Summary of Reinforcement Learning with Quasi-hyperbolic Discounting, by S.r. Eshwar et al.
Reinforcement Learning with Quasi-Hyperbolic Discounting
by S.R. Eshwar, Mayank Motwani, Nibedita Roy, Gugan Thoppe
First submitted to arxiv on: 16 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 A new approach to reinforcement learning tackles the limitations of traditional exponential discounting and average reward setups by introducing Quasi-Hyperbolic (QH) discounting, which better captures human behavior’s bias towards immediate gratification. The optimal QH-policy can differ depending on the starting time, leading to sub-optimal overall returns if an agent is naive or impatient. To mitigate this issue, the paper proposes a model-free algorithm for finding a Markov Perfect Equilibrium (MPE) policy, which is shown to converge to an MPE using a two-timescale analysis. The algorithm is validated numerically for the standard inventory system with stochastic demands. This work significantly advances the practical application of reinforcement learning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way to learn through rewards tries to make it more like how humans behave. Right now, we use ways that are easy to calculate mathematically, but they don’t really match up with how people act. People tend to focus on what’s happening right away, rather than waiting for something better in the future. This new approach uses “quasi-hyperbolic” discounting, which takes this human behavior into account. It also introduces a way to find the best policy that works at all times, not just at the start or end. This makes it possible to make more informed decisions and get better results. |
Keywords
* Artificial intelligence * Reinforcement learning
