Mushroom Edibility Rules

Apply both Neural Networks and Genetic Algorithms to extract logical rules for mushroom edibility. Firstly use NumPy and Pandas to assign continuous variables into discrete one-hot representation. The Neural Network is implemented by PyTorch. It involves an auxiliary cost function term to force weights to approach ±1 or 0, so the network can be interpreted as a tree-like logic graph. For the second approach, GA cooperates with Decision Tree, where the accuracy of Decision Tree built by GA selected features serves back as the GA fitness function. The training process is visualized by Matplotlib. The project report written by LaTeX is available here.

Source Code: https://github.com/ybhan/Mushroom-Edibility-Rules

• Python
• PyTorch
• Pandas
• Machine Learning
• LaTeX