Using Contextual Bandit models in large action spaces at Instacart

Contextual bandit models are a popular approach for personalizing the user experience by recommending relevant products. However, these models become challenging to train and evaluate when the number of actions, or products, in the recommendation pool become large. In this blog post, we will explore the difficulties associated with using contextual bandit models in large action spaces and propose potential solutions to overcome these challenges. One of these solutions was recently launched into production at Instacart.

Link