180 likes | 319 Views
Student’s Research Group. Intro. A student’s research group is forming Institute of Business Informatics Aims & scope extend the interest of students provide some interesting topics to work on promote cooperation between the academic staff and students who want to extend their knowledge
E N D
Intro • A student’s research group is forming • Institute of Business Informatics • Aims & scope • extend the interest of students • provide some interesting topics to work on • promote cooperation between the academic staff and students who want to extend their knowledge • Many topics are proposed, but… 2
My personal favourites • Artificial Intelligence • Softcomputing • Evolutionary algorithms • Multiobjective optimization • Combinatorial optimization • Dynamic optimization 3
Evolutionary Algorithms Main Population Parent Population Next Population evolutionary operators 4
Evolutionary Algorithms You can grow an antenna if you like This was done using a technique called Genetic Programming 8
Multiobjective Optimization • Many criteria that have to be optimized • Computing power vs. cost • Investment return vs. risk • Strength of an element vs. weight • EAs are well-suited for this 9
Multiobjective Optimization • Constrained problems • Can be solved using EMOO algorithms • Constraint violation as a criterion • IDEA: Infeasibility-Driven Evolutionary Algorithm 10
Combinatorial Optimization • Want a fast and cheap travel? • sure, but these are conflicting criteria (usually)… • …and the TSP is not so easy • fortunately, suboptimal solutions are quite good source: http://gtresearchnews.gatech.edu/reshor/rh-f04/tsp.html 11
Combinatorial Optimization • It’s possible to tackle harder problems • Q3AP is O((n!)2) • so, for n = 20, we have (20!)2 5.91 1036possible solutions • there are a mere 2.5 109 transistors in a CPU (and we’re talking about a 10-core Xeon Westmere-EX here!) • it performs up to about 38 GFLOPS 3.8 1010 floating-point operations per second (the estimated performance of X5365) • electronic computers are known for (much) less than 3.15 109seconds ( 100 years) (3.8 1010) (3.15 109) 1.2 1020 << 5.91 1036 12
Dynamic Optimization • Goal(s) and constraints change over time • Typical situation in real life • The algorithm has to adapt to the new situation • The evolution does not see into the future… • … but you can combine it with prediction Source: P. Filipiak, K. Michalak, P. Lipiński Infeasibility Driven Evolutionary Algorithm with ARIMA-Based Prediction Mechanism Lecture Notes in Computer Science, volume 6936, pp. 345-352. Springer, 2011. 13
Applications • Finance • stock market trading rules • portfolio optimization • credit scoring 14 10:30 10:10 13:10 14:10 9:50 12:50 13:50 11:30 12:30 13:30 11:10 12:10 11:50 10:50
Applications • Robotics • inverse kinematics 15
Questions… • Do I have to know all that to start? • No! You will learn as you go • But, be willing to learn and work • How do I start? • formal structure is yet forming • but, let me know ASAP that you want to join krzysztof.michalak@ue.wroc.pl • we‘ll arrange a meeting for participants 16
Questions… • Do I have to know programming? • No, at least not at first • But, it will become useful later • So, again, be willing to learn and work • Are we limited to the topics discussed? • No, not at all • But, I can offer most help with these • I would like to work moslty on AI / Softcomputing • Other topics will be handled by other people 17
Questions… • What happens if I try and fail? • Not much • It does not have to happen • There will be plenty of work at various difficulty levels • There are no bad grades or penalties • It’s no shame to try and fail • It is a shame to underestimate yourself :P • You will learn a few things anyway Other questions? 18