Multilayer Perceptron
The goal
- Implement from scratch a multilayer perception
- Implement stochastic gradient-descent and the backpropagation algorithm
- Test effectiveness with a toy dataset
The tools
- Python
- Numpy (Fast computation on matrices)
The data
The project requires us to work with the Breast Cancer Wisconsin Diagnostic dataset. It is a csv file of 32 columns, the column diagnosis being the label you want to learn given all the other features of an example, it can be either the value M or B (for malignant or benign)
The project structure
It is divided into two folders:
bonus/subject/
The subject folder refers everything related to the Wisconsin Dataset.
The bonus utilises the same code, but is trained on the famous MNIST dataset.
In both folders, I wrote the corresponding code with the pytorch framework.