An Ensemble of Simple Convolutional Neural Network Models for MNIST Digit Recognition

TL;DR

This paper demonstrates that combining simple CNN models with different kernel sizes and data augmentation techniques can achieve near state-of-the-art accuracy on MNIST digit recognition, emphasizing the effectiveness of ensemble methods.

Contribution

The study introduces a simple yet highly accurate ensemble of CNN models with varied kernel sizes, achieving over 99.9% accuracy on MNIST without complex architectures.

Findings

Achieved 99.87% accuracy with majority voting ensemble.

Two-layer heterogeneous ensemble reached 99.91% accuracy.

Simple CNN models with data augmentation outperform more complex models.

Abstract

We report that a very high accuracy on the MNIST test set can be achieved by using simple convolutional neural network (CNN) models. We use three different models with 3x3, 5x5, and 7x7 kernel size in the convolution layers. Each model consists of a set of convolution layers followed by a single fully connected layer. Every convolution layer uses batch normalization and ReLU activation, and pooling is not used. Rotation and translation is used to augment training data, which is frequently used in most image classification tasks. A majority voting using the three models independently trained on the training data set can achieve up to 99.87% accuracy on the test set, which is one of the state-of-the-art results. A two-layer ensemble, a heterogeneous ensemble of three homogeneous ensemble networks, can achieve up to 99.91% test accuracy. The results can be reproduced by using the code at: https://github.com/ansh941/MnistSimpleCNN