{"id":"arxiv/2602.11206","url":"https://arxiv.org/abs/2602.11206","tags":["cs.LG","cs.AI","cs.CV","math.RA","q-bio.NC"],"score":0,"title":"UltraLIF: Fully Differentiable Spiking Neural Networks via Ultradiscretization and Max-Plus Algebra","domain":"cs.LG","source":"arxiv","authors":["Jose Marie Antonio Miñoza"],"reviews":0,"subject":"cs.LG","summary":"Spiking Neural Networks (SNNs) offer energy-efficient, biologically plausible computation but suffer from non-differentiable spike generation, necessitating reliance on heuristic surrogate gradients. This paper introduces UltraLIF, a principled framework that replaces surrogate gradients with ultradiscretization, a mathematical formalism from tropical geometry providing continuous relaxations of discrete dynamics. The central insight is that the max-plus semiring underlying ultradiscretization naturally models neural threshold dynamics: the log-sum-exp function serves as a differentiable…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-02-10","reactions":{}}