Natural and artificial intelligence
My main research focus is on nature-inspired approaches to artificial intelligence, especially neuroscience-inspired techniques and evolutionary algorithms. My most recent projects have covered extensions of backpropagation algorithm to networks with plastic connections and variable structure, biologically plausible learning in recurrent neural networks, and mechanistic models of visual attention and visual search in the primate brain.
See also the statement of previous research and future research plans at the bottom of this page.
My research objective is to understand how high-level perception and behavior are implemented in the brain. I build computational models in which various brain areas interact through basic neural operations (excitation, inhibition, modulation etc.) to give rise to phenomena such as attention, object recognition, and decision making.
Evolutionary algorithms and artificial life
During my PhD (University of Birmingham, UK) I worked on evolutionary robotics. I built a 3D environment with realistic physics in which "artificial creatures" (simulated articulated structures controlled by distributed neural networks) were automatically designed and optimized by Darwinian evolution. I evolved creatures for various tasks, ranging from simple locomotion to full-fledged physical combat.
While the original code was lost, it has been refactored and published on Github by Shane Celis.
You can also look at Youtube videos of evolved creatures.