John Uckele

I'm a software engineer living in the Greater Boston Area. I've been working in industry for 8 years. My interests include scientific computing, artificial intelligence, evolutionary biology, linguistics, gaming, and music theory.

Vivarium

Vivarium is my project for generating artificial life capable of creating cognitive association, specifically aimed at linguistics. The simulation is driven by reccurent nueral networks trained by an inline genetic algorithm.
Visit the project page.

Kexing

Kexing is a general board game playing AI using Monte Carlo decision trees as the core control with genetic algorithms and genetic programming for huerestic creation and optimization.

Hybrid SVM

To try and compensate for the higher runtime of RBF SVMs, I used K-Means to find meaningful centroids and then ran an RBF-like kernel function where the features become the distance from the K-centroids. This technique becomes interesting when RBF on the full data set is prohibitively expensive.

RMS Titanic

The RMS Titanic data is a common Machine Learning exercise data set. I applied half a dozen models to the data. Linear regression, logistic regression, decision trees, random forests, and an ensemble learning of the previous models.

Swamptalk

While working at Tertl Studos I developed a mobile game. Swamptalk is a word game that resembles a hybrid between Tetris and Scrabble.
Visit the project page.