Logan Dunbar

Profile

My name is Logan Dunbar. I hail from sunny South Africa, where I completed my undergraduate degree at the University of Stellenbosch in Mechatronic Engineering. For my final year thesis I used two low cost webcams to triangulate a hand in 3D space for recognising hand gestures. I then moved to the beautiful Cape Town where I completed my masters degree in Electrical Engineering at the University of Cape Town. My dissertation looked at using the human visual system for inspiration in dealing with complex environments. I did this using visual saliency, which is a measure of how salient/interesting/surprising regions of the image/video are based on colour statistics, shape, motion, etc. I uncovered some biases in the current metrics for scoring saliency and proposed a more visually plausible metric and applied it to some common datasets. I am currently pursuing my PhD in Artificial Intelligence at the University of Leeds where I am looking to integrate symbolic and connectionist paradigms to build systems which are able to generalise and abstract rapidly in novel, unknown and uncertain environments. I plan to use the connectionist paradigm to ground the symbolic system in the real world statistics, and to use the symbolic system to do hierarchical abstraction which will then be used to develop curricula with the goal of speeding up learning and improving the exploration of the state space.

Research interests

Artificial intelligence, reinforcement learning, curriculum learning, building machines that think and act autonomously whilst learning in an open-ended fashion. Bringer of the robotic uprising.

Qualifications

• Candidate - PhD in Artificial Intelligence
• MscEng (Electrical Engineering)
• BEng (Mechatronic Engineering)