Decoder Class

neuraldecoding.decoder

Decoder class for converting offline or online neural data into behavioural data using methods provided.

Classes

Decoder

class Decoder(ABC)

Abstract base class for decoders. This class provides the common interface and functionality for all decoder implementations. It handles model instantiation, model loading, and defines the abstract methods.

Parameters:

• cfg : dict

  Configuration dictionary with the following structure, typically parsed by config parser:

  decoder_struct = {'model': {'name': str, 
                               'parameters': DictConfig, 
                               'input_shape': ListConfig, 
                               'output_shape': ListConfig}, 
                     'stabilization': {'name': str, 
                                       'parameters': DictConfig, 
                                       'date_0': str, 
                                       'date_k': str},
                     'fpath': str
                  }

Attributes:

• model : object

  The instantiated model object

• fpath : str

  File path to saved model

• input_shape : tuple

  Expected input tensor dimensions

• output_shape : tuple

  Expected output tensor dimensions

Methods

__init__(cfg)

Initialize the decoder with configuration, instantiates model, sets path to saved model, sets input and output shape, and loads the model.

Parameters:

• cfg : dict

  Config dictionary as described above

load_model(fpath=None)

Load model parameters from file. Option to override the file path to the saved model.

Parameters:

• fpath : str, optional

  Path to model parameters. If None, uses the attribute fpath.

get_decoder()

Create a deep copy of the decoder with loaded parameters.

Returns:

• decoder : Decoder

  Deep-copied decoder instance with parameters loaded from file

predict(neural_data) (abstract)

Predict outputs from neural data.

Parameters:

• neural_data :

  Input neural data

Returns:

• predictions :

  Decoded predictions

get_input_shape()

Get the expected input shape.

Returns:

• shape : numpy.ndarray

  Input dimensions

get_output_shape()

Get the expected output shape.

Returns:

• shape : numpy.ndarray

  Output dimensions