search

yoon-gu / ezlab-rl

1 stars 2 forks source link

11-experiment-sir-in-dqn-large-scale #12

Closed boyeon-kim closed 9 months ago

boyeon-kim commented 9 months ago

인구 1000만 S0=9999990 I0=10 time = 300 beta = 0.00000007 gamma = 1/10

boyeon-kim commented 9 months ago

reward = - I

SIR_wo_control SIR_w_control SIR_control_u SIR_score

boyeon-kim commented 9 months ago

reward = -I - action * S/1e6 SIR_wo_control SIR_w_control SIR_control_u SIR_score

boyeon-kim commented 9 months ago
image image image image image
boyeon-kim commented 9 months ago

Learning rate와의 상관관계 Normalize (reward scaling?) 의 상관관계