René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Multi Objective Optimization","articleBody":"Hello!\r\n\r\nUsing the \"ParBayesianOptimization\" package, it possible to use this package for \"multi objective optimization\" (e.g. optimize several cost functions together)?\r\n\r\nFor example, below I have included an example of multi-objective optimization using an algorithm called \"particle swarm optimization\":\r\n\r\n```\r\n#Load library:\r\nlibrary(mopsocd)\r\n\r\n#load libraries\r\nlibrary(dplyr)\r\n\r\n\r\n# create some data for this example\r\na1 = rnorm(1000,100,10)\r\nb1 = rnorm(1000,100,10)\r\nc1 = sample.int(1000, 1000, replace = TRUE)\r\ntrain_data = data.frame(a1,b1,c1)\r\n\r\n#define function:\r\n\r\nfunct_set \u003c- function (x) {\r\n    \r\n    \r\n    \r\n    #bin data according to random criteria\r\n    train_data \u003c- train_data %\u003e%\r\n        mutate(cat = ifelse(a1 \u003c= x[1] \u0026 b1 \u003c= x[3], \"a\",\r\n                            ifelse(a1 \u003c= x[2] \u0026 b1 \u003c= x[4], \"b\", \"c\")))\r\n    \r\n    train_data$cat = as.factor(train_data$cat)\r\n    \r\n    #new splits\r\n    a_table = train_data %\u003e%\r\n        filter(cat == \"a\") %\u003e%\r\n        select(a1, b1, c1, cat)\r\n    \r\n    b_table = train_data %\u003e%\r\n        filter(cat == \"b\") %\u003e%\r\n        select(a1, b1, c1, cat)\r\n    \r\n    c_table = train_data %\u003e%\r\n        filter(cat == \"c\") %\u003e%\r\n        select(a1, b1, c1, cat)\r\n    \r\n    \r\n    \r\n    #calculate  quantile (\"quant\") for each bin\r\n    \r\n    table_a = data.frame(a_table%\u003e% group_by(cat) %\u003e%\r\n                             mutate(quant = ifelse(c1 \u003e x[5],1,0 )))\r\n    \r\n    table_b = data.frame(b_table%\u003e% group_by(cat) %\u003e%\r\n                             mutate(quant = ifelse(c1 \u003e x[6],1,0 )))\r\n    \r\n    table_c = data.frame(c_table%\u003e% group_by(cat) %\u003e%\r\n                             mutate(quant = ifelse(c1 \u003e x[7],1,0 )))\r\n    \r\n    f1 = mean(table_a$quant)\r\n    f2 = mean(table_b$quant)\r\n    f3 = mean(table_c$quant)\r\n    \r\n    \r\n    #group all tables\r\n    \r\n    final_table = rbind(table_a, table_b, table_c)\r\n    # calculate the total mean : this is what needs to be optimized\r\n    \r\n    f4 = mean(final_table$quant)\r\n    \r\n    #multiple functions are being optimized \r\n    return (c(f1, f2, f3, f4));\r\n}\r\n\r\n#constraints (I know this is not currently possible in ParBayesianOptimization)\r\n  gn \u003c- function(x) {\r\n    g1 \u003c- x[2] - x[1] \u003e 0.0\r\n    g2 \u003c- x[4] - x[3] \u003e 0.0\r\n    g3 \u003c- x[7] - x[6] \u003e0\r\n    g4\u003c- x[6] - x[5] \u003e0\r\n    return(c(g1,g2,g3, g4))\r\n}\r\n\r\n## Set Arguments/Bounds \r\n\r\nvarcount \u003c- 7\r\nfncount \u003c- 4\r\nlbound \u003c- c(80,90,80,90,100, 200, 300)\r\nubound \u003c- c(90,110,90,110,200, 300, 500)\r\noptmin \u003c- 0\r\n\r\n\r\n\r\n#optimization of multiple cost functions\r\nex1 \u003c- mopsocd(funct_set,gn, varcnt=varcount,fncnt=fncount,\r\n               lowerbound=lbound,upperbound=ubound,opt=optmin)\r\n```\r\n\r\nWould it be possible to solve a similar style problem (i.e. optimization of multiple objective functions) using the ParBayesianOptimization library?\r\n\r\nThanks so much!","author":{"url":"https://github.com/swaheera","@type":"Person","name":"swaheera"},"datePublished":"2021-07-13T16:51:06.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/40/ParBayesianOptimization/issues/40"}