Answers are grounded in this paper and cite their sources.
This work presents a resistively coupled 2 × 2 all-to-all oscillatory neural network (ONN) based on ring oscillators (ROSCs) that implements a hardware oscillator-based Potts machine (OPM) for grayscale associative memory and graph coloring. Fabricated in 65-nm CMOS technology, the ONN supports multi-state spin operation through nine-stage ROSCs with accessible delay elements and a five-bit tunable resistor bank enabling programmable coupling. Leveraging the Potts model, the OPM reliably retrieves horizontal, vertical, and diagonal (inverse) patterns encoded in 5 gray tones and efficiently solves multiple graph coloring problems using phase-encoded multi-state spins. Operating at 1.91 MHz with a power consumption of 12.6 μ W , the network consistently reaches steady-state solutions within three oscillation cycles for associative memory tasks and within 37 cycles for one iteration of the planted graph coloring task. Although compact in scale, this is, to the best of our knowledge, the first OPM chip to demonstrate both associative memory and combinatorial optimization with up to nine distinguishable phase states.
No takes yet. Share an insight, caveat, or question.
Bodegraven et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: