Quantum Rényi-Jarzynski Equality

What the math gives to ML

The paper derives a non-destructive quantum Jarzynski relation in which nonequilibrium bath drift is quantified by a tunable Renyi divergence rather than only by an ordinary free-energy difference. Its transferable asset is a principled family of divergence objectives whose order k changes sensitivity to rare, high-ratio regions, with a sharp transition between competing minima at a critical k. A practical neural-network analogue is to regularize fine-tuning or control-policy training with a Renyi divergence to a reference predictive distribution and explicitly track the order at which the optimizer switches between solutions.