A novel technology for fabricating customizable VLSI artificial neural network chips

The authors describe an implementation of hardware neural networks using highly linear thin-film resistor technology and an 8-b binary weight circuit to produce customizable artificial neural network chips and systems. These neural networks are programmed using precision laser cutting deposition. The fast turnaround of laser-based customization allows the exploration of different neural network architectures and rapid programming of the synaptic weights. The customizable chip allows expansion of an artificial network laterally and vertically. This flexibility permits the building of very large neural network systems