information, nor the layer class name. These are handled by Network (one layer of abstraction above).

Note that get_config() does not guarantee to return a fresh copy of dict every time it is called. The callers should make a copy of the returned dict if they want to modify it.

Returns:: Python dictionary.

property n_non_zero

Number of non-zero unique elements in precision matrix (excluding symmetry)

Returns:: int: Count

class tfConstrainedGauss.solve_me.model_me.ModelME(*args, **kwargs)

Bases: tensorflow.python.keras.engine.training.Model

Model for the MaxEnt method

__init__(inv_lyr: tfConstrainedGauss.solve_me.model_me.LayerPrecToCovMat, **kwargs)

Constructor

Args:: inv_lyr (LayerPrecToCovMat): Precision to covariance matrix layer

call(input_tensor, training=False)

Given some input_tensor that is only used to set the batch size (must be batch_size >= 2, else there are problems!): Calls the precision matrix to covariance matrix layer and returns the vector of the elements in the covariance matrix corresponding to the non-zero entries in the precision matrix.
Args:: input_tensor ([type]): Anything; used only to determine batch size: size should be batch_size x whatever
Returns:: [type]: Elements in the covariance matrix corresponding to the non-zero entries in the precision matrix.: batch_size x no_non_zero

classmethod from_config(config)

get_config()

Returns the config of the layer.

A layer config is a Python dictionary (serializable) containing the configuration of a layer. The same layer can be reinstantiated later (without its trained weights) from this configuration.

The config of a layer does not include connectivity information, nor the layer class name. These are handled by Network (one layer of abstraction above).

Note that get_config() does not guarantee to return a fresh copy of dict every time it is called. The callers should make a copy of the returned dict if they want to modify it.

Returns:: Python dictionary.

tfConstrainedGauss.solve_me.solve_me module

class tfConstrainedGauss.solve_me.solve_me.InputsME(n: int, non_zero_idx_pairs: List[Tuple[int, int]], target_cov_mat_non_zero: numpy.array, epochs: int = 100, learning_rate: float = 0.01, use_weighted_loss: bool = False, verbose: bool = False)

Bases: object

Inputs to the MaxEnt problem

__init__(n: int, non_zero_idx_pairs: List[Tuple[int, int]], target_cov_mat_non_zero: numpy.array, epochs: int = 100, learning_rate: float = 0.01, use_weighted_loss: bool = False, verbose: bool = False) → None

convert_mat_non_zero_to_inv_mat_non_zero(mat_non_zero: numpy.array): Wrapper around convert_mat_non_zero_to_inv_mat_non_zero

epochs: int = 100

learning_rate: float = 0.01

n: int

property n_non_zero

The number of unique non-zero elements in the precision matrix (excluding symmetry)

Returns:: int: Count

non_zero_idx_pairs: List[Tuple[int, int]]

report(): Print some information

target_cov_mat_non_zero: numpy.array

use_weighted_loss: bool = False

verbose: bool = False

class tfConstrainedGauss.solve_me.solve_me.ResultsME(inputs: tfConstrainedGauss.solve_me.solve_me.InputsME, trained_model: tfConstrainedGauss.solve_me.model_me.ModelME, init_prec_mat_non_zero: numpy.array, init_cov_mat_reconstructed_non_zero: numpy.array, learned_prec_mat_non_zero: numpy.array, learned_cov_mat: numpy.array)

Bases: object

Results of the MaxEnt problem

__init__(inputs: tfConstrainedGauss.solve_me.solve_me.InputsME, trained_model: tfConstrainedGauss.solve_me.model_me.ModelME, init_prec_mat_non_zero: numpy.array, init_cov_mat_reconstructed_non_zero: numpy.array, learned_prec_mat_non_zero: numpy.array, learned_cov_mat: numpy.array) → None