[Notes] To those who have failed to get two OK's If you still want to pass this course, then try last task of "Sorting network" to find a sorting network with minumum comparisons to sort 16 items, using evolution of, say, a population of 1000 chromosomes with 240 genes each of which is one of the integer from 1 to 16. Show fitness vs generation + 5 set of best network (16 lines + arrows) from the start to the end as usual. Report should be sent to me by email by 28 October 2016. After that I really couldn't do any action to rescue you. ============================================================================================================= II-10-A Sep Oct Nov total eval Wednesday 16:50-20:00 07 07 21 21 05 05 19 19 02 02 (09)(09) 16 16 Monday 16:50-20:00 24 24 Friday 15:10-16:40 28(411) Saturday 16:50-18:20 14 14(310) additional ------------------------------------------------------------------------------------------------------------- Siarhei Shalyhaila 5 4.4 (4)* I #(2) 4 | 5 | 24.4/24 OK Ihar Yachnik 5 (4)* (4)* am (2) 3.5 | 3 | 4 25.5 OK Maxim Malcev 5 4 5 (2) 4.5 | 2.5 | 23 failed Alesya Polik (5)* (2) (4)* in (2) 4.5 | 3 <--- 3 23.5 ok Nickita Yurchuck (2) 4 (4)* a (2) 4.5 | 2.5 o 19.0 ok Alex Hreben (5)* 4 (4)* conf (2) 4.5 | 2 4.5 26.0 OK Siarhei Savaniuk 5 4 4.5 3 4.5 <-- 3 24.0 OK Yauhen Sampir 5 4.4 4.8 at (2) - 5 2.5 23.7 failed Yauheni Shablouski 5 4 (4)* Poland (2) 4.5 2.5 4 26.0 OK Dzmitry Rybalko 5 (2) 5 (2) (2) 2.5 18.5 failed Yury Yuryn 5.1 4 (4)* &5 4 5 27.1 OK Maxim Tatachka 5 4 (4) 3 4.5 3.5 24.0 OK Dzianis Palchuk 5 (2) (2) (2) 4.5 5 4 24.5 OK ------------------------------------------------------------------------------------------------------------- %=>10/12; #=>10/26; &=10/27 10 ------------------------------------------------------------------------------------------------------------- 09/07 All One Problem starting with a population of 100 rondom binary chromosomes with 1000 genes. Evolve the population with Trancate-selection, One-point-crossover. Fitness is the number of 1. 09/21 Minimization of 20-D Schwefel function and its 2-D versioan. 10/05 Traveling Salesperson Problem of 20 cities. 10/24 Iterated Prisonner's Dilemma (IPD) with 40 players. 10/28 Multi-fitness problem: Seek non-dominated solutions of y=(x-1)^2 and y=(x-5)^2 0<=(x-1)^2 ============================================================================================================= II-10-B Sep Oct Nov total eval Wednesday 16:50-20:00 14 14 28 28 12 12 26 26 09 09 Monday 16:50-18:20 (14) ------------------------------------------------------------------------------------------------------------- Andrey Avdey 3.5 - 4 4 5 4 20.5/20 OK Dmitriy Boyko - 5 2 4 5 3 19.0 failed Vladislav Golovchik 3.5 5 5.1 4.5 5 23.1 OK Artyom Drapezo 5.1 5 4 4 5.1 23.2 OK Oksana Zanko 5 5 4.5 4 5 23.5 OK Kirill Kirilyuk 4 5 4 4 - 3 20.0 OK Kirill Kovalyuk 3.5 5 4.5 4 5 22.0 OK Maxim Kravchuk 5 5 4.5 4 5 23.5 OK Vladislav Mashchuk (2) 2 4.5 3.5 - ---- failed Artyom Yudenkov 3.5 5 4.5 4 5 22.0 OK Polina Rachkovskaya 5 5.1 5.1 $4 5 24.2 OK Vladimir Dmitruk 4.9 5 4 4 5 22.9 OK -------------------------------------------------------------------------------------------------------------- $=>10/28 10 -------------------------------------------------------------------------------------------------------------- 09/14 All One Problem starting with a population of 100 rondom binary chromosomes with 1000 genes. Evolve the population with roulette-selection and one-point-crossover. Fitness is the number of 1. show a table of fitness, probability-to-be-selected, how-often-selected-actually each generation in addition to fitness vs generation 09/27 Lucky dog with 20 dogs at (500,500) at start looking for the ham at (800,800). 1000 steps are allowed. 10/12 Maximization of 3D Schwefel function z=x*sin(|x|)+y*sin(|y|) -5<=x,y<=5 20 chrosomes with 22 binary genes. by Fitness Sharing algorithm. 