Training different types of Neural Networks in order to achieve as higher as possible accuracy on the MNIST image classification problem.
- Feed Forward Neural Network: test accuracy up to 98%,
- Convolutional Neural Network: test accuracy of 98.8%,
- Ensemble of Convolutional Neural Networks with the same structures: test accuracy up to 99.09%,
- Ensemble of Convolutional Neural Networks with different structures: test accuracy up to 99.12%.