Inspired by Adrian Colyer, Denny Britz and Daniel Seita
This contains my notes for research papers that are relevant for my PhD on Machine Learning. First, I list papers that I've read and papers that I want to read. Then, read papers are numbered on a (1) to (5) scale where a (1) means I have only barely skimmed it, while a (5) means I feel confident that I understand almost everything about the paper. The links here go to my paper summaries if I have them, otherwise those papers are on my TODO list.
Contents:
- Learning to Act by Predicting the Future (5)
- Universal Value Function Approximators (4)
- Horde: A Scalable Real-time Architecture for Learning Knowledge from Unsupervised Sensorimotor Interaction (5)
- Learning by Playing – Solving Sparse Reward Tasks from Scratch (3)
- Reinforcement Learning with Unsupervised Auxiliary Tasks (4)
- Successor Features for Transfer in Reinforcement Learning (4)
- Hindsight Experience Replay (4)
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks (4)
- Meta-Learning and Universality: Deep Representations and Gradient Descent can Approximate any Learning Algorithm (2)
- A Distributional Perspective on Reinforcement Learning (4)
- Transfer for Reinforcement Learning Domains: A Survey (3)
- World Models (4)
- Simple random search provides a competitive approach to reinforcement learning (5)
- Levine's lecture on "what do to when you have a forward model ?" (5)
- Trust Region Policy Optimization (4)
- Proximal Policy Optimization (4)
- Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images (3)
- Imagination-Augmented Agents for Deep Reinforcement Learning (3)
- Deep Successor Reinforcement Learning (4)
- Learning to Navigate in Complex Environments (4)
- Universal Option Models (3)
- Options: Temporal abstraction in Reinforcement Learning (4)
- A Laplacian Framework for Option Discovery in Reinforcement Learning (3)
- Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation (2)
- Temporal Difference Models: Model-Free Deep RL for Model-Based Control (3)
- The Predictron: End-To-End Learning and Planning (2)
- Action-Conditional Video Prediction using Deep Networks in Atari Games (3)
- Decoupling Dynamics and Reward for Transfer Learning (5)
- Diversity is All You Need: Learning Skills without a Reward Function (2)
- Generative Temporal Models with Spatial Memory for Partially Observed Environments (?)
- Learning and Querying Fast Generative Models for Reinforcement Learning (2)
- Model-Based Value Expansion for Efficient Model-Free Reinforcement Learning (4)
- A Deep Hierarchical Approach to Lifelong Learning in Minecraft (4)
- Deep Reinforcement Learning that Matters (5)
- Continuous Deep Q-Learning with Model-based Acceleration (3)
- Overcoming catastrophic forgetting in neural networks (4)
- Progressive Neural Networks (4)
- Learning without Forgetting (3)
- Distilling the Knowledge in a Neural Network (3)
- Policy Distillation (4)
- Progress & Compress: a scalable framework for continual learning (4)
- Progressive Reinforcement Learning with Distillation for Multi-Skilled Motion Control (3)
- Residual Loss Prediction: Reinforcement Learning With No Incremental Feedback (3)
- Distral: Robust Multitask Reinforcement Learning (?)
- Mixture Density Networks (5)
- DARLA: Improving Zero-Shot Transfer in Reinforcement Learning (4)
- Divide-and-Conquer Reinforcement Learning (?)
- Mix & Match – Agent Curricula for Reinforcement Learning (?)
- Generative Temporal Models with Spatial Memory for Partially Observed Environments (?)
- Been There, Done That: Meta-Learning with Episodic Recall (?)
- Continual Reinforcement Learning with Complex Synapses (?)
- Importance Weighted Transfer of Samples in Reinforcement Learning (?)
- Self-Imitation Learning (?)
- Efficient Model-Based Deep Reinforcement Learning with Variational State Tabulation (?)
- A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning (?)
- Hybrid computing using a neural network with dynamic external memory (?)
- Transfer in variable-reward hierarchical reinforcement learning (?)
- Learning Actionable Representations with Goal-Conditioned Policies (?)
- Disentangling Controllable and Uncontrollable Factors by Interacting with the World (?)
- The Laplacian in RL: Learning Representations with Efficient Approximations (?)
- Associative Compression Networks (?)
- Online Structured Laplace Approximations for Overcoming Catastrophic Forgetting (?)
- Learning Latent Subspaces with Variational Autoencoders (?)
- APIAE: Representation Learning and Planning for Dynamical Systems (?)
- Isolating Sources of Disentanglement in Variational Autoencoders (?)
- Learning Latent Dynamics for Planning from Pixels (?)
- Learning by watching Youtube (?)
- Policy Optimization via Importance Sampling (?)
- Learning Attractor Dynamics for Generative Memory (?)
- End-to-end Differentiable Physics for Learning and Control (?)
- HIRO: Data-Efficient Hierarchical Reinforcement Learning (?)
- Evolved Policy Gradients (?)
- Variational Memory Autoencoder (?)
- Emergence of Invariance and Disentanglement in Deep Representations (?)
- Adaptative Path-Integral Autoencoder: Representation Learning and Planning for Dynamical Systems (?)
- Discrete Variational Autoencoders (?)
- Towards a Definition of Disentangled Representations (5, Reading group presentation)
- Learning Latent Dynamics for Planning from Pixels (?)
- Model Based Reinforcement Learning for Atari (?)
- How do Mixture Density RNNs Predict the Future? (?
- CURIOUS: Intrinsically Motivated Multi-Task Multi-Goal Reinforcement Learning (4)
- Universal Successor Features Approximators (2)
- A new dog learns an old trick: RL learns classical algorithms (?)
- Near-Optimal Representation Learning for Hierarchical RL (?)
- Recurrent experience replay in Distributed RL (3)
- Large Memory Layers with Product Keys (4)
- VQ-VAE: Neural Discrete Representation Learning (4)
- StyleGAN: A Style-Based Generator Architecture for Generative Adversarial Networks (5)
- BigGAN: Large Scale GAN Training for High Fidelity Natural Image Synthesis (5)
- (?)
- (?)
- (?)
- (?)
Special: Notes and thoughts about VAEs and how to make them work! Based on the following papers:
- β-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework (β-VAE) (4)
- Understanding disentangling in β-VAE (CCI-VAE) (4)
- Disentangling by Factorising (FactorVAE) (4)
- Isolating Sources of Disentanglement in VAEs (β-TCVAE (3)
- Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations (4)
- Fixing a Broken ELBO (Fixed-rate objective function for VAEs) (4)
- Taming VAEs (GECO VAEs) (3)
- FeUdal Networks for Hierarchical Reinforcement Learning
- Diversity is All You Need: Learning Skills without a Reward Function
- Learning to Search Better than Your Teacher
- Transfer in Variable-Reward Hierarchical Reinforcement Learning
- Curriculum Learning
- Theoretical TL papers from TL survey
I need the answers to these questions. Any help is welcome ! Reach me at [email protected]