Biologically inspired architectures for sample-efficient deep reinforcement learning
arXiv:1911.11285
Abstract
Deep reinforcement learning requires a heavy price in terms of sample efficiency and overparameterization in the neural networks used for function approximation. In this work, we use tensor factorization in order to learn more compact representation for reinforcement learning policies. We show empirically that in the low-data regime, it is possible to learn online policies with 2 to 10 times less total coefficients, with little to no loss of performance. We also leverage progress in second order optimization, and use the theory of wavelet scattering to further reduce the number of learned coefficients, by foregoing learning the topmost convolutional layer filters altogether. We evaluate our results on the Atari suite against recent baseline algorithms that represent the state-of-the-art in data efficiency, and get comparable results with an order of magnitude gain in weight parsimony.
Deep Reinforcement Learning Workshop, NeurIPS 2019, Vancouver, Canada
References in corpus (16)
- Playing Atari with Deep Reinforcement Learning
- Prioritized Experience Replay
- Understanding deep learning requires rethinking generalization
- Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
- Optimizing Neural Networks with Kronecker-factored Approximate Curvature
- Model-Based Reinforcement Learning for Atari
- TensorLy: Tensor Learning in Python
- Model-Free Episodic Control
- Classification and Geometry of General Perceptual Manifolds
- When to use parametric models in reinforcement learning?
- Neural Episodic Control
- Solid Harmonic Wavelet Scattering for Predictions of Molecule Properties
- Scattering Networks for Hybrid Representation Learning
- Linear Readout of Object Manifolds
- Tensor Regression Networks with various Low-Rank Tensor Approximations
- Random Expert Distillation: Imitation Learning via Expert Policy Support Estimation