Identifying mirror symmetry density with delay in spiking neural networks
The ability to rapidly identify symmetry and anti-symmetry is an essential attribute of intelligence. Symmetry perception is a central process in human vision and may be key to human 3D visualization. While previous work in understanding neuron symmetry perception has concentrated on the neuron as an integrator, the invention provides the coincidence detecting property of the spiking neuron can be used to reveal symmetry density in spatial data. A synchronized symmetry-identifying spiking artificial neural network enables layering and feedback in the network. The network of the invention can identify symmetry density between sets of data and present a digital logic implementation demonstrating an 8×8 leaky-integrate-and-fire symmetry detector in a field-programmable gate array. The efficiency of spiking neural networks can be harnessed to rapidly identify symmetry in spatial data with applications in image processing, 3D computer vision, and robotics.
1. A symmetry detector to detect symmetry of input data, said symmetry detector comprising:
a plurality of input nodes, each input node receiving a discrete data point of the input data, each input node representing a data point at a fixed position in an N dimensional metric space, and having a spike producer configured to produce spiked data as a threshold of the input data and delay devices configured to delay the produced spiked data by a fixed time constant representing distance in the metric space and provide delayed output data; and
a plurality of output nodes, each output node connected to one or more of the plurality of input nodes, each output node representing output data at a fixed position also in the same N dimensional metric space as the input nodes, receiving the delayed output data, from each of the connected one or more plurality of input nodes, where each delay is constant and set proportionally to the distance in the metric space between the input node and the output node, and each output node having a coincidence detector configured to detect symmetry based on delayed output data that is spiked.
2. The symmetry detector of claim 1 , said plurality of input nodes and said plurality of output nodes each comprising a neuron.
3. The symmetry detector of claim 1 , said coincidence detector comprising a plurality of multipliers each configured to multiply the delayed output data from each of the connected one or more said plurality of input nodes by a corresponding weight value, an accumulator configured to receive and accumulate the multiplied delayed output data and a threshold detector configured to detect a threshold number of spiked data in the accumulated multiplied delayed output data.
4. The symmetry detector of claim 1 , wherein the threshold number represents symmetry of input data.
5. The symmetry detector of claim 1 , said spike producer comprising a multiplier configured to multiply the input data by a corresponding weight value, an accumulator configured to receive the multiplied input, a threshold detector configured to detect spiked data in the input data and provide the detected spiked input data.
6. The symmetry detector of claim 5 , said accumulator having leak configured to adjust the sensitivity of the output node to detecting symmetry.
7. The symmetry detector of claim 1 , said delay device comprising a set of clocks, or a delay counter, or a routing delay.
8. The symmetry detector of claim 1 , wherein the delay is proportional to a distance between each respective input node and each respective output node, in the data space.
9. A symmetry detector to detect symmetry of input data, said symmetry detector comprising:
a plurality of input nodes, each input node receiving a discrete data point of the input data, each input node representing a data point at a fixed position in an N dimensional metric space, and having a spike producer configured to produce spiked data by a fixed time constant representing distance in the metric space as a threshold of the input data; and
a plurality of output nodes, each output node connected to one or more of the plurality of input nodes, each output node representing output data at a fixed position also in the same N dimensional metric space as the input nodes, receiving the input node output data from each of the one or more connected plurality of input nodes, each output node having a delay device delaying the input node output data to provide a delayed input node output data, where each delay is constant and set proportionally to the distance in the metric space between the input node and the output node, and each output node having a coincidence detector configured to detect a threshold number of spikes in the delayed input node output data in a given period of time.
10. The symmetry detector of claim 9 , said plurality of input nodes and said plurality of output nodes each comprising a neuron.
11. The symmetry detector of claim 9 , said coincidence detector comprising an accumulator configured to receive and accumulate the delayed input node output data and a threshold detector coupled to the accumulator configured to identify the threshold number of spikes.
12. The symmetry detector of claim 11 , said accumulator having leak configured to adjust the sensitivity of the output node to detecting symmetry.
13. The symmetry detector of claim 9 , wherein the threshold number represents symmetry of input data.
14. The symmetry detector of claim 9 , said delay device comprising a set of clocks, or a delay counter, or a routing delay.
15. The symmetry detector of claim 1 , the spiked data comprising a temporal pulse.