Abstract:The cold-start problem is a long-standing challenge in recommender systems due to the lack of user-item interactions, which significantly hurts the recommendation effect over new users and items. Recently, meta-learning based methods attempt to learn globally shared prior knowledge across all users, which can be rapidly adapted to new users and items with very few interactions. Though with significant performance improvement, the globally shared parameter may lead to local optimum. Besides, they are oblivious to the inherent information and feature interactions existing in the new users and items, which are critical in cold-start scenarios. In this paper, we propose a Task aligned Meta-learning based Augmented Graph (TMAG) to address cold-start recommendation. Specifically, a fine-grained task aligned constructor is proposed to cluster similar users and divide tasks for meta-learning, enabling consistent optimization direction. Besides, an augmented graph neural network with two graph enhanced approaches is designed to alleviate data sparsity and capture the high-order user-item interactions. We validate our approach on three real-world datasets in various cold-start scenarios, showing the superiority of TMAG over state-of-the-art methods for cold-start recommendation.

What problem does this paper attempt to address?

### The Problem the Paper Attempts to Solve This paper attempts to address the cold-start problem in recommendation systems. The cold-start problem arises due to the lack of interaction data between users and items, leading to poor recommendation performance for new users and new items. Although traditional Matrix Factorization (MF) and popular deep learning models have achieved success in recommendation systems, they still face challenges in handling the cold-start problem. Specifically, new users may abandon the system due to poor initial recommendations, making it crucial to solve this issue. ### Background and Challenges 1. **Feature-level Strategies**: - Utilize user profiles, item attributes, and cross-domain knowledge to enhance the representation of new users or items. - Model feature interactions through Graph Neural Networks (GNN) and Heterogeneous Information Networks (HIN) to capture high-order collaborative signals. - These methods rely on the availability and quality of features. 2. **Model-level Strategies**: - Few-shot Learning and Meta-learning have made significant progress in addressing data sparsity issues. - Existing meta-learning methods learn global parameters through optimization algorithms (e.g., MAML) to initialize personalized model parameters. - These methods extract meta-knowledge in meta-training tasks and quickly adapt to new users and items during the meta-testing phase. ### Limitations of Existing Methods 1. **Local Optimum Problem**: - Existing meta-learning methods treat each user as a task and learn globally shared meta-knowledge. This coarse-grained global knowledge may lead the model to fall into local optima when dealing with users with different gradient descent directions. - As shown in Figure 1(a), due to age differences, ∇L4 dominates the direction of gradient descent, causing parameter 𝜃 to deviate towards the optimal solution. 2. **Failure to Fully Utilize Intrinsic Information and Feature Interactions**: - Existing methods lack in utilizing intrinsic information and feature interactions, which are crucial for modeling new users and items. ### Proposed Method To address the above issues, this paper proposes a Task Aligned Meta-learning based Augmented Graph (TMAG) model, aiming to solve the cold-start recommendation problem from both feature-level and model-level. 1. **Fine-grained Task Alignment Constructor**: - Use attribute-oriented autoencoders to extract latent representations of users and items. - Cluster users with similar representations into a group, treating each group as a task with consistent optimization direction, thereby alleviating the local optimum problem during meta-training. 2. **Augmented Graph Neural Network**: - Combine graph structure information and attribute information to enhance the adjacency matrix, alleviating the data sparsity problem. - Employ two graph augmentation methods to explore potential interaction signals from the perspectives of attributes and graph structure, capturing high-order user-item interactions. 3. **Contrastive Regularization**: - Introduce task-level attribute contrastive regularization to enhance latent clustering prior knowledge. ### Main Contributions - Propose a fine-grained task alignment constructor that can quickly adapt to new users, solving the local optimum problem. - Alleviate the data sparsity problem by combining graph structure information and attribute information to enhance the adjacency matrix. - Introduce task-level attribute contrastive regularization to enhance latent clustering prior knowledge. - Conduct extensive experiments on three real-world datasets, validating the superior performance of TMAG in various cold-start scenarios. ### Experimental Results - TMAG outperforms existing cold-start recommendation methods on multiple metrics. - Analysis of the impact of different components on recommendation performance shows that each component significantly contributes to the overall performance. - TMAG demonstrates good robustness in terms of support set sparsity and hyperparameter selection. - Provide qualitative analysis of the learned representations, showcasing the effectiveness of task alignment. Through these contributions, TMAG addresses the cold-start problem in recommendation systems.

Task Aligned Meta-learning based Augmented Graph for Cold-Start Recommendation

MetaKG: Meta-learning on Knowledge Graph for Cold-start Recommendation

CoMeta: Enhancing Meta Embeddings with Collaborative Information in Cold-start Problem of Recommendation.

Boosting Meta-Learning Cold-Start Recommendation with Graph Neural Network

Task-Difficulty-Aware Meta-Learning with Adaptive Update Strategies for User Cold-Start Recommendation

TAML: Time-Aware Meta Learning for Cold-Start Problem in News Recommendation

Improving the Performance of Cold-Start Recommendation by Fusion of Attention Network and Meta-Learning

They Like Comedy, Don't You? A Cluster-Based Meta-Learning for Cold-Start Recommendation

AdaMO: Adaptive Meta-Optimization for cold-start recommendation

Meta-Learned User Preference Estimator with Attention Network for Cold-Start Recommendation

Multimodal Meta-Learning for Cold-Start Sequential Recommendation.

Learning to Learn a Cold-start Sequential Recommender

Cold-Start Sequential Recommendation with Collaborative Boosted Meta Transitional Learning

Multifaceted Relation-Aware Meta-Learning with Dual Customization for User Cold-Start Recommendation

Meta-Learning with Adaptive Weighted Loss for Imbalanced Cold-Start Recommendation

Cold-Start Recommendation Based on Knowledge Graph and Meta-Learning under Positive and Negative Sampling

Relation-propagation Meta-Learning on an Explicit Preference Graph for Cold-Start Recommendation

MARec: Metadata Alignment for cold-start Recommendation

AdaML: an Adaptive Meta-Learning Model Based on User Relevance for User Cold-Start Recommendation.

Graph attention networks with adaptive neighbor graph aggregation for cold-start recommendation

Cold-start Sequential Recommendation via Meta Learner