Memory Is No Longer a Bottleneck: Memory-Efficient Graph Filtering for Scalable Collaborative Filtering

The article introduces Mem-GF, a memory-efficient graph filtering method for collaborative filtering that addresses the memory bottleneck associated with storing full item similarity graphs. By utilizing Krylov subspaces for approximating polynomial graph filters, Mem-GF achieves up to 5.74 times lower memory usage and 4.38 times faster runtime compared to existing methods, while also improving recommendation accuracy. This advancement is significant for practitioners as it allows for scalable collaborative filtering on large datasets without the prohibitive memory costs typically associated with traditional graph convolutional networks.