In this repository, I have written some code on understanding softmax-regression. In addition, I also tried to get familiar with Pytorch by programming softmax regression in the Pytorch flow as well as using Dataset and DataLoader.
The source code, in my view, is easy to read. You can change the mean values of the data or the number of classes in the main.py and softmax_regression.py, respectively. However, the boundaries drawn after training are valid only if the data is 2-dimensional.
Here is the example of classifying 3 classes:
