This is a Github repo where I develop toy experiments for understanding various aspects of machine learning. In each folder, I'll perform a self-contained experiment that explores either a method, concept, or property of machine learning algorithms (not just deep learning).
The key thing with many of these experiments is that I want to dive deep into the core of concepts: unless I absolutely must, I won't be looking at the standard ML datasets like MNIST: the idea here is to explore the underlying principles, which is most easily done on relatively simple (one dimensional) data sets, which are just randomly generated functions.
I envision many of these will also be written up as blog posts, either on my own website, or on Medium.
Please feel free to use the code in this repo for your own experiments, and contact me with any questions or comments! Forks/Issues/PRs are welcome!
It seems like Binder's not working so well for for my conda
environment!
For the time being, if you'd like to go beyond looking at the code, please
use environment.yml
to create a local conda
environment to be able to
run the notebooks.