Decoder

class behavenet.models.decoders.Decoder(hparams)

Bases: BaseModel

General wrapper class for encoding/decoding models.

Methods Summary

build_model()	Construct the model using hparams.
forward(x)	Process input data.
loss(data[, accumulate_grad, chunk_size])	Calculate negative log-likelihood loss for supervised models.

Methods Documentation

build_model()

Construct the model using hparams.

forward(x)

Process input data.

loss(data, accumulate_grad=True, chunk_size=200, **kwargs)

Calculate negative log-likelihood loss for supervised models.

The batch is split into chunks if larger than a hard-coded chunk_size to keep memory requirements low; gradients are accumulated across all chunks before a gradient step is taken.

Parameters:

• data (dict) – signals are of shape (1, time, n_channels)

• accumulate_grad (bool, optional) – accumulate gradient for training step

• chunk_size (int, optional) – batch is split into chunks of this size to keep memory requirements low

Returns:

• ‘loss’ (float): total loss (negative log-like under specified noise dist)

• ’r2’ (float): variance-weighted $R^2$ when noise dist is Gaussian

• ’fc’ (float): fraction correct when noise dist is Categorical

Return type:

dict