Nearest Neighbor Classification Based on Activation Space of Convolutional Neural Network
Xinbo Ju, Shuo Shao, Huan Long, Weizhe Wang · 2021
In this paper, we propose a new image classifier based on the incorporation of the nearest neighbor algorithm and the activation space of convolutional neural network. The classifier has been successfully implemented on some state-of-the-art models and further improve their performance. The main technique tool we use is convex hull based classification and its acceleration. We find several phenomena which we believe are of both theoretical and application interest: 1) in several cases, the new classifier outperforms original CNN by reaching higher accuracy; 2) the classifier can work more efficiently by combining with sampling strategy; 3) centroid of each convex hull shows surprising ability in classification. Most of the work have strong geometric meanings, which helps us have a new understanding about convolutional layers.