GA3C: GPU-based A3C for Deep Reinforcement Learning

75 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures (2018, methodology): “The actor/learner asynchrony in GA3C leads to instabilities during learning, which (Babaeizadeh et al., 2016) only partially mitigates by adding a small constant to action probabilities Environment steps Forward pass Backward pass”
Accelerated Methods for Deep Reinforcement Learning (2018, methodology): “In GA3C (Babaeizadeh et al., 2016), a speedup over CPU-only A3C was achieved by using a GPU.”
Efficient Parallel Methods for Deep Reinforcement Learning (2017, methodology): “The results and their comparison to Gorila (Nair et al., 2015), A3C (Mnih et al., 2016) and GA3C (Babaeizadeh et al., 2016) are presented in Table 1.”
Accelerating Reinforcement Learning through GPU Atari Emulation (2019, methodology): “Ape-X DQN [8] 5 days 22,800M 50K 376 cores, 1 GPU Rainbow [7] 10 days 200M — 1 GPU Distributional (C51) [4] 10 days 200M — 1 GPU A3C [12] 4 days — 2K 16 cores GA3C [1, 2] 1 day — 8K 16 cores, 1 GPU Prioritized Dueling [22] 9.”
Distributed Deep Reinforcement Learning: Learn how to play Atari games in 21 minutes (2018, methodology): “) shows that our implementation is much faster and achieves as good or better scores8 (see table 3). Importantly our experiments used 64 CPU nodes of 12 cores each, while the experiments presented in [5] were all single node. However the results show that using distributed computations on CPU cluste”