Notes on the Theory of Markov Chains in a Continuous State Space

These are lecture notes that were delivered as a seminar (chalk talk) at the Pande Group lab meeting on February 12, 2016. They introduce some aspects theory of reversible Markov chains in a continuous state space, with an orientation towards models of the classical conformational dynamics of molecular systems.

The …

Read more →

Efficient maximum likelihood parameterization of continuous-time Markov processes

This afternoon, I spoke on our recent work on arXiv on efficient estimation of continuous-time Markov processes (master equations) from discrete-time data. Check out the slide deck!

Read more →

Building a Conditional Random Fields Tagger for Academic Citations

Over the weekend, I built a system that identifies, parses and formats unstructured academic citations. The system is not running anymore, because I didn't want to pay for hosting, but the source code is still available on github.

It can take a raw string like "Wang, L.-P.; Titov, A …

Read more →

What's new in MSMBuilder3!

The long awaited beta of MSMBuilder3 is here! I put together a short slide deck, which builds on some of Kyle Beauchamp slides, showing how the pieces fit together.

Read more →

Introducing Osprey

Last week, I started work on a new open source software project whose goal is to streamline hyperparameter optimization for machine learning algorithms. The tool is called osprey, and it's available on github, pypi, and readthedocs. It integrates closely with scikit-learn.

Osprey is designed to make hyperparameter optimization as easy …

Read more →

Passwordless SSH with Kerberos

In research computing, passwordless SSH makes everything much more convenient, especially when your data and computations are split across an array of cluster resources.

For most clusters, this is really easy by exchanging public keys. For Kerberos, it's a little more complex. Here are the steps I used.

  1. Follow your …
Read more →

Optimal Markov Models: Formulation & Pursuit

I spoke this afternoon about our recent work [on arXiv] on the parameterization and cross-validation of Markovian models of molecular kinetics. The talk focus on the theory for a variational formulation of low-rank approximations to the kinetics of reversible dynamical systems. Check out the slide deck!

Read more →

Deploying with Travis-CI and S3

Travis-CI is fantastic for testing software (continuous integration) during open-source development. Less widely known, I think is how useful it can be for deployment.

With travis-ci, you have full access (passwordless sudo) access to a fresh Ubuntu virtual machine that runs automatically on every commit to your github repository. Sure …

Read more →

Cross-validating tICA and MSMs

Some of our new work at the intersection of chemical physics and machine learning on the construction of Markov models is now out on arXiv and under review. The title of our manuscript is "Variational cross-validation of slow dynamical modes in molecular kinetics".

The question that lead down this road …

Read more →

New website!

This is my first post on this new website! I went ahead and purchased the domain rmcgibbo.org to try to consolidate my material on the web.

The domain was registered through GoDaddy, and its DNS settings were configured through Amazon Route53. This site itself is simply static HTML/CSS …

Read more →