Self-modifying information processing structures - hierarchy and flexibilityin neural systems

Lines connect the coordinator with its partners. CORDIS does not always give exact coordinates for projects before 2014. These points are placed at city or country level.

Project objective

Research objectives and content This research investigates the mechanisms underlying intelligent human cognitive abilities by bringing together neural systems and the hypercycle theory, which accounts for the stabilization of structures in the self-organization of life by hierarchical reaction cycles. The principle of the hypercycle is promising to imbue novel neural network models with the abilities to define the problem spaces in which to search for solutions, and to learn hierarchical structures of concepts. One approach is to investigate how the state transitions of recurrent neural networks can be learned to have the structure of a hypercycle. Another approach is based on artificial life models, for instance cellular automata, where a way to build up cycle hierarchies in conjunction with neural processing still has to be found. It is expected that the results are applicable to language translation, for example. Training content (objective, benefit and expected impact) The applicant would specialize in cognitive structures and receive a PhD, and he would benefit from the scientific support and international contacts of the host research group. Through his results and experience he will increase his chances for a career in science or e.g. the telecommunications industry. Links with industry / industrial relevance (22) The host institution has ongoing cooperation with British Telecom about cooperative agents that are capable of language.

Original text from CORDIS.

Participants

• IMPERIAL COLLEGE OF SCIENCE, TECHNOLOGY AND MEDICINE · LONDONCoordinatorUnited Kingdom
• Not availableCity levelGermany

Links

• View on CORDIS