Exponentially Weighted Imitation Learning for Batched Historical Data

Authors: Qing Wang, Jiechao Xiong, Lei Han, peng sun, Han Liu, Tong Zhang

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Under mild conditions, our algorithm, though surprisingly simple, has a policy improvement bound and outperforms most competing methods empirically. Thorough numerical results are also provided to demonstrate the efficacy of the proposed methodology.
Researcher Affiliation Collaboration 1Tencent AI Lab 2Northwestern University {drwang, jcxiong, lxhan, pythonsun}@tencent.com hanliu@northwestern.edu, tongzhang@tongzhang-ml.org
Pseudocode Yes Algorithm 1 Monotonic Advantage Re-Weighted Imitation Learning (MARWIL)
Open Source Code No The paper does not provide any concrete access information (e.g., repository link, explicit statement of code release) for the source code of the described methodology.
Open Datasets No The paper mentions using environments like HFO and TORCS, and collecting 'human replay files' for King of Glory, but it does not provide concrete access information (links, citations with author/year, or specific repository names) to any publicly available or open datasets used for training.
Dataset Splits No The paper does not provide specific dataset split information (percentages, sample counts, or citations to predefined splits) for training, validation, and testing.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. It only mentions 'A DNN based function approximator is adopted' for King of Glory.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x) that are needed to replicate the experiment.
Experiment Setup Yes For the HFO game, we model the 3 discrete actions with multinomial probabilities and the 2 continuous parameters for each action with normal distributions of known σ = 0.2 but unknown µ... Then we optimize the policy loss Lp and the value loss Lv simultaneously, with a mixture coefficient cv as a hyper-parameter (by default cv = 1). Each experiment is repeated 3 times and the average of scores is reported in Figure 1. Additional details of the algorithm settings are given in Appendix B.2.