Publications

Posts

Below are some images, produced with ggplot2, for our work on spatiotemporal smoothing with the graph-fused lasso. The figures on the left (red) show the productivity (dollars per hour) at the airport, and the figures at the right in downtown (blue). The two top figures show the models without postprocessing, and the bottom two use Bernstein polynomials smoothing and 400 degrees of freedom. To do The image may improve using facet instead of four separate gifs

CONTINUE READING

Current Draft: 2018-11-26 Here’s a list to Reinforcement Learning ideas and papers. It is mostly for personal research, as part of my work as PhD student at the University of Texas at Austin. Course: Research Elective Fall 2019 Advisor: Prof. Stephen G. Walker Note: This is not an extensive literature review, but a broad overview to guide our research, with the specific goal of exploring and extracting the statistical and probabilistic ideas and areas of opportunity in reinforcement learning literature.

CONTINUE READING

In this tiny post, I show with a minimal example the main difference between the distributed computing approaches of Julia and MPI. For this we will use Stampede 2, the University of Texas’ supercomputer (the faster at any university as of now). This is an Intel Skylake cluster that uses SLURM as it’s job queue system. We will show step-by-step how to run a distributed hello-world program in both languages.

CONTINUE READING

A common issue when starting a new project involving data is choosing a format in which to store the data. The answer to the question of course depends on several aspects: the sources, the data types, the purpose, and its relational structure. While modern approaches in Big Data endorse the strategy of first extracting high volumes of raw data and worry about processing later; in here we shall be concerned with the opposite case, when data has to be captured manually by a human team, and we don’t want to spend hours later trying to make it amenable for the computer to read.

CONTINUE READING

I sincerely hope that this Hugo website is–finally–the beginning of the blog I’ve been searching for a long time. Every year or so ‘I feel it’s time’, but I’m am always detained by the boredom of the existing technology: neither I want to use log-in solutions such as Wordpress, nor I want to spend years writing html code, just to communicate a simple thought. Knots and the name of this blog… Before becoming a PhD student, I spent a some time thinking about geometry and knots.

CONTINUE READING

Contact