contact
Todd Johnson
johnsong@ics.uci.edu
In February 2012 I completed a PhD in computer science at the University of California, Irvine. My advisor was Eric Mjolsness. The main webpage for Eric's research group is Computable Plant.
johnsong@ics.uci.edu
In February 2012 I completed a PhD in computer science at the University of California, Irvine. My advisor was Eric Mjolsness. The main webpage for Eric's research group is Computable Plant.
research
My research interests are in machine learning, artificial
intelligence, and computational biology.
My PhD thesis focused on two novel approaches in probabilistic graphical models. Each of these is described on its own page. Each page also includes links to download software packages for the Mathematica programming langauge, and example notebooks.
Dependency Diagrams extend factor graphs (a superset of Bayesian networks and Markov random fields). I used them to implement a package that, among other things, automatically generates runnable source code for the process of performing Metropolis-Hastings sampling on arbitrary distributions.
Graph-Constrained Correlation Dynamics formalizes a method of representing probability distributions that evolve continuously in time. The software package leverages Dependency Diagrams to optimize GCCD models.
My thesis is also available as a PDF.
My PhD thesis focused on two novel approaches in probabilistic graphical models. Each of these is described on its own page. Each page also includes links to download software packages for the Mathematica programming langauge, and example notebooks.
Dependency Diagrams extend factor graphs (a superset of Bayesian networks and Markov random fields). I used them to implement a package that, among other things, automatically generates runnable source code for the process of performing Metropolis-Hastings sampling on arbitrary distributions.
Graph-Constrained Correlation Dynamics formalizes a method of representing probability distributions that evolve continuously in time. The software package leverages Dependency Diagrams to optimize GCCD models.
My thesis is also available as a PDF.
selected side projects
- I wrote this implementation of parallel Lam-Delosme simulated annealing for statistical numerical optimization in Mathematica.
- I implemented a Bayesian Logic model that predicts the outcomes of basketball games. This required implementing an approximate Gibbs sampler for BLOG, making me (according to Brian Milch) maybe the first person outside of his and Russell's research groups to implement a custom proposer class for BLOG. If this is something you need to do, I would be happy to talk about the process.
- David Orendorff, Drew Frank, and I placed third in the 2008 UCSD data mining competition using bagged combinations of neural networks, SVMs, and naive bayes classifiers.
personal
I read, when I can;
my Goodreads
page is here.
I can be found on
Facebook
here.
I am lucky enough to be married
to Ruth
Barrett, an excptionally talented neuroscientist.
We have a weblog,
where we post pictures and occassional
goings-on, for friends and family.
I also maintain
a weblog with some
friends from college.
Coincidentally, all of the males in my immediate family-in-law are also computer scientists. Sorted by the degree to which our research is related, these fine fellows are:
Coincidentally, all of the males in my immediate family-in-law are also computer scientists. Sorted by the degree to which our research is related, these fine fellows are:
- Leon Barrett (PhD from UC Berkeley, now at Google)
- Sam Barrett (PhD student at UT Austin)
- Marty Barrett (a professor at East Tennessee State)