Summary of Revisiting Nearest Neighbor For Tabular Data: a Deep Tabular Baseline Two Decades Later, by Han-jia Ye et al.
Revisiting Nearest Neighbor for Tabular Data: A Deep Tabular Baseline Two Decades Later
by Han-Jia Ye, Huai-Hong Yin, De-Chuan Zhan, Wei-Lun Chao
First submitted to arxiv on: 3 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper explores the revitalization of classical methods, such as K-nearest neighbors (KNN), with modern deep learning techniques to improve performance on tabular data. The authors revisit Neighbourhood Components Analysis (NCA), a differentiable version of KNN designed for semantic similarity capture, and adapt it using stochastic gradient descent (SGD) without dimensionality reduction. Surprisingly, this implementation achieves decent performance on tabular datasets, outperforming existing toolboxes like scikit-learn. By adding deep representations and training stochasticity, the authors further enhance NCA’s capabilities, rivaling CatBoost and outperforming existing deep tabular models in classification and regression tasks on 300 datasets. The study concludes by analyzing factors contributing to these improvements, including loss functions, prediction strategies, and deep architectures. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper takes a classic approach used for recognizing patterns in data, called K-nearest neighbors (KNN), and updates it with modern techniques from artificial intelligence. They test this updated version on different types of data and show that it works well, even better than some existing methods. By combining these old and new ideas, they were able to make the approach work even better, beating other popular methods in many cases. |
Keywords
* Artificial intelligence * Classification * Deep learning * Dimensionality reduction * Regression * Stochastic gradient descent
