ó \ÿˆIc@sºdZddlZddlZddlZddlZddlZddlmZddlm Z ddl m Z ddl m Z ddlmZddlmZdd lmZd Zd „Zd „Zd „Zd„Zdejfd„ƒYZdd d„ƒYZdd!d„ƒYZdd"d„ƒYZdZedkr¬eej ƒdkrke!ej dƒZq¬eej ƒdkr¬dGHdGHdGHdGHdGHej"dƒq¬neeƒdS(#srTry out the N-queens problem for an arbitrary number of queens. This program uses Genetic Algorithms to try to solve the N queens problem, in which you place N different queens on an N by N chess board such that no two queens will be attacking each other. We represent queens on the board as a tuple like (1, 2, 3, 4, 5) which would be 5 queens diaganol across the board. Usage: python test_GAQueens.py where is just a number specifying how many queens you want to try to calculate this for. When called as part of the Biopython unit test suite, 5 queens are used. iÿÿÿÿN(tAlphabet(tGenerationEvolver(tOrganism(tConversionMutation(tSinglePointCrossover(tRouletteWheelSelection(tTournamentSelectionicCsVd|GHd}d|GHt|ƒ}tj|||tƒ}dGHtddƒ}ttddd d ƒ}tƒ}t|||d ƒ}tj tjƒƒ}t ||ƒ} | j t ƒ} tj tjƒƒ} g} x<| D]4} | j |krÛ| | kr| j| ƒqqÛqÛWtrRd || fGHx'| D]} d G| GHt| jƒq/WndS(NsCalculating for %s queens...iès3Generating an initial population of %s organisms...s7Evolving the population and searching for a solution...t mutation_rategš™™™™™©?tcrossover_probgš™™™™™É?tmax_crossover_sizeiis$Search started at %s and ended at %ss We did it!(tQueensAlphabetRtrandom_populationtqueens_fitnesstQueensMutationtQueensCrossovert QueensRepairRttimetctimeRtevolvet queens_solvedtfitnesstappendtVERBOSEt display_boardtgenome(t num_queenstnum_orgstqueen_alphabettstart_populationtmutatort crossovertrepairt t_selectort start_timetevolvert evolved_poptend_timetunique_solutionstorganism((stest_GAQueens.pytmain(s4           cCs€ddt|ƒdGHxKtt|ƒƒD]7}dGx%|D]}||krTdGq;dGq;WdGHq*Wddt|ƒdGHdS(s¦Display a genome in the N-queens problem. Inspired by the display function in the queens.py solution to the N-queens problem in the Python demo scripts. s+-s--t+t|tQt.N(tlentrange(Rtrowt genome_item((stest_GAQueens.pyRNs   cCs1x*|D]"}|jt|jƒkrdSqWdS(súDetermine if we have solved the problem. We just search through the population for an organism that has a fitness that is equal to the number of queens in the population. If so, we have a solution, otherwise we need to keep looking. ii(RR,R(t organismstorg((stest_GAQueens.pyRas cCsÌd}x¿tt|ƒƒD]«}d}x‰tt|ƒƒD]u}||kr8t||ƒ}t||ƒ}||kr€d}Pq­t||ƒt||ƒkr­d}Pq­q8q8W|s|d7}qqW|S(sèCalculate the fitness of an organization of queens on the chessboard. Arguments: o genome -- A MutableSeq object specifying an organism genome. The number returned is the number of unattacked queens on the board. ii(R-R,tinttabs(RRtcheck_queen_colt is_attackedtother_queen_coltcheck_queen_rowtother_queen_row((stest_GAQueens.pyR os"     R cBseZd„ZRS(cCs4g|_x$t|ƒD]}|jj|ƒqWdS(sEInitialize with the number of queens we are calculating for. N(tlettersR-R(tselfRtnum((stest_GAQueens.pyt__init__•s (t__name__t __module__R<(((stest_GAQueens.pyR ”sRcBs5eZdZdd„Zd„Zd„Zd„ZRS(s+A repair function to help create correct N-Queens solutions. This attempts to help generate correct solutions by offering some amount of repair to remove queens that are located in the same rows. After repair, a sequence should have no queens in the same row. So, if you start with something infeasible like (1, 2, 2, 3, 3, 4), after running it through repair you'll get a feasible individual like (1, 2, 5, 3, 6, 4). This should greatly reduce the number of individuals that need to be searched through in a population. icCs ||_dS(s¡Initialize the repairer. Arguments: o repair_prob -- The probability that we'll repair a genome. By default, we always repair. N(t _repair_prob(R:t repair_prob((stest_GAQueens.pyR<ªscCsLg}x?