English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Reinforcement learning (RL) has become a proven method for optimizing aprocedure for which success has been defined, but the specific actions neededto achieve it have not. We apply the so-called "black box" method of RL to whathas been referred as the "black art" of simulated annealing (SA), demonstratingthat an RL agent based on proximal policy optimization can, through experiencealone, arrive at a temperature schedule that surpasses the performance ofstandard heuristic temperature schedules for two classes of Hamiltonians. Whenthe system is initialized at a cool temperature, the RL agent learns to heatthe system to "melt" it, and then slowly cool it in an effort to anneal to theground state; if the system is initialized at a high temperature, the algorithmimmediately cools the system. We investigate the performance of our RL-drivenSA agent in generalizing to all Hamiltonians of a specific class; when trainedon random Hamiltonians of nearest-neighbour spin glasses, the RL agent is ableto control the SA process for other Hamiltonians, reaching the ground statewith a higher probability than a simple linear annealing schedule. Furthermore,the scaling performance (with respect to system size) of the RL approach is farmore favourable, achieving a performance improvement of one order of magnitudeon L=14x14 systems. We demonstrate the robustness of the RL approach when thesystem operates in a "destructive observation" mode, an allusion to a quantumsystem where measurements destroy the state of the system. The success of theRL agent could have far-reaching impact, from classical optimization, toquantum annealing, to the simulation of physical systems.

Controlled Online Optimization Learning (COOL): Finding the ground state of spin Hamiltonians with reinforcement learning