- Implementation of a Feed-Forward Neural Network model in JAVA for Intelligent Character Recognition.
- Working with the MNIST dataset.
- Preprocessing the data in MATLAB using various methods such as binarization, normalization, reduction and different types of transformations (edge detection, elastic distortions).
- Training the model with gradient descent optimization algorithm.
- 98.22% accuracy on the test set using 784-255-10 NN architecture.