Simple Multi-Layer Perceptron (MLP) used for teaching BSc/MSc Data Science students. The teaching was split across 2 lab sessions (2 hours each) and aimed to teach students about the basics of neural networks.
Important aspects included:
-
Demonstrating the importance of vectorized code implementation (acheived using 'for loops' vs 'numpy inbuilt functionality'. See the 'Vectorization Teaching Version' Notebook.
-
Demonstarting the importance of activation functions. See the 'Activation Functions Teaching Version' Notebook.
-
Walking through the Python implementation of a Multi-Layer Perceptron (MLP). The Notebook also allowed students to investigate the impacts of changing network parameters (such as number of neurons in each layer & alpha backpropigation paramater). See the 'MLP Walkthrough Teaching Version' Notebook.
The 'wine' and 'iris' datasets are available at: https://archive.ics.uci.edu/ml/datasets.php