Generate a new array that chops the given array along the given axis into overlapping frames. This code has been implemented by Anne Archibald

Yield successive n-sized chunks from sequence.

WRITEME Parameters ---------- X : list of complex vectors Notes ----- This function assumes X as 2D

Compute KL divergence between a bunch of univariate Gaussian distributions with the given means and log-variances. We do KL(N(mu_left, lo

Initialize parameters following the way proposed in the paper.

Apply real inverse FFT to X (scipy) Parameters ---------- X : list of lists or ndArrays

States if a number is a power of two (Author: A.Polino)

Apply real inverse FFT to X (numpy) Parameters ---------- X : list of lists or ndArrays

Apply real FFT to X (numpy) Parameters ---------- X : list of lists or ndArrays

Apply real FFT to X (scipy) Parameters ---------- X : list of lists or ndArrays

Most parts are copied from torch.nn.LSTMCell.

Given variable lengths sequences, pad zeros w.r.t to the maximum length sequences and create a dense design matrix

Given variable lengths sequences, pad zeros w.r.t to the maximum length sequences and create a dense design matrix

Args: input_: A (batch, input_size) tensor containing input features. hx: A tuple (h_0, c_0), which c

Globally normalize X into zero mean and unit variance Parameters ---------- X : list of lists or ndArrays

Infer latent variables for a given batch of sentences ``x''.

Apply short-time Fourier transform to X Parameters ---------- X : list of lists or ndArrays

Unify the norm of each sequence in X Parameters ---------- X : list of lists or ndArrays avr_nom : Scal

Reverse each sequence of X Parameters ---------- X : list of ndArrays or lists

Standardize X such that X \in [0, 1] Parameters ---------- X : list of lists or ndArrays X_max : Scalar

Apply short-time Fourier transform to X Parameters ---------- X : list of lists or ndArrays