Cookies help us deliver our services. By using our services, you agree to our use of cookies. Learn more

crowdAI is shutting down - please read our blog post for more information

Train Schedule Optimisation Challenge

Optimizing train schedules


multiple solutions

Posted by Raluca almost 2 years ago

What would be the best way to manage the problem of the time limit when the solution is rather sensitive (high objective in let’s say 12 min and a low objective in around 17-20 min on our local computers)? Could we submit both solutions in case one won’t be allowed due to the time limit? It is very hard to tell how much faster the 32 cores 256 GB computer will be.

Thank you

Posted by jordiju  almost 2 years ago |  Quote

Hi @Raluca

As mentioned in the other discussion, the best way is to submit the best solution you have been able to obtain in 15 minutes.

We will ask participants to show us the solver “live” before deploying it ourselves. So we would expect that participants are able to reproduce the solutions they submitted.

Also, there is probably no guarantee that 32 cores will be at all faster than, say, 4 or 8 cores. Depending on seeds and other random effects, it could be even slower. Other people have mentioned that the results from a 32-core AWS-machine was actually worse than from the local laptop.

Last but not least, our cloud provider does not specify what kind of physical CPU they are actually using. We can configure a pod with “32’000 millicores”, but we don’t know what is actually behind this.

Meaning: Don’t rely on us to be able to run your algorithms faster than you.

Posted by hur  almost 2 years ago |  Quote

We will ask participants to show us the solver “live” before deploying it ourselves.<

Will you organise a video call with screen sharing? Could you please explain how will be the live demonstration?

Posted by jordiju  almost 2 years ago |  Quote

Yes, video call. Will post info soon.