Research Interests

My current research interests include compiler optimizations and connectionist networks.

Compiler Optimization

Automated generation of parallel code is a challenging endevour even in the cases where sufficient information can be collected to determine it is safe to do so. I am particularly interested in bringing existing automated parallel compilation techniques to C language compilers; however, I also experiment with the fringe of what is feasible by considering the automated generation of code using message passing facilities such as MPI.

Connectionist Networks

Connectionist (or "neural") networks are a massively parallel model for computation involving highly interconnected sets of very simple processors in contrast to the monolithic processors typically used in modern computers.

I study a particular class of connectionist networks in which processing units use non-monotonic activation functions, offering a number of advantages over traditional networks. Of primary interest is the highly structured internal representation such networks tend to adopt facilitating interpretation and rule extraction. Other areas of interests include neural circuit complexity and applications of techniques from operations research to problems of network analysis.