tputer.gif (7747 bytes)cds-top.gif (4410 bytes)Brock University
Department of Computer Science


back.gif (1097 bytes)home.gif (1132 bytes)forward.gif (1172 bytes)

2. Research

As Computer Science kept doubling its volume every 2.5 years, Vlad grew with it. Consequently, Vlad's interests are varied and somewhat encyclopaedic in nature.

Vlad's main interests are computer vision, parallel computing and simulation, with visual machine learning and parallel simulation of parallel systems to boot!

Other computing research interests include: artificial intelligence, pattern recognition and classification, evolution of expert systems, object orientation and software engineering (Ada2005, CASE, etc.), networking and operating systems.

Vlad has strong mathematical background which supports his research interests in operations research, statistics and simulation.

In particular, Vlad finds Artificial Intelligence to be in a state of great potential. Mathematical theories, heuristics and findings abound and are very promising, but the current state of computing technology does not allow (yet!) to verify all the theories and findings experimentally. Hence his focus on parallel processing, especially as it applies to Artificial Intelligence.

Consequently Vlad and his students wrote (in Ada 2005 for smooth handling of parallelism) a parallel simulator of parallel systems of arbitrary topologies, running synthetic programs of arbitrary levels of parallelism, and putting arbitrary loads on simulated processors and inter-CPU communication channels. All this with hope of finding practical guidelines for construction of operating systems for parallel computers, as well as classifying arbitrary algorithms into affinity groups of preferences to various classes of parallel systems. Lots of experimentation is ahead ... especially in search of parallel computer architectures conducive to visual machine learning. Ill this work is biologically inspired and constitutes a novel approach to problems of handling parallelism.

Vlad's interest in machine vision led him to collaboration with his then student, Behzad "Ben" Salami (now gainfully employed at IBM Canada) formulate a theory regarding the visual learning systems of vertebrates, in terms of machine perception and learning. All this started by attacking and solving problems related to automatic identification of birefringent, thin petrographic sections using a special polarizing microscope, designed by Dr. Frank Fueten (ERSC, Brock). This is no small feat, as the findings therein clarify certain inaccuracies in the mathematical theory of metric spaces by introducing the concept of a hyperball.

Further investigation of visual learning problems led to collaboration with another then student, Pascal Comte (now a Ph.D. candidate at Memorial University of Newfoundland) that culminated in identification of retinal image processing in animals and suggestions for implementation of such algorithms in robots. At the time of this writing their joint paper is in print in the special issue on Artificial Brains of the Neurocomputing Journal. This is no small feat either, as their findings offer a model of functionality of human visual cortex.

Relying on his biological and evolutionary (i.e. recursive) inspirations Vlad feels that he is on a roll now!

Inspection of the mathematical apparatus in his theories of vision led to quick conclusion: mammalian auditory perception is very similar to vision and begs for another paper, this time in collaboration with his brother, Dr. W. Gregory Wojcik, M.D., on the new generation of cochlear implants. After all, audio signals can be thought of as simplified, one-dimensional video signals ... This work is in progress.

Currently on sabbatical, Vlad plans to touch up his theory of animal visual processing: Everybody knows that the sensors on mammalian retinas (rods and cones) are distributed on a hexagonal grid. However, analogous sensors on silicon retinas are aligned in rows and columns, on a rectangular grid.

Vlad follows his gut feeling that relevant redesign of silicon retinas could facilitate speed improvements in robotic visual processing, by better exploiting massive parallelism inherent in his algorithms.

Let us wait and see what transpires ... Knock, knock on wood!

back.gif (1097 bytes)home.gif (1132 bytes)forward.gif (1172 bytes)cameo.gif (1740 bytes)Administrator: Vlad Wojcik mail.gif (1189 bytes)
Revised: August 13, 2010
2010 Vlad Wojcik