As noted previously, I've been researching machine learning (just for fun).
MITx - 6.036
I'm starting with an undergraduate course from MIT that is online for free: MITx - 6.036: Introduction to Machine Learning.
Having just completed the course, here are some observations:
- The lectures are excellent, but note that they're theory- and math-heavy (and my math was rustier than I realized)
- I'd recommend learning about NumPy prior to starting this course (especially element-wise operations and broadcasting)
- So far, there has been no need (and no benefit) to enabling GPU acceleration (and setting it up on Windows was painful, for reasons I still don't understand)
- The provided code seems to be built for a previous version of TensorFlow's Keras interface (although the updates I have needed thus far were minor)
The course provides a great introduction to:
- Neural networks
- Convolutional neural networks
- Recurrent neural networks
- Recommender systems
- Decision trees
Note that the course doesn't cover some of the machine learning topics I'm most interested in:
Overall, it was a fun course and good Python practice, but I wasn't really inspired to go start a machine learning project because I'm not terribly interested in classification and regression problems at the moment.