mpvasilis / activeconlearn Goto Github PK

This repository contains the code for the active constraint acquisition systems described in our CP2023 paper "Guided Bottom-up Interactive Constraint Acquisition".

You can cite as follows: "Tsouros, D. C., Berden, S., & Guns, T. (2023). Guided Bottom-up Interactive Constraint Acquisition. In: 29th International Conference on Principles and Practice of Constraint Programming, CP 2023, Toronto, Canada, 2023

@inproceedings{tsouros2023guided, title={Guided Bottom-up Interactive Constraint Acquisition}, author={Tsouros, Dimosthenis C and Berden, Senne and Guns, Tias}, booktitle={29th International Conference on Principles and Practice of Constraint Programming, CP 2023, Toronto, Canada}, year={2023}, }

main.py [-h] -a {quacq,mquacq,mquacq2,mquacq2-a,growacq} [-ia {quacq,mquacq,mquacq2,mquacq2-a}] [-qg {baseline,base,tqgen,pqgen}] [-obj {max,sol,p,prob,proba,class}] [-fs {1,2,3}] [-fc {1,2}] [-gqg] [-gfs] [-gfc] [-c {counts,random_forest,MLP,GaussianNB,CategoricalNB,SVM}] [-hls HIDDEN_LAYERS [HIDDEN_LAYERS ...]] [-lr LEARNING_RATE] [-t TIME_LIMIT] -b {9sudoku,4sudoku,jsudoku,random122,random495,new_random,golomb8,murder,job_shop_scheduling,exam_timetabling,exam_timetabling_simple,exam_timetabling_adv,exam_timetabling_advanced,nurse_rostering,nurse_rostering_simple,nurse_rostering_advanced,nurse_rostering_adv} [-nj NUM_JOBS] [-nm NUM_MACHINES] [-hor HORIZON] [-s SEED] [-nspd NUM_SHIFTS_PER_DAY] [-ndfs NUM_DAYS_FOR_SCHEDULE] [-nn NUM_NURSES] [-nps NURSES_PER_SHIFT] [-ns NUM_SEMESTERS] [-ncps NUM_COURSES_PER_SEMESTER] [-nr NUM_ROOMS] [-ntpd NUM_TIMESLOTS_PER_DAY] [-ndfe NUM_DAYS_FOR_EXAMS] [-np NUM_PROFESSORS]

-h, --help show this help message and exit

-a {quacq,mquacq,mquacq2,mquacq2-a,growacq}, --algorithm {quacq,mquacq,mquacq2,mquacq2-a,growacq} The name of the algorithm to use

-ia {quacq,mquacq,mquacq2,mquacq2-a}, --inner-algorithm {quacq,mquacq,mquacq2,mquacq2-a} Only relevant when the chosen algorithm is GrowAcq - the name of the inner algorithm to use

-qg {baseline,base,tqgen,pqgen}, --query-generation {baseline,base,tqgen,pqgen} The version of the query generation method to use

-obj {max,sol,p,prob,proba,class}, --objective {max,sol,p,prob,proba,class} The objective function used in query generation

-fs {1,2,3}, --findscope {1,2,3} The version of the findscope method to use

-fc {1,2}, --findc {1,2} The version of the findc method to use

-gqg, --guide-qgen Use this to guide query generation

-gfs, --guide-findscope Use this to guide FindScope

-gfc, --guide-findc Use this to guide FindC

-c {counts,random_forest,MLP,GaussianNB,CategoricalNB,SVM}, --classifier {counts,random_forest,MLP,GaussianNB,CategoricalNB,SVM} Only relevant when the chosen query generation method is 4 - the machine learning classifier to use

-hls HIDDEN_LAYERS [HIDDEN_LAYERS ...], --hidden-layers HIDDEN_LAYERS [HIDDEN_LAYERS ...] Only relevant when the chosen query generation method is 4 and the chosen classifier is MLP - the hidden layer shapes, e.g. -hls 16 32 64 creates 3 hidden layers of sizes 16, 32 and 64, respectively

-lr LEARNING_RATE, --learning-rate LEARNING_RATE Only relevant when the chosen query generation method is 4 and the chosen classifier is MLP - the learning rate

-t TIME_LIMIT, --time-limit TIME_LIMIT An optional time limit

-b {9sudoku,4sudoku,jsudoku,random122,random495,new_random,golomb8,murder,job_shop_scheduling,exam_timetabling,exam_timetabling_simple,exam_timetabling_adv,exam_timetabling_advanced,nurse_rostering,nurse_rostering_simple,nurse_rostering_advanced,nurse_rostering_adv}, --benchmark {9sudoku,4sudoku,jsudoku,random122,random495,new_random,golomb8,murder,job_shop_scheduling,exam_timetabling,exam_timetabling_simple,exam_timetabling_adv,exam_timetabling_advanced,nurse_rostering,nurse_rostering_simple,nurse_rostering_advanced,nurse_rostering_adv} The name of the benchmark to use

-nj NUM_JOBS, --num-jobs NUM_JOBS Only relevant when the chosen benchmark is job-shop scheduling - the number of jobs

