I recently organised a Machine Learning masterclass for the Macquarie Astrophysics department. A collaborator of mine, Professor Benoit Liquet-Weiland from the statistics department, is writing a Machine Learning textbook (which you can see here) and very kindly agreed to present a two-afternoon course on supervised and unsupervised learning.
The first afternoon covered a wide range of things, starting from linear models and logistic regression, then moved on to regularisation and finished on softmax models (i.e. a shallow neural network) for multi-class classification problem.
The second day…
It was great to see lots of the statistics and ML tools I’m familiar with being explained from a statistics perspective.
To add
- Day 2
- Really nice classification task with non-linear inputs (end of day 1). Can also do the same thing with multiclass data, even when we have more labels than features.
- Interesting to compare the available
Randpythontools. Benoit had to work hard to getglmnetworking properly inpython!