# Visualization, Discriminability and Applications of Interpretable Saak Features

## TL;DR

This paper explores Saak features as an interpretable alternative to deep learning features, demonstrating their visualization, properties, and effectiveness in image classification across multiple datasets.

## Contribution

It introduces a multi-stage Saak transform inspired by CNNs, providing a transparent, low-complexity feature extraction method with competitive classification performance.

## Key findings

- Saak features can be visualized to interpret their properties.
- Saak features achieve high discriminant power in image classification.
- Experimental results show competitive accuracy on MNIST, CIFAR-10, and STL-10.

## Abstract

In this work, we study the power of Saak features as an effort towards interpretable deep learning. Being inspired by the operations of convolutional layers of convolutional neural networks, multi-stage Saak transform was proposed. Based on this foundation, we provide an in-depth examination on Saak features, which are coefficients of the Saak transform, by analyzing their properties through visualization and demonstrating their applications in image classification. Being similar to CNN features, Saak features at later stages have larger receptive fields, yet they are obtained in a one-pass feedforward manner without backpropagation. The whole feature extraction process is transparent and is of extremely low complexity. The discriminant power of Saak features is demonstrated, and their classification performance in three well-known datasets (namely, MNIST, CIFAR-10 and STL-10) is shown by experimental results.