For fast on-demand sampling from large categorical distributions,
pip install categorical, and check out the API examples, and the benchmark against numpy.
An implementation of word2vec in Python using Lasagne and Theano: pip install theano-word2vec or git clone https://github.com/enewe101/word2vec.git, then check out the API examples. Extend it or incorporate it into your Lasagne neural architectures.Research Interests
The dynamics of large populations—from social movements to natural disasters—now unfolds in the digital commons. My research is motivated by a desire to understand collective behavior. I pursue that end using social media and open data sources, coupled with machine learning, and models of human behavior.
There are two major challenges in understanding collective dynamics. One is to understand and model the underlying mechanisms. The other is to
decodethe streams of data that carry information about populations—information like tweets, Reddit posts, or Instagram photos. This
decodinginvolves training computers, using machine learning, to understand the meaning and significance of such content.
My PhD work focuses on text understanding, specifically on understanding how a particular article or set of articles frames an event, and how it depicts interactions between agents. My work combines natural language processing, neural networks, and communication theory, to build systems that understand the depiction of relationships between entities in text.
Recently I investigated how US residents differ geographically in how they frame immigration and US relations with Latin American countries. Currently I am working on a neural architecture that learns embeddings for entities and their relationships, using a word2vec-inspired approach.
Edward Newell, David Jurgens, Haji Mohammad Saleem, Hardik Vala, Jad Sassine, Caitrin Armstrong, and Derek Ruths. User Migration in Online Social Networks: A Case Study on Reddit During a Period of Community Unrest. AAAI Conference on Web and Social Media, 2016.