examples-counterexamples

Examples of some machine learning-related facts and libraries

Requirements

To run examples using code from src you'll need to run python setup.py develop.

Some examples also require graphviz installed (on Ubuntu you can do apt-get install graphviz)

For loading data for word embeddings or coil20 dataset see Makefile.

Notebooks

• Kernel SVMs: fitting sine function, nonlinear classification with RBF kernel

• Transforming nonlinear data with Kernel PCA

• (counterexample) if a dataset is linearly separable then it is linearly separable after applying PCA

• Extractive Summarization in Python, includes one word-embedding based method.

• Nonnegative Matrix Factorization (NMF) for topic modeling - very simple exaple on 20 Newsgroups

Deep learning

• Symmetric Deep Dream - added rotation/mirroring invariance to original Deep Dream code

• Variational Autoencoder (VAE) in Keras - Example of running generative model on very small dataset (Olivetti faces dataset - 400 examples). Adapted and refactored code from Keras project.

• Logistic regression - using Theano (code)

• Basic MXNet - logistic regression & MLP

• GANs in Gluon and in PyTorch on Colab! Using Gluon's Hybrid Blocks which utilize dynamic graphs by default, but can be compiled to static ones

• Visualizing Convolutional Neural Networks in Keras with Quiver

Other

• Minimal CoreNLP Scala notebook