Winners of Neukom Fellowships have been announced for the 2009-2010 inaugural year.

Fellowships will provide a full year of funding, including stipend and benefits, to Ph.D. students engaged in faculty-advised research in the development of novel computational techniques as well as the application of computational methods to problems in the Sciences, Social Sciences, and Humanities.

The inaugural winners are:


“SAVVY: Scalable Audio-Visual creatiVitY”
Ramona Behravan
Michael Casey – Faculty Advisor (Music)

A significant proportion of human cultural output is now available on-line in the form of images, video and audio. However, most creative tools are organized around the concept of constructing a single document using a small collection of locally stored clips. We propose to enhance creative applications via research
into audio-visual feature encoding, matching and retrieval at the scale of the Web, and to develop methods to embed such technologies into the creative tools of today1. SAVVY will enable audio-visual materials to be retrieved from large collections in real-time, synchronously with the actions of playing music clips or
video clips in a creative application such as Adobe Audition, Adobe Flash, Final Cut Pro or AfterEffects.


“Distributed Representation and Transformation of Information In the Brain”
Jyothi Swaroop Guntupalli
James Haxby – Faculty Advisor
(Psychological and Brain Sciences)

Understanding information processing mechanisms in the human brain remains an important challenge in both neuroscience and computer science. In order to understand these processing mechanisms, we first need to know how information is represented in the brain. Our proposed research is aimed at answering this fundamental question of how information is represented in the brain and understanding how this representation changes along information processing pathways.


“A Learning Classifier System for the Detection, Characterization, and Modeling of Genetic Heterogeneity”
Ryan J. Urbanowicz
Jason Moore – Faculty Advisor (Genetics)

Modern human disease research recognizes that there are many complicating factors which make the detection and modeling of associated or causative genes and environmental factors extremely difficult. Over the last decade, a variety of algorithms have been developed in order to address some of these complications. One such complicating phenomenon, genetic heterogeneity (GH), poses a particularly difficult challenge but has received little attention. GH refers to the presence of different underlying genetic mechanisms resulting in the appearance of the same or a similar disease phenotype. From an epidemiological perspective, the presence of GH denotes the occurrence of separate etiologies which fall under the same phenotypic characterization. While some methods have been suggested to “side-step” this problem, (I.e. clustering and data stratification) these methods often rely on arbitrary cutoffs and lead to an inherent loss of power. Additionally, these methods conform to a standard analytical paradigm which seeks to identify a single model representing the best solution to a given data set. We propose to develop a bioinformatics algorithm that can evolve multiple rules/models which collectively represent a solution, for the detection, modeling, and characterization of GH.