10/26 Minimization of y = sin^6(5*pi*x) 0<=x<=1 by Crowding Algorithm with 20 chromosomes of 10 binary genes. ============================================================================================================= II-11-A Sep Oct Nov Total eval Thursday 08:10-11:20 08 08 22 22 06 06 20 20 --> Tuesday 16:50-20:00 25 25 Wednesday 16:50-18:00 (09)(09) (12) (14) ------------------------------------------------------------------------------------------------------------- Igor Kondrashuk 5 4.4 3 I 3.5 5 20.9/16 OK Alexandra Letun (2) * &4.5 3 am 4.5 5 19.0 OK Pavel Kugaev 4 (2) (2) (2) 5 5 20.0 OK Karina Berezina 4 3.6 (2) * in 3.5 4.5 17.6 OK Ilya Babich 4.5 * 4.4 3 a 3.5 5 20.4 OK Kirill Zabrodsky 4.5 4.4 3 conf 3.5 5 20.4 OK Alexey Golodko (2) * 4.7 3 3.5 3 16.2 OK Vladislav Shukalo @2.5 &4.5 3 in 4.5 5 19.5 OK Mikhail Viktorenkov (2) * 4.6 (2) * Poland - 5 5 18.6 OK Ivan Bakunovich 4 3.4 @(2) * (2) 5 16.4 OK Roman Bursevich (2) * 4.5 (2) * 3.5 4.5 16.5 OK Andrey Abramchuk 3.9 3.6 (2) * 3.5 5 18 OK ------------------------------------------------------------------------------------------------------------- @=>10/13; &=>10/27 12 ------------------------------------------------------------------------------------------------------------- 09/08 The same as II-10-A on 09/07 but apply a mutation to every children created with a probability of 1/1000. 09/21 Minimization of 20-D Griewank function and its 2-D version. 10/06 Neural Network for N-even-parity with N-N-1 feedforward structure. 10/25 Dimension Reduction: 30 random points on the 10-D hyper-sphere to 2-D space. ============================================================================================================= II-11-B Sep Oct Nov total eval Thursday 08:10-11:20 15 15 29 29 13 13 27 27 Wednesday 16:50-18:00 (09) (09) ------------------------------------------------------------------------------------------------------------- Alexey Cherkasow @(2)* 3.5 3.4 4** 5 17.9/16 OK Denis Ramskiy 2.9=>4.5 5 4 4.5 18 OK Roman Rudskiy 1.5=>4 4 3.5 4.5 16 OK Ewgeniy Semenuk 2.5=>3.5 4.2 2.5 4.5 5 19.7 OK Mihail Sosnowskiy 2 =>3.5 2.5 4.5 4.5 5 20 OK Andrey Cheslow 2 =>5 4 3 4.5 16.5 OK Kirill Tsibikov 3 =>3 4.2 2.5 4.5 5 19.2 OK Wladimer Shukaylo @(2)* 3.9 (2) 5 5 17.9 OK Aleksandr Matweychuk 2.9=>5.1 4 3.5 4.1 17.1 OK ------------------------------------------------------------------------------------------------------------- @=>10/06; *=>claimed not by email but by UPS, but ** was found in inbox 8 ------------------------------------------------------------------------------------------------------------- 09/15 The same as II-10-B on 09/14 but uniform crossover instead of one-point-crossover! I.e. roulette + uniform 09/29 Maximization of 3-D Schwefel function. z=x*sin(|x|)+y*sin(|y|) -5<=x,y<=5 20 chrosomes with 22 binary genes. 10/13 Lucky dog with 4 sausage. 20 dogs start from (500,500) looking for 4 sausages at (200,200), (200,800), (800,200), (800,800) 10/27 Knapsack Problem wit`h 20 items with price and size of each item are x_i and w_i (i=1,2,...,20). all x_i and w_i and size of knapsace A should be randomly specified at the beginning. E.g. | i | 1 2 3 4 ... 20 | --------------------------- A = 500, maximize (n_1 * x_1 + n_2 * x_2 + ....... + n_20 * x_20) | x_i | 5 3 2 9 ... 8 | such that (n_1 * w_1 + n_2 * w_2 + ....... + n_20 * w_20) <=A | w_i | 2 6 4 5 ... 3 | upper limit of n might be A/w_minimum ============================================================================================================= [Previous Notes] When you send your results by e-mail, file name should be like muzyka-aleksander-0205.pdf not like my_lab1.pdf or something else! Those attended but disappeared were given the minimum point of <1.0> without checking theri result. If you don't like this, attend another group on another day. Then the point will be replaced with a new. Check your line by yourself!!! Mistake would not replaced the fact!!! If you confirm you really have OK, see Who-attend-when.txt ---------- All those who have not OK you should attend practice on 14 November => This will be your last chance! Those who have failed to get OK in this practice should write a report in PDF on Hinton Nowlan's Simulation (1) Repeat generation untill N of the best chromosome becomes 0 (1000 at the beginning) draw the graph of (i) N vs generation. (ii) number of 1, 0, and ? vs generation. of the best chromosomes in each generation. (2) Repeat (1) 10 times by changing random number seed and obtain avrage generation in which best N becomes 0 Repeat the above with 5 bit chromosomes, then 15 and 20 draw the bar graph of that average number vs number of genes of the chromosome (5, 10, 15, 20) Report should include title, author, explanation, souce cords too. Send the report to me by e-mail till Sunday night (20 November)! =============================================================================================================