A Parameter Update Balancing Algorithm for Multi-task Ranking Models in Recommendation Systems
Authors:
- 9 December 2024
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 929-934
Abstract
Multi-task ranking models have become essential for modern real-world recommendation systems. While most recommendation researches focus on designing sophisticated models for specific scenarios, achieving performance improvement for multi-task ranking models across various scenarios still remains a significant challenge. Training all tasks naively can result in in-consistent learning, highlighting the need for the development of multi-task optimization (MTO) methods to tackle this challenge. Conventional methods assume that the optimal joint gradient on shared parameters leads to optimal parameter updates. However, the actual update on model parameters may deviates significantly from gradients when using momentum based optimizers such as Adam. In this paper, we propose a novel Parameter Update Balancing algorithm for multi-task optimization, denoted as PUB. In contrast to traditional MTO method which are based on gradient level tasks fusion or loss level tasks fusion, PUB is the first work to optimize multiple tasks through parameter update balancing. Comprehensive experiments on benchmark multi-task ranking datasets demonstrate that PUB consistently improves several multi-task backbones and achieves state-of-the-art performance. Furthermore, we deployed our method for an industrial evaluation on the real-world commercial platform, HUAWEI AppGallery, where PUB significantly enhances the on-line multi-task ranking model, efficiently managing the primary traffic of a crucial channel.Keywords
This publication has 12 references indexed in Scilit:
- Improving Training Stability for Multitask Ranking Models in Recommender Systems2023
- AdaTask: A Task-Aware Adaptive Learning Rate Approach to Multi-Task LearningProceedings of the AAAI Conference on Artificial Intelligence, 2023
- Progressive Layered Extraction (PLE): A Novel Multi-Task Learning (MTL) Model for Personalized Recommendations2020
- Recommending what video to watch next2019
- Why I like it2018
- Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts2018
- Variations and extension of the convex–concave procedureOptimization and Engineering, 2015
- Convex Optimization: Algorithms and Complexity2015
- Multiple-gradient descent algorithm (MGDA) for multiobjective optimizationComptes Rendus Mathématique, 2012
- The Concave-Convex ProcedureNeural Computation, 2003