Automatic Self-supervised Learning for Social Recommendations

TL;DR

AusRec tackles the problem of leveraging multiple self-supervised auxiliary tasks (SS-A tasks) in social recommendations without manual tuning. It introduces a meta-learning based automatic weighting network to assign adaptive weights to SS-A tasks, balancing their contributions to the primary recommendation task. The method integrates a LightGCN-based encoder with diverse SS-A tasks (triangle relations, k-hop neighbors, and meta-paths) and optimizes via a bi-level objective to mitigate negative transfer. Empirical results on three real-world datasets show that automatic task weighting yields consistent improvements over strong baselines and existing SSL-based approaches, highlighting the practical impact of adaptive SS-A balancing in social recommendation systems.

Abstract

In recent years, researchers have attempted to exploit social relations to improve the performance in recommendation systems. Generally, most existing social recommendation methods heavily depends on substantial domain knowledge and expertise in primary recommendation tasks for designing useful auxiliary tasks. Meanwhile, Self-Supervised Learning (SSL) recently has received considerable attention in the field of recommendation, since it can provide self-supervision signals in assisting the improvement of target recommendation systems by constructing self-supervised auxiliary tasks from raw data without human-annotated labels. Despite the great success, these SSL-based social recommendations are insufficient to adaptively balance various self-supervised auxiliary tasks, since assigning equal weights on various auxiliary tasks can result in sub-optimal recommendation performance, where different self-supervised auxiliary tasks may contribute differently to improving the primary social recommendation across different datasets. To address this issue, in this work, we propose Adaptive Self-supervised Learning for Social Recommendations (AdasRec) by taking advantage of various self-supervised auxiliary tasks. More specifically, an adaptive weighting mechanism is proposed to learn adaptive weights for various self-supervised auxiliary tasks, so as to balance the contribution of such self-supervised auxiliary tasks for enhancing representation learning in social recommendations. The adaptive weighting mechanism is used to assign different weights on auxiliary tasks to achieve an overall weighting of the entire auxiliary tasks and ultimately assist the primary recommendation task, achieved by a meta learning optimization problem with an adaptive weighting network. Comprehensive experiments on various real-world datasets are constructed to verify the effectiveness of our proposed method.

Figures (4)

• Figure 1: Examples of various social relations in social networks. $u$ indicates a user, and $v$ indicates an item.
• Figure 2: Performance of multiple self-supervised auxiliary tasks for the advanced social recommendation method MHCN yu2021self under Recall@5 and NDCG@5 metrics over three datasets (i.e., LastFM, Epinions, and DBook). MHCN(w/o tasks) denotes the MHCN model without any auxiliary tasks enhanced. The versions of MHCN model with single self-supervised auxiliary task enhanced are represented by ssl 1 to 6, including social triangle relation, joint triangle relation, meta-path relation ($u_i \to v_j \to u_{i'}$), 1-hop neighbors, 2-hop neighbors, 3-hop neighbors.
• Figure 3: The overall framework of the proposed method. The parameters updating process consists of three main stages in the automatic weighting mechanism.
• Figure 4: The change of automatic weights on various SS-A tasks during training (Best viewed in color).