Little ML projects I did from training when I was discovering the field.
You can find :
-
predict_presidents
: Can we predict if somebody was a USA president with a Neural Network using only the hyperlinks from his wikipedia page as features (except the POTUS hyperlink, of course) ? Is it better than a Personalised Page Rank using the hyperlinks between wikipedia pages as links in a graph ? -
random_walk
: We train a LSTM network to predict the next step of a markov random walk and study how the network stores its infos. -
official_pytorch_tutorials
: coded following the official Pytorch tutorials (for instance https://pytorch.org/tutorials/beginner/nlp/word_embeddings_tutorial.html#sphx-glr-beginner-nlp-word-embeddings-tutorial-py). Predict a name's origin, or generate a name from an origin. -
kaggle
: First try at a kaggle challenge (predict Kobe Bryant shot), comparing different methods (SVM, random forest, neural networks...). Not top performance but a good training nonetheless. -
First tests
: first NN implemented to get a grasp of how it's working