ML Research

Research projects in Machine Learning

Environment

conda create -n ml-research
source activate ml-research
make deps

Install OpenML

git clone https://github.com/openml/openml-python
cd openml-python
python setup.py install
cd ..
rm -rf openml-python

Install jupyter lab extensions:

jupyter labextension install @jupyterlab/plotly-extension

Projects

Generative Adversarial Networks for Gaussian Distributions

The gan_gaussian project is a basic intuition-builder for GANs. It implements as simple GAN that tries to learn a gaussian (normal) distribution. The project results can be reproduced by running:

cd gan_gaussian
python gan_gaussian.py

Running the script produces a few artifacts that are meant to help build some intuition about how GANs work and what kinds of distributions it learns when the generator is trying to trick the discriminator into accepting the generator's fake samples.

• gan_gaussian_plot.png
• gan_gaussian_evolution.png

METAlearn

MetaRL-based Estimator using Task-encodings for Automated machine Learning

META Learn is a deep learning approach to automated machine learning that parameterizes the API of machine learning software as a sequence of actions to select the hyperparameters of a machine learning pipeline in an end-to-end fashion, from raw data representation, imputation, normalizing, feature representation, and classification/regression. Currently the [sklearn API][sklearn] is the only supported ML framework.