Learning by Examples Techniques, Neural Networks and similar applied to the following problem:
there are n jobs to be processed through m machines in a described order. The job processing on a machine requires a duration known as the processing time. Minimizing the makespan ( Cmax ) may be the objective function.
Some more complicated modifications of this problem is also interested for me.
Dear @Bozena, here is the link with many articles about the issue of machine learning methods applied to solving Job Shop Scheduling Problem. Good luck with your research.
https://scholar.google.com/scholar?q=machine+learning+methods+applied+to+solving+Job+Shop+Scheduling+Problem&hl=en&as_sdt=0&as_vis=1&oi=scholart&sa=X&ved=0ahUKEwiWyunlia7UAhXBWBoKHdnyA-YQgQMILDAA
As suggested by @Patricia, here are some articles that cover hyper-heuristic application!
A Genetic Based Hyper-Heuristic Algorithm for the Job Shop Scheduling Problem
Hyper-Heuristic Approach For Solving Scheduling Problem: A Case Study
http://ieeexplore.ieee.org/document/7334675/
http://www.academia.edu/8254384/Hyper-Heuristic_Approach_For_Solving_Scheduling_Problem_A_Case_Study
Dear ma'am ,definitely the hyper-heuristic approach is very useful in this case..or you can use another techniques of scheduling(generally used in production and operation management) for N job and M machines..it depends on your research area..for scheduling problems genetic algorithms are quite ok
Thank you all.
I am starting reading on hybrid genetic algorithms.
Do you know anything on learning by examples in this context (scheduling problem)?
Hi,
Combining Tabu Search and Genetic Algorithm in a Multi-agent System for Solving Flexible Job Shop Problem (2012)
Diversifying TS using GA in multi-agent system for solving Flexible Job Shop Problem
Sign In or Purchase(2015)
Flexible Job-shop Scheduling Problem with Sequence-dependent Setup Times using
Genetic Algorithm (2016)
Good luck
It's true that the problem is not typically a "machine learning" problem, but an optimisation problem.
Even so, the search space is so large (non-trivial instances are NP-complete), so it's a good idea to use machine-learning methods to learn more about the search space, and direct the search to find good-quality solutions.
Personally, I would try a dynamic programming approach with a value or policy function approximation -- intuitively, I would imagine something like Q-learning or SARSA might give good results on this problem.
As a starting point for this particular type of approach, you can check out the work being done at CASTLE Lab, Princeton (http://castlelab.princeton.edu/).