Reproducibility

[do-me]

I was wondering if reproducibility of results is an active feature or a more of a side-effect of VRP.

I ran some very quick tests with the very simple problems with the WebAssembly version and it seems indeed to provide the same results. I couldn't find anything about it in the docs, so I thought I'd ask here.

Am I correct to assume that:

• as long as the input remains exactly the same and
• any termination criterion hits before the time limit is exceeded

the results should always be the same or is there some kind of random seed somewhere for more complex problems?

[reinterpretcat]

No, algorithm is non-determenistic, so for different runs it can explore solution space differently and return different solutions. Search stability is one of the main topics among with solution quality and good termination criteria.

[do-me]

You're referring to this poll right? #82
This is good to know, thank you!

[reinterpretcat]

Yes, it is mentioned here too.
At the moment, I'm lazily researching how to improve my current algorithm to address these issues rather working on new features.