A neural network built from scratch to classify handwritten digits from the MNIST dataset.
- Technologies Used: numpy
- Highlights:
- Custom architecture with 3 dense layers
- Achieved 97% accuracy
- Confusion matrix and prediction examples
We assume the image data of 28x28 is flatern into 784x1 dimension: