MNIST Digit Classification Comparative Analysis of Machine Learning and Deep Learning Approaches

— Handwritten digit recognition has long served as a foundational problem in the field of machine learning, offering a controlled environment to evaluate and compare different modeling approaches. This study presents a comparative analysis of classical machine learning and deep learning techniques for digit classification using the MNIST dataset. Specifically, we examine the performance of Logistic Regression as a linear baseline, a feedforward Artificial Neural Network (ANN) as a non-linear model, and a Convolutional Neural Network (CNN) as a spatial feature extractor.Rather than focusing solely on accuracy, this work explores a broader set of evaluation criteria, including computational efficiency, model complexity, robustness to input variations, and suitability for realworld deployment. Through systematic experimentation and controlled benchmarking, we analyze how each model behaves under standard conditions as well as in the presence of noise and distortions.The results highlight a clear trade-off between simplicity and performance. While CNNs achieve the highest classification accuracy due to their ability to capture spatial hierarchies, simpler models like Logistic Regression and ANN remain competitive in terms of speed, interpretability, and resource efficiency. These findings provide practical insights for selecting appropriate models based on application constraints, particularly in environments where computational resources are limited.