Abstract:Collaborative filtering (CF) remains essential in recommender systems, leveraging user--item interactions to provide personalized recommendations. Meanwhile, a number of CF techniques have evolved into sophisticated model architectures based on multi-layer perceptrons (MLPs). However, MLPs often suffer from catastrophic forgetting, and thus lose previously acquired knowledge when new information is learned, particularly in dynamic environments requiring continual learning. To tackle this problem, we propose CF-KAN, a new CF method utilizing Kolmogorov-Arnold networks (KANs). By learning nonlinear functions on the edge level, KANs are more robust to the catastrophic forgetting problem than MLPs. Built upon a KAN-based autoencoder, CF-KAN is designed in the sense of effectively capturing the intricacies of sparse user--item interactions and retaining information from previous data instances. Despite its simplicity, our extensive experiments demonstrate 1) CF-KAN's superiority over state-of-the-art methods in recommendation accuracy, 2) CF-KAN's resilience to catastrophic forgetting, underscoring its effectiveness in both static and dynamic recommendation scenarios, and 3) CF-KAN's edge-level interpretation facilitating the explainability of recommendations.

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper attempts to address the issue of Catastrophic Forgetting in recommendation systems. Specifically: 1. **Background**: - Collaborative Filtering (CF) remains crucial in recommendation systems, providing personalized recommendations by leveraging user-item interactions. - Over time, many CF techniques have evolved into complex model architectures based on Multi-Layer Perceptrons (MLPs). - However, MLPs often suffer from catastrophic forgetting when performing continual learning in dynamic environments, meaning the model loses previously acquired knowledge when learning new information. 2. **Problem**: - Catastrophic forgetting leads to a decline in the performance of recommendation systems, especially in dynamic environments requiring continual learning. - Existing MLP methods struggle to effectively retain information from previous data instances when dealing with sparse user-item interaction data. 3. **Solution**: - To address this issue, the authors propose CF-KAN, a novel CF method that leverages Kolmogorov-Arnold Networks (KANs). - KANs resist catastrophic forgetting more robustly than MLPs by learning nonlinear functions at the edge level. - CF-KAN is designed as an autoencoder based on KANs, capable of effectively capturing the complexity of sparse user-item interactions and retaining information from previous data instances. ### Main Contributions 1. **Methodology**: - Proposed CF-KAN, a new CF method developed by utilizing the unique properties of KANs. - Compared to traditional MLP-based CF methods, CF-KAN can directly learn and adapt nonlinear functions at the edge level, addressing the issue of catastrophic forgetting. 2. **Comprehensive Analysis**: - Conducted extensive experiments systematically to validate the superiority of CF-KAN in terms of recommendation accuracy, robustness in continual learning scenarios, and scalability. - Experimental results show that CF-KAN outperforms state-of-the-art methods by 8.2% in Recall@20 metric, excels in dynamic recommendation environments, and has faster training speed. 3. **Enhanced Interpretability**: - Demonstrated CF-KAN's outstanding performance in highlighting the importance of individual user-item interactions during edge-level learning and pruning processes through visualizations and case studies. - This interpretability is crucial for model transparency and user confidence. ### Conclusion By introducing CF-KAN, the paper successfully addresses the issue of catastrophic forgetting in recommendation systems, enhancing the performance and stability of recommendation systems while maintaining model interpretability.

CF-KAN: Kolmogorov-Arnold Network-based Collaborative Filtering to Mitigate Catastrophic Forgetting in Recommender Systems

Improving Recommendations With Collaborative Factors

FourierKAN-GCF: Fourier Kolmogorov-Arnold Network -- An Effective and Efficient Feature Transformation for Graph Collaborative Filtering

CKAN: Collaborative Knowledge-aware Attentive Network for Recommender Systems

Continual knowledge graph embedding enhancement for joint interaction-based next click recommendation

Online Collaborative Filtering With Implicit Feedback

Collaborative Reflection-Augmented Autoencoder Network for Recommender Systems

A Collaborative Filtering-Based Recommender Systems approach for Multifarious applications

Hybrid Collaborative Filtering with Autoencoders

Kernel-CF: Collaborative filtering done right with social network analysis and kernel smoothing

Causal Collaborative Filtering

Knowledge Graph Enhanced Contextualized Attention-Based Network for Responsible User-Specific Recommendation

Recommendation via Collaborative Autoregressive Flows

A Complex Network Approach for Collaborative Recommendation

KGCFRec: Improving Collaborative Filtering Recommendation with Knowledge Graph

Hypergraph Contrastive Collaborative Filtering

Attribute-based Neural Collaborative Filtering

A Neural Autoregressive Approach to Collaborative Filtering

Knowledge-Adaptive Contrastive Learning for Recommendation

A Multiscale Neighbor-Aware Attention Network for Collaborative Filtering

RGCF: Refined Graph Convolution Collaborative Filtering with concise and expressive embedding