Neuron circuits for spiking neural networks

A spiking neural network apparatus comprising:

• multiple neuron circuits interconnected by links, wherein each link connecting respective neuron circuits is associated with a respective weight for transmission of signals between the neuron circuits,
• wherein a neuron circuit of the multiple neuron circuits comprises:
• a digital transmitter for generating trigger signals that indicate a state of the neuron circuit, on outgoing links of the circuit, wherein the state is encoded in a time interval defined by the generated trigger signals;
• and
• a digital receiver comprising:
• a signal detector for detecting trigger signals on incoming links of the circuit, and
• digital accumulator logic adapted to:
• generate, in response to detection of a first trigger signal on a first incoming link, a weighted signal dependent on (i) a first time interval defined by the first trigger signal and (ii) a weight associated with the first incoming link, the weighted signal being generated based on a first thermometer encoded output value;
• accumulate a set of weighted signals generated from detected trigger signals on the incoming links;
• and
• determine the state of the neuron circuit based on the accumulated weighted signals,
• wherein the digital accumulator logic includes counter logic adapted to generate, for each incoming link, a corresponding thermometer-encoded output value, the corresponding thermometer-encoded output value being dependent on a time interval defined by a detected trigger signal on a corresponding incoming link.

A computer-implemented method comprising:

• generating a plurality of trigger signals on a plurality of outgoing links of a neuron circuit of a spiking neural network comprising:
• multiple neuron circuits each connected by a plurality of links, wherein the generated trigger signals indicate a state of the neuron circuit that is encoded in a time interval defined by the generated trigger signals;
• in response to detecting a first trigger signal of the generated trigger signals on a first incoming link of the neuron circuit, generating a weighted signal for the first incoming link that is dependent on a first time interval defined by the detected trigger signal, wherein a thermometer-encoded output value dependent on the first time interval defined by the first trigger signal is generated for the first incoming link, the weighted signal being generated based on the thermometer-encoded output value;
• and
• encoding the weighted signal associated with the first incoming link in a pulse signal embedded in the generated trigger signals on the first incoming link.

A computer system comprising:

• one or more computer processors;
• one or more computer readable storage media;
• and
• program instructions stored on the one or more computer readable storage media for execution by at least one of the one or more computer processors, the program instructions comprising:
• program instructions to generate a plurality of trigger signals on a plurality of outgoing links of a neuron circuit of a spiking neural network comprising:
• multiple neuron circuits each connected by a plurality of links, wherein the generated trigger signals indicate a state of the neuron circuit that is encoded in a time interval defined by the generated trigger signals;
• program instructions to, in response to detecting a first trigger signal of the generated trigger signals on a first incoming link of the neuron circuit, generating a weighted signal for the first incoming link that is dependent on a first time interval defined by the detected trigger signal, wherein a thermometer-encoded output value dependent on the first time interval defined by the first trigger signal is generated for the first incoming link, the weighted signal being generated based on the thermometer-encoded output value;
• and
• program instructions to encode the weighted signal associated with the first incoming link in a pulse signal embedded in the generated trigger signals on the first incoming link.