nfp.layers.RBFExpansion

class RBFExpansion(*args, **kwargs)[source]

Bases: tensorflow.keras.layers.Layer

Layer to calculate radial basis function ‘embeddings’ for a continuous input variable. The width and location of each bin can be optionally trained. Essentially equivalent to a 1-hot embedding for a continuous variable.

Parameters

  • dimension (The total number of distance bins) –

  • init_gap (The initial width of each gaussian distribution) –

  • init_max_distance (the initial maximum value of the continuous variable) –

  • trainable (Whether the centers and gap parameters should be added as trainable) – NN parameters.

  • args (Any) –

  • kwargs (Any) –

Return type

Any

Methods

build

call

compute_mask

get_config

__call__(*args, **kwargs)

Call self as a function.

Parameters

  • args (Any) –

  • kwargs (Any) –

Return type

Any