In football there is a saying “Don’t go where the ball is now, go where it is going”. The same applies to technology. In this research theme we explore emerging technologies and identify natural ways of computing with such technologies in order to meet the demands of the future information processing. The aim is to demonstrate the value and validity of the approaches in real life applications – like the skunkworks of computing, but balancing between risk and expected gain; instead of trying to find new solutions to old problems, we aim at enabling technologies/approaches that re-define the problems.
The application areas we are targeting are reconfigurable, analog and mixed-mode near sensor signal processing, probabilistic computing, and holistic inference and associative computing for cognitive information processing. We believe that certain key approaches are crucial in finding the right solutions for maximally power-efficient computing with emerging technologies: Signals should be continuous valued in continuous time, and computing should take advantage of the device physics and take place close to the sensor, which requires algorithms to be topologically mapped on the processing architecture. Sparse and distributed analog computing is considered for power saving and information processing benefits. Our target implementation mediums are CMOS-nanotechnology hybrids, FPGAs and embedded microprocessors.
- Nano-Friendly Computing Engines for Cognitive Tasks
- MEMENCO: Memristive and memcapacitive event-based neuromorphic computing
- MEMCAP: Solid-state memcapacitor for neural hardware and tunable analog circuits
Academy Research Fellow, Adjunct prof. of Parallel Computing & Systems
+358 (0)50 436 7577