This repository contains my own implementation of machine learning algorithms that stimulate my interest. At the beginning of each notebook you will find a reference to the paper that describes the algorithm. My goal is not to reproduce every singe table and figure in the paper, I am satisfied when I have enough evidence that my implementation is correct. Sometimes the notebooks are not very refined because I try to limit the amount of time I work on a paper, because I lose interest, because I find something else that diverts my attention, etc. The current implementations are:
bayes_by_backprop.ipynb: Charles Blundell et al., Weight Uncertainty in Neural Networks, May 2015. Pure numpy implementation for linear regression, tensorflow and pymc3 on MNIST.coteaching.ipynb: Bo Han et al., Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labels, 2018. Implemented in keras.deep_gaussian_processes.ipynb: Jaehoon Lee, Yasaman Bahri, Roman Novak, Samuel S. Schoenholz, Jeffrey Pennington, Jascha Sohl-Dickstei, Deep Neural Networks as Gaussian Processes, International Conference of Learning Representation, 2018. Implemented in numpy and tested on subsamples of MNIST, also includes hyper-parameter optimization using gpyopt.relevance_vector_machine.ipynb: Tipping, Michael E. (2001). Sparse Bayesian Learning and the Relevance Vector Machine. Journal of Machine Learning Research. 1: 211–244. Implementation in numpy, both for classification and regression.sparse_gaussian_processes.ipynb: M. K. Titsias. Variational learning of inducing variables in sparse Gaussian processes, Proceedings of the Twelfth International Conference on Artificial Intelligence and Statistics. 2009. Implemented in numpy and tensorflow.