@stefan-k Let's write all documentation in Markdown?
- The files in the open_jobshop directory consider the Open Shop Scheduling Problem (OSSP) and use the evolib.
- Test OSSP results against benchmark db
- Takes a long time, Parallel processing? --> @stefank
- Is there a bug in hybrid genetic? Results are poor. @jjbayer
- Take a look at LibGA by Corcoran Wainwright
- What is elitism? Is this a good library? (This is the library the guys in the papers use)
- [OPTIONAL] Try using different algorithms
- Try different algorithm parameters in benchmark_test.jl
- [OPTIONAL] Are there other problems we can solve? E.g. normal jobshop, flow shop, the "CPU-problem", ...
- Does Khuri use mutation in Selfish Gene? See chapter 4 experimental results
-
Add a field "Percentage of problems for which the lower bound was met" to the evaluation
-
Alternative to "load(...)" so we don't get "redefinition of constant" warnings?
-
Maybe use alias Int/Uint instead of Int64?
-
OSSP efficiency
-
DONE dynamic stopping condition for selfish gene algorithm (now only max iterations)
-
grep -rin TODO .