EHW

Fra Robin

Revisjon per 10. mar 2007 kl. 23:21 av Hansbe (Diskusjon | bidrag)
(diff) ← Eldre versjon | Nåværende versjon (diff) | Nyere versjon → (diff)
Gå til: navigasjon, søk

Innhold


Research

Main goal

Improve hardware evolutionary schemes to be able to implement complex real-world applications.

Main topics

  • Incremental evolution
  • Building blocks and domain-specific
  • Online adaptable
  • Applying commercial hardware technology
  • Real-world applications

Outline

Evolvable hardware (EHW) has recently been introduced as a new scheme for designing systems for real-world applications. However, so far the number of applications is limited. One of the main problems in evolving hardware systems seems to be the limitation in the chromosome string length. A long string is required for representing a complex system. However, a larger number of generations are required by genetic algorithms (GA) as the string increases. This often makes the search space too large and explains why only small circuits have been evolvable so far. Thus, work has been undertaken to try to diminish this limitation. There are several ways of solving this problem:

  • Dividing the GA.
  • Compressing the chromosome string.
  • Increasing the building block complexity.
  • Dividing the application.

We think that the most promising approach is to divide the application. Dividing the application is based on the principle of divide-and-conquer. It was proposed for EHW as a way to incrementally evolve the application (paper). The scheme is called increased complexity evolution, since a system is evolved by evolving smaller sub-systems. Increased building block complexity is also a part of this approach, where the building blocks are becoming more complex as the system complexity increases. Experiments show that the number of generations required for evolution by the new method can be substantially reduced compared to evolving a system directly in one operations. This is both for a character recognition problem (paper) and a road image recognition problem (paper).

Our present research is concerned with targeting:

  • Further developing the increased complexity evolution scheme.
  • Improving generalization of evolved digital circuits.


People

Faculty

Ph.D. Students

Master Students

  • Nancy C. Flores
  • Geir Aa. Senland


Previous Master Students

  • Jens Petter Sandvik
  • Vidar Engh Skaugen
  • Knut Arne Vinger

Ressurser

Personlige verktøy