Analog Neuromorphic Circuits With Resistive-Memory Parallel Multiply-Add

The computing system is segmented into multiple independent processing units (neurons) that can operate simultaneously. Each neuron processes operations in parallel rather than sequentially, dividing the computational workload into concurrent tasks that execute independently, thereby resolving the chronological execution limitation of conventional microprocessors.

The system transitions from one-dimensional sequential processing to multi-dimensional parallel processing. By implementing spatial architecture where multiple neurons are arranged in parallel, the system executes operations across multiple dimensions simultaneously, enabling exponential increase in computational efficiency without proportionally increasing time loss.

The patent replaces conventional mechanical/electronic computing architectures with a neuromorphic system that uses biological-inspired principles. By substituting traditional processor-based computation with membrane potential-based neural networks, the system achieves high computational efficiency while reducing physical footprint and power consumption through distributed parallel processing.

Each neuron in the network performs multiple functions simultaneously - processing inputs, applying weights, generating outputs, and participating in parallel computations. This multi-functionality allows the system to achieve high computational efficiency without requiring separate dedicated components for each function, thereby reducing overall physical size.

The neural network operates through periodic action potentials and membrane potential changes rather than continuous high-power processing. By using event-driven computation where neurons fire only when necessary to process inputs, the system maintains high computational efficiency while significantly reducing average power consumption compared to conventional continuous-processing architectures.

The neuromorphic system performs self-regulation through membrane potential dynamics and threshold-based firing mechanisms. Each neuron automatically adjusts its state based on incoming signals without requiring external control, enabling efficient energy usage where computation occurs only when needed, thereby reducing overall power consumption while maintaining high productivity.