Downsample by a factor of 2 using reflecting boundary conditions. This function convolves `x` with filter `f` with reflecting boundary conditions

Upsample by a factor of 2 using transposed reflecting boundary conditions. This function undecimates `x` along the axis specified by `direction` an

Returns an image of scales.

get modes on frequency domain: 'random' means sampling randomly; 'else' means sampling the lowest modes;

The inverse of tf.nn.softplus().

Computes the loss on a matrix. Args: x: The residual for which the loss is being computed. Must be a rank-2 tensor, where the inner

Sets up the loss function. Args: num_dims: The number of dimensions of the input to come. float_dtype: The floating point precision o

Checks the inputs to _downsample() and _upsample().

Returns an image of alphas.

Generates the analysis and synthesis filters for a kind of wavelet. Currently only supports wavelet types where all filters have an odd length. T

r"""Approximate the distribution's log-partition function with a 1D spline. Because the partition function (Z(\alpha) in the paper) of the distri

r"""Implements the negative log-likelihood (NLL). Specifically, we implement -log(p(x | 0, \alpha, c) of Equation 16 in the paper as nllfun(x

Pads `x` with reflecting conditions above and/or below it along some axis. Pads `x` with reflecting conditions for `padding_below` entries below th

Transforms a batch of images to a matrix.

Results visualization

Returns an image of degrees of freedom, for the Student's T model.

Flattens a wavelet decomposition into an image-like single Tensor. construct() produces wavelet decompositions in the form of nested tuples, whic

Compute the CRPS of observations x relative to gaussian distributed forecasts with mean, sigma. CRPS(N(mu, sig^2); x) Formula

Probability density function of a univariate standard Gaussian distribution with zero mean and unit variance.

Returns the maximum number of levels that construct() can support. Args: sz: A tuple of ints representing some input size (batch, width, height

The inverse of tf.nn.sigmoid().

Applies a curve to alpha >= 0 to compress its range before interpolation. This is a weird hand-crafted function designed to take in alpha values an

from: https://github.com/husnejahan/DeepAR-pytorch/blob/master/net.py Compute using gaussian the log-likehood which needs to be maximized. Ig

SpeedUp version of Autocorrelation (a batch-normalization style design) This is for the inference phase.

SpeedUp version of Autocorrelation (a batch-normalization style design) This is for the training phase.

Returns a list of time features that will be appropriate for the given frequency string. Parameters ---------- freq_str Frequ