Spring 2018 CS595I Advanced NLP/ML Seminar

Time: To be announced Location: HFH 1132.

If you registered this class, you should contact the instructor to present one paper *and* be the discussant of one paper below.

• Presenter: prepare a short summary of no more than 15 mins of presentation.
• Discussant: by presenting a paper in one session, you automatically become the discussant of the other paper. Please prepare two questions for discussion about the paper.

If you don't present or lead the discussion, you will then need to write a 2-page final report in ICML 2018 style, comparing any two of the papers below. Due: TBA to william@cs.ucsb.edu.

Machine Learning

• Variance-based Regularization with Convex Objectives. Hongseok Namkoong, John Duchi. https://arxiv.org/abs/1610.02581

Reinforcement Learning

• NEURAL MAP: STRUCTURED MEMORY FOR DEEP REINFORCEMENT LEARNING, Parisotto and Salakhutdinov, ICLR 2018 https://arxiv.org/pdf/1702.08360.pdf
• Counterfactual Multi−Agent Policy Gradients", Foerster et al., AAAI 2018, Outstanding Student Paper, http://www.cs.ox.ac.uk/people/shimon.whiteson/pubs/foersteraaai18.pdf
• Shallow Updates for Deep Reinforcement Learning, Levine et al., NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=9098
• Imagination-Augmented Agents for Deep Reinforcement Learning Racanière et al., NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=10081
• Robust Imitation of Diverse Behaviors, Wang et al. 2017, https://arxiv.org/pdf/1707.02747.pdf
• Compatible Reward Inverse Reinforcement Learning Metelli et al., NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=8993
• Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation, Wu et al., NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=10087
• Expected Policy Gradients, Kamil Ciosek, Shimon Whiteson, https://arxiv.org/abs/1706.05374
• Reinforcement Learning with Deep Energy-Based Policies Haarnoja et al, ICML 2017 http://proceedings.mlr.press/v70/haarnoja17a/haarnoja17a.pdf
• Interpolated Policy Gradient: Merging On-Policy and Off-Policy Gradient Estimation for Deep Reinforcement Learning, Gu et al., NIPS 2017 https://arxiv.org/abs/1706.00387
• Distral: Robust multitask reinforcement learning, Teh et al., NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=9227
• Repeated Inverse Reinforcement Learning, Amin et al., NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=10107
• Hybrid Reward Architecture for Reinforcement Learning, Van Seijen et al., NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=9314
• Cold-Start Reinforcement Learning with Softmax Policy Gradient, Ding and Soirut, NIPS 2017 https://nips.cc/Conferences/2017/Schedule?showEvent=9067