This repo is for the paper: Theory-inspired Parameter Control Benchmarks for Dynamic Algorithm Configuration
The DAC environment is based on the Dynamic Algorithm Configuration benchmark library DACBench.
For computing the optimal DAC policy for RLS (Randomized Local Search) on LeadingOne problem, please check the instructions in rls_lo/optimal_policy/
.
For training/evaluating a DDQN agent on this benchmark, please see the instructions in rls_lo/rl
.