myMNIST: Benchmark of PETNN, KAN, and Classical Deep Learning Models for Burmese Handwritten Digit Recognition

2026-03-19 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, medium

Summary

A new benchmark study, published on March 19, 2026, systematically evaluates eleven deep learning architectures on myMNIST (formerly BHDD), a publicly available Burmese handwritten digit dataset. The research compares classical models like Multi-Layer Perceptron, CNN, LSTM, GRU, and Transformer with recent alternatives such as FastKAN, EfficientKAN, the energy-based JEM, and physics-inspired PETNN variants (Sigmoid, GELU, SiLU). Using Precision, Recall, F1-Score, and Accuracy, the CNN model achieved the highest overall performance with an F1-Score of 0.9959 and Accuracy of 0.9970. The PETNN (GELU) closely followed with an F1-Score of 0.9955 and Accuracy of 0.9966, outperforming LSTM, GRU, Transformer, and KAN variants. JEM also performed competitively, while KAN-based models provided a meaningful alternative baseline.

Key takeaway

For AI scientists and research scientists working on regional script recognition or evaluating novel neural network architectures, this benchmark provides critical performance data. You should consider CNNs as a robust baseline and investigate PETNN (GELU) models for their competitive accuracy, potentially offering better performance than Transformer or KAN variants for similar tasks. This data can guide your model selection and experimental design for future research.

Key insights

CNNs remain a strong baseline for Burmese handwritten digit recognition, with PETNNs showing competitive performance.

Principles

• CNNs offer robust performance in digit recognition.
• PETNNs can outperform Transformers and KANs.
• Energy-based models are competitive alternatives.

Method

The study systematically benchmarks eleven deep learning architectures on the myMNIST dataset using Precision, Recall, F1-Score, and Accuracy to establish reproducible baselines for Burmese handwritten digit recognition.

In practice

• Consider CNNs as a primary baseline.
• Evaluate PETNNs for strong performance.
• Explore JEM for energy-based modeling.

Topics

• Burmese Digit Recognition
• Deep Learning Benchmarking
• Kolmogorov-Arnold Networks
• PETNN
• Convolutional Neural Networks

Code references

• DLYuanGod/MLP-KAN

Best for: AI Scientist, Research Scientist, AI Researcher, Machine Learning Engineer, Deep Learning Engineer

Related on AIssential

• IEEE Transactions on Neural Networks and Learning Systems, Volume 37, Issue 5, May 2026 — The volume showcases diverse advancements in neural networks, from foundational surveys to specialized applications and novel architectures.
• Urdu Katib Handwritten Dataset: A Historical Document Dataset for Offline Urdu Handwritten Text Recognition with CRNN-Based Baseline Evaluation — The Urdu Katib Handwritten Dataset (UKHD) and CRNN-BGRU-CTC model establish a baseline for historical Urdu handwritten text recognition.
• Non-frontal face recognition using GANs and memristor-based classifiers — Combining lightweight GANs for pose frontalisation with memristor-based neuromorphic recognition enables efficient non-frontal face recognition at the edge.
• P1-KAN: an effective Kolmogorov-Arnold network with application to hydraulic valley optimization — P1-KAN offers superior accuracy and stability for high-dimensional function approximation compared to MLPs and ReLU-KANs.
• On Sequence-to-Sequence Models for Automated Log Parsing — Mamba and Transformer models offer superior log parsing accuracy and sample efficiency, with Mamba providing cost advantages.
• Introducing Wake Vision: A High-Quality, Large-Scale Dataset for TinyML Computer Vision Applications — High-quality, large-scale datasets like Wake Vision are crucial for advancing TinyML person detection.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.