This was a personal project, where I coded a neural network from scratch using Python.

To see it in action, run the Tests.py file. If you just want a quick glance at the results, check the Charts folder. It contains charts and images with samples and results.

The Tests.py file should run in about 30s. For a more in depth look on performance, I recommend checking out the Charts folder.

• Training samples: 60 000 (MNIST dataset)
• Batch size: 60
• Epochs (per batch): 50
• Learning rate (per batch): 0.01

• Layer 1: 784x16, ReLU activation
• Layer 2: 16x16, ReLU activation
• Output Layer: 16x10, Softmax activation

• Test samples: 10 000 (MNIST dataset)
• Tests: 10
• Average accuracy: 87%