Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pareto Front Approximation for Multi-Objective Session-Based Recommender Systems

Timo Wilm, Philipp Normann, Felix Stepprath

TL;DR

MultiTRON addresses the challenge of optimizing multiple business objectives in session-based recommendations by learning a parameterized Pareto front within a Transformer-based model. It uses Dirichlet-sampled preference vectors to condition the model and a regularization term to improve coverage, enabling a single model to approximate the entire Pareto front. Offline experiments on Diginetica, Yoochoose, and OTTO plus an online A/B test demonstrate competitive trade-offs between click-through rate and conversion rate, with higher lambda yielding broader Pareto coverage. The work offers a scalable, practical approach for multi-objective recommendation in e-commerce, and releases open-source code for broader adoption.

Abstract

This work introduces MultiTRON, an approach that adapts Pareto front approximation techniques to multi-objective session-based recommender systems using a transformer neural network. Our approach optimizes trade-offs between key metrics such as click-through and conversion rates by training on sampled preference vectors. A significant advantage is that after training, a single model can access the entire Pareto front, allowing it to be tailored to meet the specific requirements of different stakeholders by adjusting an additional input vector that weights the objectives. We validate the model's performance through extensive offline and online evaluation. For broader application and research, the source code is made available at https://github.com/otto-de/MultiTRON. The results confirm the model's ability to manage multiple recommendation objectives effectively, offering a flexible tool for diverse business needs.

Pareto Front Approximation for Multi-Objective Session-Based Recommender Systems

TL;DR

MultiTRON addresses the challenge of optimizing multiple business objectives in session-based recommendations by learning a parameterized Pareto front within a Transformer-based model. It uses Dirichlet-sampled preference vectors to condition the model and a regularization term to improve coverage, enabling a single model to approximate the entire Pareto front. Offline experiments on Diginetica, Yoochoose, and OTTO plus an online A/B test demonstrate competitive trade-offs between click-through rate and conversion rate, with higher lambda yielding broader Pareto coverage. The work offers a scalable, practical approach for multi-objective recommendation in e-commerce, and releases open-source code for broader adoption.

Abstract

This work introduces MultiTRON, an approach that adapts Pareto front approximation techniques to multi-objective session-based recommender systems using a transformer neural network. Our approach optimizes trade-offs between key metrics such as click-through and conversion rates by training on sampled preference vectors. A significant advantage is that after training, a single model can access the entire Pareto front, allowing it to be tailored to meet the specific requirements of different stakeholders by adjusting an additional input vector that weights the objectives. We validate the model's performance through extensive offline and online evaluation. For broader application and research, the source code is made available at https://github.com/otto-de/MultiTRON. The results confirm the model's ability to manage multiple recommendation objectives effectively, offering a flexible tool for diverse business needs.
Paper Structure (12 sections, 4 equations, 2 figures, 2 tables)

This paper contains 12 sections, 4 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Contributions
  4. Methods
  5. Recommender Model and Loss Functions
  6. Pareto Front Approximation
  7. Regularization and Pareto Front Coverage
  8. Overall Loss Function
  9. Experimental Setup
  10. Evaluation
  11. Conclusion
  12. Speaker Bio

Figures (2)

  • Figure 1: The best performing Pareto fronts from the offline evaluation on all three datasets showing the trade-off between $\mathcal{L}_c$ and $\mathcal{L}_o$ for 26 increasing values of $\pi_o$.
  • Figure 2: Evaluation results from the offline dataset (left) and live A/B test (right). It demonstrates that points on the predicted offline front translate into real-world trade-offs of CTR and CVR. The colored points correspond to differing values of $\pi_o$, representing the four different A/B test groups.