Artificial Neuron Synaptic Weights Implemented with Variable Dissolvable Conductive Paths
Description
Artificial Neural Network (ANN) is an information processing system that is based on the biological neural networks of the human brain. In this computational model, the network contains nodes or units, replicating the role of neurons. The nodes receive input from other nodes or external sources, computing an output. The connectivity between the nodes determines the functionality of the neural network. Like the biological neural system, synaptic weight refers to the strength/amplitude of the connection between two nodes. Each input has an associated weight that can be modified, modeling synaptic learning. ANNs are designed by variable and dissolvable conductive paths. The technology implements a novel method that gradually increases the current circulating in the resistive random access memory (ReRAM) cells with a variable current source. As quickly as the current returns to zero, the resistance of the cells returns to its initial value with no trace of prior states. The formation of conductive paths into the cells of ReRAM devices can be exploited to design low-power, controllable, and reconfigurable weights for use in ANN. Different stable resistance values in each ReRAM cell can be controlled by injecting different current values.
Additional information
Patent number and inventor
Patent pending
Bertrand Cambou, Donald Telesca Jr., and Brayden Villa.
Potential applications
This technology is designed for use with artificial intelligence.
Benefits and advantages
This technology demonstrates better control of an artificial neuron with circuitry that includes resistive RAMs and memristors.
Case number and licensing status
2018-002
This innovation is available for licensing.