Gustavo Lacerda de Melo
Gustavo Lacerda
 

+1 604 345 8742

I am a graduate student at UBC, specializing in machine learning. I am currently doing an MSc thesis in statistical genomics, with Jennifer Bryan. I am seeking PhD positions in statistical computation (in areas such as Machine Learning, Statistics, Data Mining, Bioinformatics, Neuroinformatics) starting September 2010.

I have long defined myself as an "AI person", but I no longer think this is a good label for me. I like working with structured data, and my competitive advantage is in projects that involve mathematical programming, spatial intuitions and algorithm invention.

Although I'm firmly committed to empirical work, I'm a bit of a theoretician by temperament: I'm rather fond of theoretical ideas, and in a past life I dedicated lots of time to thinking about paradoxes and foundational questions.

CV(as of 9 September 2009 or later)     Projects

previous affiliations

Carnegie Mellon University 2006-2008 (programmer for HCII, researcher at Machine Learning Department)

Universiteit van Amsterdam 2003-2005 (MSc in Logic at ILLC)

conferences and summer schools

IPAMGSS 2007       ICML/UAI 2008       SFI Summer School 2009
NIPS 2008, 2009       CogSci 2008, 2009.

some things I like

argument mapping, bikes, bluegrass, Emacs, functional programming, infoviz, Kiva, musical instruments, open data, wikis.

some handy links (updated 2010)

mathoverflow




picture of Gustavo


blog
Google is my friend. I'm the first hit each of these queries (checked on 11 Feb 2010): "cyclic+discovery", correlation+"unmatched+data", correlation+"triangle+inequality", ICA+identifiability+Gaussian


Underneath this page, this website is run on a wiki, so...
Note: I take full responsibility for the content on this page and any other pages that don't have an "edit" link, as they are only editable by me. -- Gustavo Lacerda

papers (see all publications)

Discovering Cyclic Causal Models by ICA (UAI2008) (paper, video lecture with slides) extends LiNGAM to discover cyclic models; The non-Gaussian model leads to a finer level of identifiability than what can be achieved in the Gaussian case (e.g. by Richardson's CCD), and allows us to relax the faithfulness assumption. We prove theorems about identifiability, specifically about when a unique model can be identified.

(draft) Upper-Bounding Proof Length with the Busy Beaver (2008) (PDF) - This note presents a Chaitin-esque result. I derive an (uncomputable) upper bound on the length of the shortest proof of any given statement, as a function of the length of the statement; and briefly discuss implications. Mathematically trivial, but original (to the best of my knowledge). Could possibly be useful if we ever have good estimates of BB for n large enough to encode an interesting question.

see all papers



tutorials

- Independent Component Analysis (ICA) (slides) Introduces ICA, and tackles some very common misconceptions, 30 minutes.

- Introduction to Kolmogorov Complexity (with Liliana Salvador) (slides), 45 minutes.

- Introduction to Machine Learning and Bayesian inference (slides), 45 minutes.




Older Content



This website is permanently under construction. You may notice that behind this frontpage is a MediaWiki site. Someday I'd like to have indexing. For now, keyword searches will have to do. RIP Xanadu
.