Data-driven equation discovery reveals nonlinear reinforcement learning in humans - PMC
Our article offers an answer to a foundational question in psychology and neuroscience: how do people learn from rewards and punishments? Specifically, we introduce a computational model of human reinforcement learning (RL) that points to a ...
Significance Our article offers an answer to a foundational question in psychology and neuroscience: how do people learn from rewards and punishments? Specifically, we introduce a computational model of human reinforcement learning (RL) that points to a nonlinear updating of the probability of reward. The strength of our model lies also in the process through which it was developed. Specifically, we discovered our model in a bottom–up fashion using symbolic regression—a class of machine learning tools applied primarily in physics and engineering. We believe that, in addition to the…
saved by
related reading
- Q-learning is not yet scalableseohong.me
- Reinforcement Learning in Newcomblike Problemsproceedings.neurips.cc
- Deep Reinforcement Learning from Human Preferencesproceedings.neurips.cc
- [1710.10044] Distributional Reinforcement Learning with Quantile Regressionarxiv.org
- Scaling Laws for Reward Model Overoptimizationarxiv.org
- [2208.10687] The Effect of Modeling Human Rationality Level on Learning Rewards from Multiple Feedback Typesar5iv.labs.arxiv.org
- Models Don't "Get Reward" — LessWronglesswrong.com
- [1709.06560] Deep Reinforcement Learning that Mattersarxiv.org
- RL in Cognitionsubstack.com
- [2607.18966] Measuring Reward-Seeking via Contrastive Belief Updatesarxiv.org
- Reinforcement Learning: Machine Learning Meets Control Theoryyoutube.com
- [2602.11399] Can We Really Learn One Representation to Optimize All Rewards?arxiv.org