''Truth doesn't care about opinions''


Non-Technical Description

With the digital revolution, data is everywhere. Text, images, videos and sound are being digitized at an astonishing rate. Much more than that is becoming digital however. High-throughput data acquisition systems are routine in microbiology laboratories, medical imaging devices, particle accelerators, telescopes. News, literature, knowledge and even people's diaries (blogging, social networking) are becoming digital on the Internet. Digital technology is transforming everything into 0's and 1's.

My research is devoted to creating and applying methods that extract relevant knowledge out of this ocean of data. In other words, I create automatic computational tools that analyze the 0's and 1's, telling us what is important in the big picture, according to the question we want to answer. Typical types of problems exemplifying where my expertise is useful include the following:
  • Microbiology: given high-throughput data from cells of an individual, infer the likelihood of him/her developing cardiovascular disease.
  • Video Surveillance: given a long video from a surveillance camera, infer automatically and quickly if it contains the face of a given suspect.
  • Internet: given the textual content of a webpage, determine automatically not only the topics it is concerned with, but also if it has a neutral, positive or negative sentiment about the topics (e.g., topic = `politics', sentiment = `negative towards party A').
I've applied these methods to solve practical problems in several areas, including image recognition, immunology and webpage ranking.

Technical Description

I work primarily in machine learning and its applications to computer vision. Most of my work is on graphical models, structured estimation and matching. I have also recently published on computational biology, complex networks and text analysis. In the far past I did research in theory and simulation of nonlinear dynamics in particle accelerators.