Exploration for Countering the Episodic Memory

Table 2

Mean human-normalized scores for listed algorithms at different training frames.

Frames (M)	Nature DQN (%)	Retrace(λ) (%)	A3C (%)	MFEC (%)	CounterEM (MFEC) (%)	NEC (%)	CounterEM (NEC) (%)
1	−10.5	−10.5	5.2	28.4	40.8	45.6	61.7
2	−5.8	−5.4	8.0	39.4	68.3	58.3	84.9
4	8.8	6.2	11.8	53.4	88.4	73.3	97.3
10	51.3	52.7	22.3	85.0	100.1	99.8	115.4
20	94.5	237.7	59.7	113.6	117.9	121.5	122.8
40	151.2	386.5	255.4	142.2	148.1	144.8	150.0