notesum.ai
Published at November 6Beyond The Rainbow: High Performance Deep Reinforcement Learning On A Desktop PC
cs.AI
cs.LG
Released Date: November 6, 2024
Authors: Tyler Clark1, Mark Towers1, Christine Evers1, Jonathon Hare1
Aff.: 1University of Southampton

| Category | BTR | w/o Munchausen | w/o IQN | w/o SN | w/o Impala | w/o Maxpool |
|---|---|---|---|---|---|---|
| Action Gap | 0.276 | 0.060 | 0.154 | 0.280 | 0.403 | 0.282 |
| % Action Swaps | 33.9% | 47.0% | 45.2% | 42.2% | 30.9% | 39.7% |
| Policy Churn | 2.9% | 8.5% | 0.5% | 2.1% | 3.5% | 2.8% |
| Score ColorJitter | 196k | 90k | 97k | 167k | 5k | 149k |
| Score | 96k | 53k | 49k | 86k | 5k | 72k |
| Score | 211k | 88k | 89k | 173k | 5k | 138k |
| Score | 351k | 211k | 196k | 366k | 5k | 500k |