Permutation-Invariant Neural Networks for Reinforcement Learningai.googleblog.com3 pointshardmaru5 years ago