TabNSA: Native Sparse Attention for Efficient Tabular Data Learning
This addresses the problem of inefficient deep learning for tabular data, which is incremental as it builds on existing attention and MLP methods.
The authors tackled the challenge of applying deep learning to tabular data by proposing TabNSA, a framework that integrates Native Sparse Attention with a TabMixer backbone to efficiently model heterogeneous features, resulting in consistent outperformance over state-of-the-art models in supervised and transfer learning settings.
Tabular data poses unique challenges for deep learning due to its heterogeneous feature types, lack of spatial structure, and often limited sample sizes. We propose TabNSA, a novel deep learning framework that integrates Native Sparse Attention (NSA) with a TabMixer backbone to efficiently model tabular data. TabNSA tackles computational and representational challenges by dynamically focusing on relevant feature subsets per instance. The NSA module employs a hierarchical sparse attention mechanism, including token compression, selective preservation, and localized sliding windows, to significantly reduce the quadratic complexity of standard attention operations while addressing feature heterogeneity. Complementing this, the TabMixer backbone captures complex, non-linear dependencies through parallel multilayer perceptron (MLP) branches with independent parameters. These modules are synergistically combined via element-wise summation and mean pooling, enabling TabNSA to model both global context and fine-grained interactions. Extensive experiments across supervised and transfer learning settings show that TabNSA consistently outperforms state-of-the-art deep learning models. Furthermore, by augmenting TabNSA with a fine-tuned large language model (LLM), we enable it to effectively address Few-Shot Learning challenges through language-guided generalization on diverse tabular benchmarks.