2. Bi-directional incremental evolution Dr Tatiana Kalganova Electronic and Computer Engineering Dept. Bio-Inspired Intelligent Systems Group Brunel University

3. Outline b Evolutionary process b Evolvable hardware b Bi-directional incremental evolution: basic concept b Bi-directional incremental evolution in evolvable hardware b Some applications of bi-directional incremental evolution

4. What is an evolution? Evolution Chromosome: DNA Chromosome: DNA Selection Crossover Mutation

5. What is an evolvable hardware (EHW)? b The logic circuit is designed using evolutionary algorithm chromosome Evolution Selection Crossover Mutation

6. Circuit design problem in evolvable hardware

7. “Stalling” effect in evolutionary process Mult2.pla - 5.000 generations Mult3.pla – 16.000.000 generations Mult4.pla - ? … 150.000.000.000 ?

8. “Stalling” effect in evolutionary process b REASON: the task is too complex to solve at once. b SOLUTION: introduce the new evolutionary process once the “stalling” effect is appeared b IMPLEMENTATION: the new fitness function is used for each evolutionary process

9. Bi-directional Incremental Evolution b IDEA: two directions of evolution to obtain the desired solution b CONCEPT: evolve the system from complex to simple and optimise using evolution from simple to complex b REQUIREMENTS: knowledge of system evolved and identification of heuristics b SUCCESS: use of simple different evolutionary processes identified by various heuristics b EXAMPLE: evolvable hardware

10. Bi-directional Incremental Evolution b Stage 1: Evolution towards a modularised system b IDEA: Evolution performs from complex system to sub-systems b Stage 2: Evolution towards an optimised system b IDEA: Evolution performs from sub- systems to complex system

11. Bi-directional Incremental Evolution (BIE) in EHW

12. Bi-directional Incremental Evolution (BIE) for EHW b Idea Evolve the system gradually using decomposition methods 1) Decompose the system into sub- systems 2) Evolve each sub-system separately 3) Assemble the complex system 4) Evolve the complex system b Advantages Evolving the circuits of the large number of variables Evolving the circuits of any complexity No restrictions on the application task

14. BIE in EHW

15. BIE: EHW-oriented decomposition

16. BIE: sub-circuit allocation

17. Direct and incremental evolutions

18. BIE in applications b Evolution of complex combinational logic circuits b Optimisation of control the fermentation process b Prediction in investment appraisal

19. BIE in prediction in investment appraisal b Problem: Design an Intelligent System for Risk Classification of Stock Investment ProjectsDesign an Intelligent System for Risk Classification of Stock Investment Projects

20. Training network b An effective bi-directional evolutionary strategy is elaborated, as direct evolution fails to rich a solution to the complex problem of optimising the weights and shift terms in the fuzzy network over a set of investment projects.

21. BIE b The strategy involves a decomposition and an incremental part. b The integral problem is first divided into subtasks of decreasing complexity by partitioning accordingly the training set of projects. b Then the subtasks are merged incrementally to optimise the integral solution.

22. Training partitioning Training-set partitioning and increment during bidirectional incremental evolution Decomposition part: the training set is partitioned at several levels, evolving the fuzzy network towards tasks with decreasing complexity. Incremental part: the training subsets are merged incrementally in reverse direction, evolving the network towards solving the integral problem. A dynamic objective function is applied at each decomposition and incremental level.

23. BIE results Performance of bidirectional incremental evolution and direct evolution in maximum fitness per generation Black line: bidirectional incremental evolution advances through several decomposition and incremental tasks and solves the general problem in 148,243 generations. Lighter line: direct evolution makes some initial progress and then stalls. increment al part decomposition part direct evolution bidirectional incremental evolution

24. Experimental results b The bi-directional strategy evolves a fully functional fuzzy network in 148,243 generations. b Direct evolution reaches only 46.33% maximum fitness in 500,000 generations. b Thus, the empirical results prove decisively the efficiency of the developed evolutionary strategy.

25. Some results b In all applications mentioned earlier it has been obtained that the optimal solution has been obtained at least in 100 times quicker then using standard evolution b The quality of evolved solution in this case remains the same

26. BIE in applications: Summary b Design of complex combinational circuits b Use of Decomposition methodsDecomposition methods Evolutionary strategyEvolutionary strategy

27. BIE in applications: Summary b Prediction in investment appraisal b Use of Decomposition methodsDecomposition methods Automatic re-scaling fitness functionAutomatic re-scaling fitness function Neural networkNeural network Fuzzy logicFuzzy logic

28. Conclusion b Bi-directional incremental evolution is the technique that can be used in evolution of complex systems b BIE allows to use evolutionary algorithms in both online and offline calculations b BIE can be used in large range of applications