Spiking network apparatus and method with bimodal spike-timing dependent plasticity
AssigneeBrain Corporation
InventorsMicah Richert, Filip Piekniewski, Eugene Izhikevich, Sach Sokol
Life of this patent
Estimating the term…
Abstract
Apparatus and methods for learning in response to temporally-proximate features. In one implementation, an image processing apparatus utilizes bi-modal spike timing dependent plasticity in a spiking neuron network. Based on a response by the neuron to a frame of input, the bi-modal plasticity mechanism is used to depress synaptic connections delivering the present input frame and to potentiate synaptic connections delivering previous and/or subsequent frames of input. The depression of near-contemporaneous input prevents the creation of a positive feedback loop and provides a mechanism for network response normalization.
Claims24
A method of detecting a representation of an object in a sequence of frames with a spiking neuron network, the method comprising:
- communicating a spiking signal to a neuron via a plurality of connections;
- and
- based on a response generated by the neuron, the response based on a first portion of the spiking signal corresponding to a first frame of a sequence of frames:
- depressing a first connection of the plurality of connections, the first connection providing the first portion of the spiking signal;
- and
- potentiating a second connection of the plurality of connections, the second connection providing a second portion of the spiking signal corresponding to a second frame of the sequence of frames, the second frame adjacent the first frame;
- wherein the spiking signal is configured based on one or more frames of the sequence of frames.
A computerized spiking neuron apparatus configured to encode sensory input comprising a plurality of views of an object, the apparatus comprising a plurality of computer-readable instructions configured to, when executed:
- encode individual ones of the plurality of views into a spike output by a plurality of first layer neurons;
- provide the spike output to a second layer neuron via a plurality of connections associated with individual ones of the plurality of first layer neurons;
- based on a response generated by the second layer neuron:
- depress first connection of the plurality of connections, the first connection configured to provide a portion of the spike output to the neuron within first time interval the response;
- and
- potentiate second connection of the plurality of connections, the second connection configured to provide a portion of the spike output to the neuron within second time interval from the response.
A method of updating a connection providing stimulus to an artificial spiking neuron, the method comprising:
- based on a response generated by the neuron:
- depressing the connection when the stimulus is within first time interval from the response;
- and
- potentiating the connection when the stimulus is outside the first time interval from the response.
A non-transitory computer-readable apparatus configured to store one or more processes thereon, the one or more processes comprising a plurality of instructions configured to, when executed:
- send an encoded digital signal to a neuron via one or more connections;
- receive a response from the neuron;
- based at least in part on the response, determine a first and second interval;
- demote a first connection of the one or more connections during the first interval;
- and
- promote a second connection of the one or more connection during the second interval;
- wherein the encoded digital signal is configured based on one or more transformations of an object in a sequence of frames.
Dependencies are inferred from the claim text.