An Optimized Toolbox for Advanced Image Processing with Tsetlin Machine Composites

TL;DR

This work addresses color image classification with interpretable Tsetlin Machines by building a toolbox of TM Specialists that each apply distinct image-processing techniques. It leverages the TM Composites architecture to fuse multiple independently trained TMs, augmented with an extensive Optuna-based hyperparameter search, achieving a CIFAR-10 accuracy of $82.8\%$ and surpassing prior TM-state-of-the-art by $7.7\%$. The study introduces seven Booleanization techniques across 22 TM Specialists and demonstrates that performance scales with clause counts and data augmentation, offering a plug-and-play foundation for future TM applications in image analysis. The results underscore the potential of interpretable, energy-efficient TM ensembles to close the gap with deep models on color-image tasks while enabling better interpretability and modular extendability.

Abstract

The Tsetlin Machine (TM) has achieved competitive results on several image classification benchmarks, including MNIST, K-MNIST, F-MNIST, and CIFAR-2. However, color image classification is arguably still in its infancy for TMs, with CIFAR-10 being a focal point for tracking progress. Over the past few years, TM's CIFAR-10 accuracy has increased from around 61% in 2020 to 75.1% in 2023 with the introduction of Drop Clause. In this paper, we leverage the recently proposed TM Composites architecture and introduce a range of TM Specialists that use various image processing techniques. These include Canny edge detection, Histogram of Oriented Gradients, adaptive mean thresholding, adaptive Gaussian thresholding, Otsu's thresholding, color thermometers, and adaptive color thermometers. In addition, we conduct a rigorous hyperparameter search, where we uncover optimal hyperparameters for several of the TM Specialists. The result is a toolbox that provides new state-of-the-art results on CIFAR-10 for TMs with an accuracy of 82.8%. In conclusion, our toolbox of TM Specialists forms a foundation for new TM applications and a landmark for further research on TM Composites in image analysis.

Figures (3)

• Figure 1: State transition diagram for a two-action Tsetlin Automaton with $2N$ states granmo_tsetlin_2021.
• Figure 2: Proposed workflow for image classification with Tsetlin Machine Specialists utilizing image processing techniques.
• Figure 3: Tsetlin Machine Composites accuracy on the CIFAR-10 dataset, plotted epoch-by-epoch.