Summary of Trajectory-wise Iterative Reinforcement Learning Framework For Auto-bidding, by Haoming Li et al.
Trajectory-wise Iterative Reinforcement Learning Framework for Auto-bidding
by Haoming Li, Yusen Huo, Shuai Dou, Zhenzhe Zheng, Zhilin Zhang, Chuan Yu, Jian Xu, Fan Wu
First submitted to arxiv on: 23 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Information Retrieval (cs.IR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper addresses the issue of performance degradation in auto-bidding algorithms used in online advertising. Current RL-based policies are typically trained in simulation, leading to a gap between simulated and actual performance. To bridge this gap, the authors propose an iterative offline RL framework that involves deploying multiple agents in parallel to collect interaction data, training offline RL algorithms on this data, and then deploying the trained policy for further data collection. However, they identify a bottleneck in this framework caused by ineffective exploration and exploitation due to the conservatism of offline RL algorithms. To overcome this bottleneck, the authors propose Trajectory-wise Exploration and Exploitation (TEE), which introduces a novel data collecting and utilization method for iterative offline RL from a trajectory perspective. Additionally, they propose Safe Exploration by Adaptive Action Selection (SEAS) to ensure safety while preserving dataset quality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about how online advertising can work better. Right now, advertisers use special computer programs to buy ad space on websites and apps. These programs are like virtual assistants that try to find the best deals for the advertiser. The problem is that these programs are only tested in fake scenarios before being used in real life, so they don’t always perform as well as expected. To fix this, the authors suggest a new way of training these programs by letting them collect data from multiple sources and then fine-tuning their performance using special algorithms. They also propose two new methods to make sure that this process is safe and effective. |
Keywords
* Artificial intelligence * Fine tuning
