Two Algorithms Race on Breakout

Two algorithms, A and B, 10 random seeds, 10M steps on Atari Breakout, which one is the best? (wrong answers recommended, A and B reveal will come later) Breakout A B Score Timesteps (1e6) 0 100 200 300 400 500 0 2 4 6 8 10

A chart compares two algorithms, A and B, trained with random seeds for 10 million steps on Atari Breakout. Algorithm A rises quickly and appears stronger early, while algorithm B improves later, making the final comparison ambiguous and inviting humorous wrong answers before the results are revealed. EN: A chart compares two algorithms, A and B, trained with random seeds for 10 million steps on Atari Breakout. Algorithm A rises quickly and appears stronger early, while algorithm B improves later, making the final comparison ambiguous and inviting humorous wrong answers before the results are revealed.

• @ithueti / 3198 2020-11-19