Exponential-Golomb Code¶

measure_expgolomb_rate(
q_module: Module,
q_step: DescriptorNN,
expgol_cnt: DescriptorNN,
) → DescriptorNN[source]¶

Get the rate associated with the current parameters.

Returns:

The rate of the different modules wrapped inside a dictionary

of float. It does not return tensor so no back propagation is possible

Return type:

DescriptorNN

Parameters:

  • q_module (Module)

  • q_step (DescriptorNN)

  • expgol_cnt (DescriptorNN)

exp_golomb_nbins(symbol: Tensor, count: int = 0) → Tensor[source]¶

Compute the number of bits required to encode a Tensor of integers using an exponential-golomb code with exponent count.

Parameters:

  • symbol (Tensor) – Tensor to encode

  • count (int, optional) – Exponent of the exp-golomb code. Defaults to 0.

Returns:

Number of bits required to encode all the symbols.

Return type:

Tensor