So, I started from the basics. I wanted to analyse the MNIST dataset and the efficiency of various machine learning algorithms in classifying the handwritten digits. The results were as follows:
- Logistic Regression: 0.9218
- Stochastic Gradient Descent: 0.8844
- SVM: DNF
- Artificial Neural Network (MLP): 0.939
- CNN: 0.9924
- Decision Tree: 0.8768
- Random Forest: 0.9285
I was surprised by the performance of the trees and random forest algorithms.