sax.nn.core module#
SAX Neural Network Core Utils
- preprocess(*params)[source]#
preprocess parameters
- > Note: (1) all arguments are first casted into the same shape. (2) then pairs
of arguments are divided into each other to create relative arguments. (3) all arguments are then stacked into one big tensor
- Parameters:
params (ComplexArrayND) –
- Return type:
ComplexArrayND
- dense(weights, *params, x_norm=(0.0, 1.0), y_norm=(0.0, 1.0), preprocess=<function preprocess>, activation=<PjitFunction of <function leaky_relu>>)[source]#
simple dense neural network
- Parameters:
weights (Dict[str, Array]) –
params (ComplexArrayND) –
x_norm (Tuple[float, float]) –
y_norm (Tuple[float, float]) –
preprocess (Callable) –
activation (Callable) –
- Return type:
ComplexArrayND
- generate_dense_weights(key, sizes, input_names=None, output_names=None, preprocess=<function preprocess>)[source]#
Generate the weights for a dense neural network
- Parameters:
key (Union[int, Array]) –
sizes (Tuple[int, ...]) –
input_names (Optional[Tuple[str, ...]]) –
output_names (Optional[Tuple[str, ...]]) –
- Return type:
Dict[str, ComplexArrayND]