8.1. Neuronal Firing

Neural network generic firing abstraction.

class fhez.nn.traverse.firing.Firing(graph=None)

Simple exhaustive neuronal firing calculation.

adaptation()

Correct nodes based on learnt gradient.

correction(signals, receptors)

Calculate/ learn correction necessary to become closer to our goal.

Parameters

  • signals – signal to be induced in corresponding receptor

  • receptors – receptor to be signaled

property graph

Get neuron graph to fire.

harvest(node_names: list)

Harvest forward response from neuronal firing, using probes.

This will replay the last node to calculate its output.

probe_shape(lst: list)

Get the shape of a list, assuming each sublist is the same length.

This function is recursive, sending the sublists down and terminating once a type error is thrown by the final point being a non-list

stimulate(neurons: numpy.ndarray, signals: numpy.ndarray, receptor='forward', debug=False)

Stimulate a set of receptors with a set of signals for response.

Breadth first stimulation of neurons/ nodes. Note that this is a single simultaneous stimulation and subsequent response. If a neuron does not fire I.E produces no (None) result then that neuron will not be followed until it does produce a result.

Parameters

  • receptors (list(str)) – list of node names to recieve stimulus

  • signals (np.ndarray or compatible) – positional list of signals for the equally positioned receptor

  • receptor (str) – Name of function/ sequence of functions to call of nodes

fhez.nn.traverse.firing.NeuronalFiring

alias of fhez.nn.traverse.firing.Firing