shigeta-technoface/public-jsp-cam

An unofficial implementation of Genetic Algorithms with Cluster Averaging Method for Solving Job-Shop Scheduling Problems by Python DEAP.

public-jsp-cam

An unofficial implementation of Genetic Algorithms with Cluster Averaging Method for Solving Job-Shop Scheduling Problems by Python DEAP.

Overview

This is an unofficial implementation of the paper(in Japanese) and the book(in Japanese) written by Hiromi Hirano.

Requirement

• deap
• line_profiler (if need)
• python3

Usage

Quick Start(docker)

> git clone [email protected]:shigeta-technoface/public-jsp-cam.git
> cd public-jsp-cam
> make build
> make run
> make login
> (login to container )
> python main.py
> exit

Quick Start

> git clone [email protected]:shigeta-technoface/public-jsp-cam.git
> cd public-jsp-cam
> pip install -r requirements.txt
> python main.py

The resulting three log files are in the logs directory.

Start Options

name	description	default value	example
seed	The number of random seed.	0	--seed 0
population	The number of individuals in one population.	100	--population 100
loop	Loop count.	1	--loop 1
processes	The number of worker processes.	os.cpu_count()	--processes 12
logdir	The directory name for log files.	logs	--logdir ./logs
no_mp	Use single processing.	Use multi processing.	--no_mp
no_cam	Use ordinal replacement.	Use CAM replacement.	--no_cam
is_test	Small problem (MT6x6) and 100 generation for development.	MT10x10 and 3000 generation.	--is_test
do_perf	Log the application performance. line_profiler is required.	-	--do_perf

Output

This program outputs three log files.

name	description	file name
root_log	main log	%Y%m%d%H%M%S%f_log.log
report_log	A TSV file that records the best individual for each loop.	%Y%m%d%H%M%S%f_report.dat
detail_log	A TSV file that records the best individual for each generation.	%Y%m%d%H%M%S%f_detail.dat

The following histogram was created using two report_log files. The CAM report_log is from python main.py --loop 300 and the NoCAM report_log from python main.py --loop 300 --no_cam.

The following line charts shows the percentiles of the best makespan for each generation. These were created by using detail_log file. The CAM detail_log is from python main.py --loop 300 and the NoCAM detail_log from python main.py --loop 300 --no_cam.

Author

[email protected]