Problem
Research questions and friction points this paper is trying to address.
Session-based Recommendation
Convolutional Models
Efficiency
Long-range Dependencies
Global Modeling
No Attention, No Problem: Rethinking Session-based Recommendation with Pure Convolution
🤖 AI Summary
This study addresses the high computational overhead of Transformers and the limited global modeling capability of traditional convolutions in session-based recommendation by proposing NextConvRec, a purely convolutional framework. The method introduces a Spatial-Positional Convolutional Encoding (SPCE) module that integrates positional biases with structural signals. Furthermore, it leverages depthwise separable convolutions to expand the receptive field as an efficient alternative to attention mechanisms, combined with graph convolutional networks for effective sequence modeling. Experimental evaluations on four benchmark datasets demonstrate that NextConvRec improves average accuracy by 1.73% while reducing inference time by 16.7%, effectively balancing recommendation performance with computational efficiency.
📝 Abstract
Session-based recommendation (SBR) predicts the next choice in a session by analyzing recent interactions. Transformer-based models are widely used because of their ability to capture long-range dependencies through self-attention mechanisms. In contrast, traditional convolutional models, although more efficient, are often limited by their weak global modeling capabilities and are losing ground in SBR tasks. In this work, we propose a Next-generation Pure Convolutional Framework (NextConvRec) for SBR tasks, aiming to balance efficiency and performance. NextConvRec uses a Structural and Positional Convolutional Encoder (SPCE) for preprocessing, combining learnable convolutional positional biases with session-level structural signals extracted through GCN layers. Its backbone convolutional module effectively expands the effective receptive field through depthwise convolutions and pointwise convolutions, enabling robust long-range preference modeling without attention mechanisms. Extensive experiments on 4 benchmark datasets show that NextConvRec outperforms several state-of-the-art baselines by around 1.73% on average, and reduces the average inference time per session by 16.7%. The convolutional architectures remain a promising direction for efficient and accurate session-based recommendations.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Session-based Recommendation
Pure Convolution
Graph Convolutional Network
Depthwise Convolution
Receptive Field Expansion
🔎 Similar Papers
No similar papers found.
