Enhancing User Sequence Modeling through Barlow Twins-based Self-Supervised Learning
Yuhan Liu, Lin Ning, Neo Wu, Karan Singhal, Philip Andrew Mansfield, Devora Berlowitz, Sushant Prakash, Bradley Green
TL;DR
The paper tackles learning scalable user representations from unlabeled interaction data where labeled signals are scarce and negative samples are costly. It adapts the Barrow Twins self-supervised objective to user sequences by using two augmented views and a shared encoder U = R ∘ E with a projection head, optimized via the BT loss $\mathcal{L}_{BT} = \sum_{i}(1-\mathcal{C}_{ii}) + \lambda \sum_{i \neq j} \mathcal{C}_{ij}^2$, where $\mathcal{C}$ is the batch cross-correlation matrix. Key contributions include the first application of BT to low-redundancy, large-discrete-domain sequence data, showing consistent improvements over a dual-encoder baseline on downstream tasks such as sequence-level classification and next-item prediction, and a thorough analysis of augmentation strategies and training dynamics. The results demonstrate that BT-based SSL learns versatile, robust sequence-level representations that transfer well with limited labeled data, reduce dependence on negative sampling, and accelerate downstream convergence, enabling more scalable personalized recommendations. The work suggests avenues for future enhancements by jointly optimizing sequence- and item-level representations, potentially via reconstruction tasks or integrated next-item objectives.
Abstract
User sequence modeling is crucial for modern large-scale recommendation systems, as it enables the extraction of informative representations of users and items from their historical interactions. These user representations are widely used for a variety of downstream tasks to enhance users' online experience. A key challenge for learning these representations is the lack of labeled training data. While self-supervised learning (SSL) methods have emerged as a promising solution for learning representations from unlabeled data, many existing approaches rely on extensive negative sampling, which can be computationally expensive and may not always be feasible in real-world scenario. In this work, we propose an adaptation of Barlow Twins, a state-of-the-art SSL methods, to user sequence modeling by incorporating suitable augmentation methods. Our approach aims to mitigate the need for large negative sample batches, enabling effective representation learning with smaller batch sizes and limited labeled data. We evaluate our method on the MovieLens-1M, MovieLens-20M, and Yelp datasets, demonstrating that our method consistently outperforms the widely-used dual encoder model across three downstream tasks, achieving an 8%-20% improvement in accuracy. Our findings underscore the effectiveness of our approach in extracting valuable sequence-level information for user modeling, particularly in scenarios where labeled data is scarce and negative examples are limited.