Model Selection for Production System via Automated Online Experiments

Abstract

A challenge that machine learning practitioners in the industry face is the task of selecting the best model to deploy in production. As a model is often an intermediate component of a production system, online controlled experiments such as A/B tests yield the most reliable estimation of the effectiveness of the whole system, but can only compare two or a few models due to budget constraints. We propose an automated online experimentation mechanism that can efficiently perform model selection from a large pool of models with a small number of online experiments. We derive the probability distribution of the metric of interest that contains the model uncertainty from our Bayesian surrogate model trained using historical logs. Our method efficiently identifies the best model by sequentially selecting and deploying a list of models from the candidate set that balance exploration-exploitation. Using simulations based on real data, we demonstrate the effectiveness of our method on two different tasks.

Related

October 2020 | CIKM

Query Understanding for Surfacing Under-served Music Content

Federico Tomasi, Rishabh Mehrotra, Aasish Pappu, Judith Bütepage, Brian Brost, Hugo Galvão, Mounia Lalmas

September 2020 | RecSys

Contextual and Sequential User Embeddings for Large-Scale Music Recommendation

Casper Hansen, Christian Hansen, Lucas Maystre, Rishabh Mehrotra, Brian Brost, Federico Tomasi, Mounia Lalmas

September 2020 | RecSys

Inferring the Causal Impact of New Track Releases on Music Recommendation Platforms through Counterfactual Predictions

Rishabh Mehrotra, Prasanta Bhattacharya, Mounia Lalmas