-nm NUM_MACHINES, --num-machines NUM_MACHINES Only relevant when the chosen benchmark is job-shop scheduling - the number of machines

-hor HORIZON, --horizon HORIZON Only relevant when the chosen benchmark is job-shop scheduling - the horizon

-s SEED, --seed SEED Only relevant when the chosen benchmark is job-shop scheduling - the seed

-nspd NUM_SHIFTS_PER_DAY, --num-shifts-per-day NUM_SHIFTS_PER_DAY Only relevant when the chosen benchmark is nurse rostering - the number of shifts per day

-ndfs NUM_DAYS_FOR_SCHEDULE, --num-days-for-schedule NUM_DAYS_FOR_SCHEDULE Only relevant when the chosen benchmark is nurse rostering - the number of days for the schedule

-nn NUM_NURSES, --num-nurses NUM_NURSES Only relevant when the chosen benchmark is nurse rostering - the number of nurses

-nps NURSES_PER_SHIFT, --nurses-per-shift NURSES_PER_SHIFT Only relevant when the chosen benchmark is nurse rostering (advanced) - the number of nurses per shift

-ns NUM_SEMESTERS, --num-semesters NUM_SEMESTERS Only relevant when the chosen benchmark is exam timetabling - the number of semesters

-ncps NUM_COURSES_PER_SEMESTER, --num-courses-per-semester NUM_COURSES_PER_SEMESTER Only relevant when the chosen benchmark is exam timetabling - the number of courses per semester

-nr NUM_ROOMS, --num-rooms NUM_ROOMS Only relevant when the chosen benchmark is exam timetabling - the number of rooms

-ntpd NUM_TIMESLOTS_PER_DAY, --num-timeslots-per-day NUM_TIMESLOTS_PER_DAY Only relevant when the chosen benchmark is exam timetabling - the number of timeslots per day

-ndfe NUM_DAYS_FOR_EXAMS, --num-days-for-exams NUM_DAYS_FOR_EXAMS Only relevant when the chosen benchmark is exam timetabling - the number of days for exams

-np NUM_PROFESSORS, --num-professors NUM_PROFESSORS Only relevant when the chosen benchmark is exam timetabling - the number of professors

###[Q1] Table 2

python main.py -a mquacq2-a -b jsudoku -qg baseline

python main.py -a mquacq2-a -b random495 -qg baseline

python main.py -a mquacq2-a -b murder -qg baseline

python main.py -a mquacq2-a -b golomb8 -qg baseline

python main.py -a mquacq2-a -b "job_shop_scheduling" -nj "10" -nm "3" -hor "15" -s "0" -qg baseline

python main.py -a mquacq2-a -b jsudoku -qg tqgen -t 2

python main.py -a mquacq2-a -b random495 -qg tqgen -t 2

python main.py -a mquacq2-a -b murder -qg tqgen -t 2

python main.py -a mquacq2-a -b golomb8 -qg tqgen -t 2

python main.py -a mquacq2-a -b "job_shop_scheduling" -nj "10" -nm "3" -hor "15" -s "0" -qg tqgen -t 2

python main.py -a mquacq2-a -b jsudoku -qg pqgen

python main.py -a mquacq2-a -b random495 -qg pqgen

python main.py -a mquacq2-a -b murder -qg pqgen

python main.py -a mquacq2-a -b golomb8 -qg pqgen

python main.py -a mquacq2-a -b "job_shop_scheduling" -nj "10" -nm "3" -hor "15" -s "0" -qg pqgen

###[Q2] Table 3: GrowAcq + MQuAcq-2

python main.py -a growacq -ia mquacq2-a -b jsudoku -qg pqgen

python main.py -a growacq -ia mquacq2-a -b random495 -qg pqgen

python main.py -a growacq -ia mquacq2-a -b murder -qg pqgen

python main.py -a growacq -ia mquacq2-a -b golomb8 -qg pqgen

python main.py -a growacq -ia mquacq2-a -b "job_shop_scheduling" -nj "10" -nm "3" -hor "15" -s "0" -qg pqgen

###[Q3 - Q4] Table 3: GrowAcq + MQuAcq-2 guided

python main.py -a growacq -ia mquacq2-a -b jsudoku -qg pqgen -obj proba

python main.py -a growacq -ia mquacq2-a -b random495 -qg pqgen -obj proba

python main.py -a growacq -ia mquacq2-a -b murder -qg pqgen -obj proba

python main.py -a growacq -ia mquacq2-a -b golomb8 -qg pqgen -obj proba

python main.py -a growacq -ia mquacq2-a -b "job_shop_scheduling" -nj "10" -nm "3" -hor "15" -s "0" -qg pqgen -obj proba

###[Q5] Table 4: GrowAcq + MQuAcq-2 guided

python main.py -a growacq -ia mquacq2-a -b "job_shop_scheduling" -nj "15" -nm "11" -hor "40" -s "0" -qg pqgen -obj proba

python main.py -a growacq -ia mquacq2-a -b "job_shop_scheduling" -nj "19" -nm "12" -hor "40" -s "0" -qg pqgen -obj proba