Statistical Reinforcement Learning for High-Risk Decision Problems

In recent years, reinforcement learning (RL) has received significant attention due to high-profile successes in areas such as chess, go, and protein folding. Furthermore, RL has long been a cornerstone in e-commerce where it is used to personalize customer recommendations.

In this Statistically Speaking, Dr. Eric Laber explores the lesser-studied use of RL for high-stakes settings in which data volume is low and the cost of experimentation is high. In various applications – such as precision medicine – statistical efficiency and safety are paramount.

Learn:

• The importance of understanding basic concepts of reinforcement learning.
• Why statistical thinking is critical for application of RL in high-stakes problems.
• How RL can facilitate the personalization of customer recommendations.
• Why exploration of RL is relevant when data volume is low and cost is high.