Product

The Product rule computes a neuron’s activation as the product of its inputs. This type of multiplicative interaction is used in certain neural architectures, including Sigma-Pi and Long Short-Term Memory (LSTM) networks.

The update equation depends on whether weights are used:

• Without weights: \(a = \prod_{i=1}^{N} x_i\)

• With weights: \(a = \prod_{i=1}^{N} x_i \cdot w_i\)

Alternatively, in networks using spiking neurons, the post-synaptic response (PSR) is used in place of \(x_i \cdot w_i\).

If a neuron has no incoming connections, its activation is set to zero. Optional noise can be added to the result.

Parameters

• Use Weights: If enabled, activation is computed using weighted inputs (or post-synaptic responses for spiking neurons). If disabled, raw activations are multiplied.

For all other parameters, see common neuron properties