← 实验室

DQN 怎样学会选动作

CPU · Python 3.12.13 · 12.93 s · 2026-10-22 08:53:43 UTC+8

本次运行结果

{'noisy_max': 0.5708461176139561, 'separate_evaluation': 0.00913648208034665}
Toy target test passed: DQN=4, Double DQN=1.
DQN held-out returns: [143.3, 105.4]
Double DQN held-out returns: [159.4, 110.9]
Experiment output Experiment output
318
Optional exercise: fill answer and enable RUN_EXERCISE.
metrics.json
{
  "DQN": {
    "evaluation_returns": [
      143.3,
      105.4
    ],
    "initial_state_q": [
      52.85267181396485,
      48.256884765625
    ]
  },
  "Double DQN": {
    "evaluation_returns": [
      159.4,
      110.9
    ],
    "initial_state_q": [
      51.59617614746094,
      47.4152530670166
    ]
  }
}