Paper Notes

Each heading is a date, displayed in YYMM format.

Each bulletpoint is a paper, with a link to the summary written in markdown. Each bulletpoint has an indented point with some "tags" for easy searching with ctrl+f. All tags use hyphens instead of spaces. Order of tags is irrelevant.

2002

• How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings (2019) - Ethayarajh
  • embeddings, bert, elmo, gpt-2

2001

• Discounted Reinforcement Learning is Not an Optimization Problem (2019) - Naik, Shariff, Yasui, Sutton
  • rl, optimization

1912

• Training Agents using Upside-Down Reinforcement Learning (2019) - Srivastava, Shyam, Mutz, Jaśkowski, Schmidhuber
  • rl, udrl, doom, a2c, dqn

1911

• Contextual Word Representations: A Contextual Introduction (2019) - Smith

  • nlp, embeddings, fine-tuning, transfer-learning

• Attention is All You Need (2017) - Vaswani, Shazeer, Parmar, Uszkoreit, Jones, Gomez, Kaiser, Polosukhin

  • transformer, nlp, nmt, translation, attention

• Universal Language Model Fine-tuning for Text Classification (2019) - Howard, Ruder

  • nlp, fine-tuning, lm, transfer-learning, embeddings

• The Measure of Intelligence (2019) - Chollet

  • rl, agi

• Deep Contextualized Word Representations (2019) - Peters, Neumann, Iyyer, Gardner, Clark, Lee, Zettlemoyer

  • lm, emlo, nlp, fine-tuning, transfer-learning, embeddings

1905

• Towards Interpretable Reinforcement Learning Using Attention Augmented Agents (2019) - Mott, Zoran, Chrzanowski, Wierstra, Rezende

  • rl, attention, atari, interpretable-models

• The Consciousness Prior (2017) - Bengio

  • thought, rl, vae, gan, unsupervised

• Reinforcement Learning, Fast and Slow (2019) - Botvinick, Ritter, Wang, Kurth-Nelson, Blundell, Hassabis

  • rl, replay-memory, biological

• Value Iteration Networks (2016) - Tamar, Wu, Thomas, Levine, Abbeel

  • rl, trpo, planning, maze

• Trust Region Policy Optimization (2015) - Schulman, Levine, Moritz, Jordan, Abbeel

  • rl, atari, trpo, policy-gradient, optimization

1904

• Prioritized Experience Replay (2016) - Schal, Quan, Antonoglou, Silver

  • rl, dqn, atari, replay-memory

• Human-level Control Through Deep Reinforcement Learning (2015) - Mnih et al.

  • rl, dqn, atari, replay-memory

• Gated Path Planning Networks (2018) - Lee, Parisotto, Chaplot, Xing, Salakhutdinov

  • rl, maze, doom, planning

• Active Neural Localization (2018) - Chaplot, Parisotto, Salakhutdinov

  • rl, maze, doom, a3c, localization

• Gated-Attention Architectures for Task-Oriented Language Grounding (2017) - Chaplot, Sathyendra, Pasumarthi, Rajagopal, Salakhutdinov

  • rl, doom, rl-text, a3c

• A Brief Survey of Deep Reinforcement Learning (2017) - Arulkumaran, Deisenroth, Brundage, Bharath

  • rl, introduction, survey

• Style-Analyzer: Fixing Code Style Inconsistencies with Interpretable Unsupervised Algorithms (2019) - Markovtsev, Long, Mougard, Slavnov, Bulychev

  • ml, decision-tree, ml4code

• Reinforcement Learning with Attention that Works (2019) - Manchin, Abbasnejad, va den Hengel

  • rl, atari, attention, ppo

• World Models (2018) - Ha, Schmidhuber

  • rl, vae, evolutionary-algorithms, world-models

1903

• To Tune or Not to Tune? Adapting Pretrained Representations to Diverse Tasks (2019) - Peters, Ruder, Smith

  • nlp, nli, sts, fine-tuning, transfer-learning, elmo, bert

• Playing Atari with Six Neurons (2019) - Cuccu, Togelius, Cudre-Mauroux

  • rl, atari, neuroevolution, evolutionary-algorithms

• Maybe Deep Neural Networks are the Best Choice for Modeling Source Code (2019) - Karamptatsis, Sutton

  • lm, ml4code