The main goal of Groningen Cognitive Systems and Materials is to create self-learning materials that will perform the tasks that are now assigned to thousands of transistors and complex algorithms in a more efficient and straightforward manner, hence, forming the basis for a new generation of computer platforms for cognitive applications, such as pattern recognition and analysis of complex data. To the best of our knowledge, this is the first initiative of such kind that unites expertise from the disciplines of physics, materials science, mathematics, computer science and artificial intelligence.
Our program on cognitive systems and materials aims to discover and develop physical building blocks (i.e. materials) with intrinsic cognitive functionality via cross-linked networks at the nanoscale, allowing more efficient and denser circuits than those of stateof-the-art solutions (e.g. neuromorphic chip TrueNorthTM). We will also investigate and design the optimal implementation of such new material structures at the system level. Hereby we keep a close connection to possible applications from the start, as the optimal characteristics of the materials and architectures will vary for different applications. Further, we aim to realize cognitive functionality on established computer platforms. To this end, the program will build on collaborations between researchers from materials science (physics and chemistry), computer science, artificial intelligence and mathematics.
For the first seven years (2018-2025), the research program has two well-defined goals:
- To demonstrate cognitive functionality using physical networks in materials. Specifically, the goal is to implement this cognitive functionality in a material system where the signal transfer paths and adaptive inter-connections are present at the nanoscale within the material itself;
- To build complex systems (capable of more complex learning functions) from the simpler, above mentioned, cognitive units.General challenges concern computer architecture and networking (complexity of hardware and software design, multilayered architecture), development of new program languages and algorithms, network synthesis, systems engineering (systems of systems), robustness to noise and variability, as well as scalability.
|Last modified:||21 March 2018 11.12 p.m.|