I hold a PhD in Mathematical Physics, obtained in 2003 from the Sao Paulo University, Sao Carlos, Brazil (supervisor: Dr ES Bernardes) and from the University of Warwick, Coventry, UK (supervisor: Professor Ian Stewart). I then held a post-doc and a research fellow position at the Computer Vision group, led by Professor L da F Costa at the IFSC (Institute of Physics of Sao Carlos). From 2007 to 2010 I was a Senior Researcher at NICTA (Canberra Lab) and an Adjunct Researcher at The Australian National University (ANU) in the Machine Learning group, led initially by Professor Alex Smola, then by Dr Wray Buntine and Dr Tiberio Caetano. I have been working exclusively at ANU since 2010 at RSB/JCSMR with Dr Andrew James and Professor Ted Maddess, with visual neuroscience, developing multi-focal methods for early diagnosis systems for cognitive diseases. I have published many peer reviewed journal papers since my PhD, and acted often as referee for major publications in the fields of solid state physics, computer science and computational neuroscience. I have taught core disciplines of the Physics and the Applied Mathematics curriculum: Thermodynamics, Electromagnetism, and Advanced Engineering Mathematics for physicists and engineers. I also taught at the Summer Schools in Logic and Learning, Canberra 2009, see http://videolectures.net/marconi_barbosa.
I like to build my own computer resources whenever it is more practical than just wait for others. I have built so far a few linux based clusters and more recently a node with multiple graphic cards. I manage all the software stack to run most common machine (deep) learning libraries and tools.
For the past 5 years or so I have been working with different aspects of Statistical Machine Learning mainly applied (with some commercial interest) to signal processing, (time series pattern analysis and classification) either derived from brain sensors or coming from sensors attached to the human body. I also have made contributions to theoretical machine learning using non-commutative harmonic analysis in the problem of learning (classification and inference on) complex structures such as preference lists or ranking. More recently, I am focusing on the mathematical (group theory, harmonic analysis, stochastic geometry) aspects of human visual perception and discrimination.
The aim is to search for neuronal mechanisms that might have evolved, giving humans the ability to recognize, visually, differences in features of surfaces and volumes of various materials--natural, man-made, mineral, organic. Those surface/volume features may be textural, geometrical and topological clues. A joint aim is to look at cognitive diseases that impair such visual mechanisms early on after its onset. See research project description below.
Areas of Expertise
- Sensory Processes, Perception And Performance
- Vision Science
- Pattern Recognition And Data Mining
- Knowledge Representation And Machine Learning
- Computer Vision
- Statistical Mechanics, Physical Combinatorics And Mathematical Aspects Of Condensed Matter
- Lie Groups, Harmonic And Fourier Analysis