Table 1

Key hyperparameters for DRL algorithms

AlgorithmLearning raten_steps/batchBuffer sizeNotes
A2C1 × 10–410/–GAE 0.95
PPO5 × 10–41,024/64entropy coef 0.01
SAC3 × 10–4–/256106ent_coef auto 0.2, log std −3
DDPG3 × 10–4–/256106noise σ = 0.1
TD31 × 10–3–/256106noise σ = 0.1
TQC3 × 10–4–/512106ent_coef auto 0.1, quantile trunc. 25%
RecurrentPPO3 × 10–41,024/128LSTM size 128, 1 layer

Note(s): GAE: generalized advantage estimation

or Create an Account

Close Modal
Close Modal