Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
Stat methodes for Susy search Daniel August Stricker-Shaver Institut für Experimentelle Kernphysik, Uni Karlsruhe 24/05/2007 Genetic algorithm Is a algorithm that finds a solition for a non analytic problem How do they do this? They chance possible solutions and combin them with each other until they solf the problem satisfying Genetic algorithm 1 2 3 4 5 6 Initialization Evaluation Selection Recombination Mutation building new Genaration with 4 and 5 and going to step 2 until termination condition is reached 1. Initialization Many individual solutions are randomly generated to form a population It depends on the nature of the problem. Typically it contain serveral 100 or 1000 of possible solutions 2. Evaluation Every individual solution is measured by a fitness function and will get a value 3. Selection Individual solutions are selected where fitter solutions are typically more likely to be selected 4. Recombination The genomes of diffrenr individuals are getting mixed and a new generation will be produced 5. Mutation Randomly changing some parts of some individuals of the new generation 6. building new Genaration with 4 and 5 and going to step 2 until termination condition is reached termination condition A solution is found The fixed number of generations is reached The highest ranking solution‘s fitness is reached Genetic algorithm Advantage: GA are the fastest evolutionary algorithms Disadvantages: You never know if is the Optimium of the fitness function Genetic algorithm Example: finding the Minimum Start with maybe 50 individual and a radom for every Genom from -50 to 50 Recombination: G0 = (18, − 3,5,9,8) and G1 = (14,13,33,2,15) => Gc = (18, − 3,33,2,15) Mutation with a posibility of maybe 1% for every change of generation and position m = (1,0,-1) of position =( a, b,c,d,e) end: termination condition has been reached Results: (4,4,4,4,4) or ( − 21, − 21, − 21, − 21, − 21) TMVA Cut optimation fitness function: quality of a retangular cut is given by good background rejection combiened with signal efficiency Decision trees What are boosted decision trees?