The tutorial may either be run via Google Colab by clicking the blue badges, or directly by installing the appropriate Python modules yourself.
When running in Google Colab, it is best to change the runtime type to GPU: Runtime -> Change Runtime Type -> Hardware Accelerator -> GPU
This tutorial is designed to present neural networks from a practical, coding perspective and focus on their implementation via PyTorch.
Several notebooks are found in the notebooks
folder:
SGD_from_scratch uses a single neuron for trivial classification and regression tasks on pseudodata, implementing backpropagation and weight updates manually in Numpy.
Basic_Classification introduces PyTorch, using a non-trvial, but basic classification problem.
Basic_Regression again uses PyTorch to solve a basic regression problem using pseudo-data
3_Classification_Application is an example using real-world tabular data to training a simple classifier.
4_Regression_Application_Exercise is an exercise using real-world tabular data to training a simple regressor.
5_Regression_Application_Completed is an example of how 4_Regression_Application_Exercise could be attempted, including a few tricks for training a regressor.