[fix] simulated anneal always accepts first iter #153
Conversation
|
Thank you for this PR! This is actually what I also suspected, apologies that I wasn't able to address this myself. For some reason the CI is broken since yesterday. I'm having the same problem in #152. I will try to have a look hopefully this evening. My suspicion is that something in the images changed. |
|
Don't worry, it was not my intention to apply any pressure to you. I just wanted to offer my help with the PR. Should I update the changelog, too? |
|
I did not interpret this as pressure :) Any help is highly appreciated. You're welcome to update the changelog if you want, otherwise I will update it once I prepare the next release. Seems like I was able to fix the CI problems. If you merge the current master into your branch then the compilation of |
|
Thanks for fixing the CI. I have updated the changelog, adding a line about the fix. |
|
Thanks a lot once more for this PR! |
This addresses #150
It appears to me that the initial
ArgminIterDatareturned by the simulated anneal solver'sinitmethod does not contain values forparamandcost. Hence theexecutor's state is not being updated upon calling thesolver'sinitmethod. This leads to the state having infinity as starting cost, which makes any first solution preferable.I have added param and corresponding initial cost to the solver's init method.