A K-State Framework for Combinatorial Optimization using P-Bits; The Rise, Fall, and Return of Electronic Analog Computing 2 views

The stalling of Moore’s law and rising demand for energy-intensive computation have led hardware designers to reconsider established computer architectures. Ising-model-based computing has shown promise in approximating combinatorial optimization problems, often implemented via p-bits, a form of binary stochastic neuron. However, current frameworks are limited to binary problems, whereas many real-world problems require more than two states to express. This in turn demands costly reductions that decrease convergence rate and solution quality. In response, we extend the current model to accept an arbitrary number of states, and we compute the max 3,4, and 5-cuts of standard benchmarking graphs via simulation. When compared to state-of-the-art k-cut solvers, this model discovers many new best-known solutions. Alternatively, other designers are considering reimplementation of analog techniques. Once the dominant form of computation, analog fully gave way to digital machines during the 1970’s and has largely been perceived as obsolete since. However, historians have established inconsistency between actual and perceived functional obsolescence, which further applies to analog computing. Evidence indicates that the perceived obsolescence of analog computing had more to do with industrial misalignment and digital snowballing effects than technological obsolescence.

BS (Bachelor of Science)

Combinatorial Optimization; Analog Computing; Ising Model; P-bit; Probabilistic Computing

School of Engineering and Applied Science Bachelor of Science in Computer Engineering Technical Advisor: Nikhil Shukla STS Advisor: Peter Norton

English

All rights reserved by the author (no additional license for public reuse)

2025-12-15