Neural dynamical systems on ferroelectric compute-in-memory for real-time forecasting
2026-06-15 • Emerging Technologies
Emerging TechnologiesHardware ArchitectureNeural and Evolutionary Computing
AI summaryⓘ
The authors present FerroNDS, a new type of brain-inspired computer that uses analog components to mimic continuous-time neural dynamics more naturally than typical digital systems. By using devices called ferrodiodes for memory and computation, their system efficiently processes time-based signals like sound waves at low power and high speed. They tested a 128-neuron model that can analyze and predict complex signal patterns with much less energy and space than regular digital approaches. This work shows how special materials and circuitry can enable better brain-like computing in hardware.
Neural dynamical systemsAnalog neuromorphic computingFerrodiodesCompute-in-memoryShort-time Fourier transformFrequency-selective filteringReal-time operationContinuous-time dynamicsPer-neuron energyLatency
Authors
Keshava Katti, Adithya Selvakumar, Pratik Chaudhari, Deep Jariwala
Abstract
Neural dynamical systems are expressive temporal predictors that capture continuous-time dynamics through fine-grained state updates. However, this sequential structure maps poorly onto digital hardware optimized for dense matrix operations, a mismatch that analog neuromorphic computing, with its native continuous-time dynamics, can resolve. We introduce FerroNDS, a neuromorphic system built from two analog primitives: an integrator for temporal accumulation and an oscillator for frequency-selective filtering. We map this system onto compute-in-memory hardware based on multi-bit ferrodiodes. A 128-neuron instance of FerroNDS computes short-time Fourier transform and forecasts a 500-ms horizon for periodic, quasi-periodic, and chaotic signals. The system achieves sub-watt real-time operation with per-neuron per-inference energy of 1.64 $μ$J (200 Hz) and 0.29 $μ$J (10 kHz), 25-40$\times$ area reduction over SRAM-based digital systems, and per-layer latency of 3.18 ms (200 Hz) and 63.87 $μ$s (10 kHz). To our knowledge, this is the first end-to-end integration of a ferrodiode into a neuromorphic computational framework, establishing ferroelectric compute-in-memory as a practical substrate for analog neural dynamical systems.