Machine learning at the mesoscale: a computation-dissipation bottleneck

arXiv:2307.02379v14 citationsh-index: 17
Originality Incremental advance
AI Analysis

This addresses a fundamental compromise in designing efficient physical computing devices, with potential implications for energy-efficient AI hardware.

The paper tackles the trade-off between computational performance and energy dissipation in mesoscale physical systems, showing that non-equilibrium conditions enhance performance in input-output tasks using real and synthetic datasets.

The cost of information processing in physical systems calls for a trade-off between performance and energetic expenditure. Here we formulate and study a computation-dissipation bottleneck in mesoscopic systems used as input-output devices. Using both real datasets and synthetic tasks, we show how non-equilibrium leads to enhanced performance. Our framework sheds light on a crucial compromise between information compression, input-output computation and dynamic irreversibility induced by non-reciprocal interactions.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes