Reconfigurable and customizable general-purpose circuits for neural networks

A method for producing spike-timing dependent plasticity in a reconfigurable neural network circuit, the method comprising:

• integrating input spikes in an integrate and fire electronic digital neuron, and upon the integrated input spikes exceeding a threshold, sending a spiking signal to digital synapses interconnecting the neuron to other neurons via a synapse array, wherein the synapse array comprises multiple digital synapses interconnecting a plurality of integrate and fire electronic digital neurons, each synapse interconnecting an axon of a pre-synaptic neuron with a dendrite of a post-synaptic neuron;
• providing control signals to reconfiguration inputs of a neuron to reconfigure neuron operation parameters including one or more of:
• spiking, integration, learning and communication;
• generating signals for phased operation of the neurons and updating the synapses, and sequential access to the synapse array by spiking;
• determining which neurons among said neurons spiked in a previous timestep, and resetting the potential of the spiking neurons;
• performing neuron spiking by pulsing an axon row in the synapse array, reading value of each synapse, and passing the value to a connected neuron to obtain an external input for each neuron;
• each neuron checking a dendrite column in the synapse array for synapses in their pulsed state, reading the synapse values, and integrating each synapse input as external input to the neuron potential;
• depending on the time elapsed since each neuron fired, probabilistically changing the synapse value and writing a new synapse value into the synapse array;
• and
• determining which neurons will spike the next timestep by comparing each neuron potential against a threshold.