|jjD]1}|jt|ƒƒdkr|j|ƒqqW|S(s&Return all of the letters in the genome that are duplicated. This checks every letter in the genome (which are the rows of the chessboard, in this case), and adds them to a list of duplicated items if there is more than one of them, and then returns this list. i(talphabetR9tcounttstrR(R:Rt duplicatestitem((stest_GAQueens.pyt_get_duplicates´s cCsLg}x?|jjD]1}|jt|ƒƒdkr|j|ƒqqW|S(sÉReturn all of the letters in the genome which are unused. This checks the letters in the genome (which are th rows on the chessboard) and returns all items which are not used. i(RAR9RBRCR(R:RtunusedRE((stest_GAQueens.pyt _get_unusedÂs cCs³tjƒ}||jkr¯x‘|j|jƒ}t|ƒdkrFPn|jj|dƒ}|j|jƒ}t|ƒdksŒtdƒ‚tj|ƒ}||j|t__doc__R<RFRHR(((stest_GAQueens.pyRžs    RcBs2eZdZddd„Zd„Zdd„ZRS(s±Crossover operation to help in solving the N-Queens problem. This tries to perform smarter crossovers by picking out regions of the genome that have high fitness. It scans through both genomes in the crossover with a window half the size of the genome, and finds the region with the highest fitness in both genomes. It then recombines these high fitness windows to form the new genome that is returned. gš™™™™™¹?icCs||_||_||_dS(syInitialize to do N-Queens optimized crossover. Arguments: o fitness_func -- A function that can calculate the fitness of a genome. o crossover_prob -- The probability of having a crossover between two passed in organisms. o max_crossover_size -- The maximum crossover size of the 'best' region to search for. N(t_crossover_probt _fitness_calct_max_crossover_size(R:t fitness_funcRR ((stest_GAQueens.pyR<øs  c CsÎ|jƒ}|jƒ}tjƒ}||jkrÄ|j|jddƒ\}}|j|jddƒ\}} t|ƒt|ƒt|ƒt| ƒks§tdƒ‚|||_|| |_n||fS(s3Perform a crossover between two organisms. tmake_best_largeriisDid not preserve genome length!(tcopyRIRSt_find_best_regionRR,RK( R:torg_1torg_2t new_org_1t new_org_2tcrossover_chancetbest_1trest_1tbest_2trest_2((stest_GAQueens.pyt do_crossover s   /  ic Csÿtt|ƒd|jƒ}t|ƒ|}|rDt||ƒ}nt||ƒ}dddg}x^tt|ƒ|ƒD]F}|j||||!ƒ}||dkry||||g}qyqyW||d|d!} |d|d!||d} | | fS(sŸFind the best region in the given genome. Arguments: o genome -- A MutableSeq object specifying the genome of an organism o make_best_larger -- A flag to determine whether the best region we should search for should be the larger region of the split caused by crossover or the smaller region. This makes it easy to split two genomes, recombine them, and get a solution that makes sense. Returns: o Two MutableSeq objects. They are both half of the size of the passed genome. The first is the highest fitness region of the genome and the second is the rest of the genome. iiiÿÿÿÿi(tmaxR,RUtminR-RT( R:RRWt first_regiont second_regiont region_sizet best_fitnesst start_indextregion_fitnesst best_regiont rest_region((stest_GAQueens.pyRY"s (R=R>RRR<RcRY(((stest_GAQueens.pyRís    R cBs#eZdZdd„Zd„ZRS(sWMutation operation to help in the N-Queens problem. This performs mutation, but instead of randomly mutating a single item to any other, it tries to mutate it to a row that is not already taken at some other position in the genome. This thus tries to generate more 'correct' mutations that will help achieve the solution. gü©ñÒMbP?cCs ||_dS(sšInititialize a mutator. Arguments: o mutation_rate -- The change of a mutation happening at any position in the genome. N(t_mutation_rate(R:R((stest_GAQueens.pyR<ZscCså|jƒ}|jjj}xÃtt|jƒƒD]¬}tjƒ}||jkr1tj|jjjƒ}x-|jD]"}||krt|j|ƒqtqtWt|ƒdkr¾|jjj}ntj |ƒ}||j|RRR<Rv(((stest_GAQueens.pyR Rs it__main__iisUsage:s4python test_GAQueens.py s:where is an optional parameters7specifying how many queens you want to try to calculates5this for. The default number of queens to place is 5.((((#RRtsystmathRIRXRtBioRtBio.GA.EvolverRtBio.GARtBio.GA.Mutation.SimpleRtBio.GA.Crossover.PointRtBio.GA.Selection.RouletteWheelRtBio.GA.Selection.TournamentRRR'RRR R RRR RR=R,targvR2texit(((stest_GAQueens.pytsB      &   % Oe/