search

dlasecki / QaoaParamsPredictorApi

Apache License 2.0
0 stars 0 forks source link

Training on a larger dataset #1

Open rsln-s opened 3 years ago

rsln-s commented 3 years ago

Hi,

Co-author of 1911.11071 here. Nice work! Glad to see that you are building an openly available API for machine learning prediction of QAOA parameters.

I would suggest incorporating the large-scale datasets in available QAOAKit (https://github.com/QAOAKit/QAOAKit) into the training set for MaxCut. It can potentially significantly improve the performance. Happy to help with any issues you might have using QAOAKit.

dlasecki commented 2 years ago

Hi @rsln-s, Thank you for pointing me to this resource. It was not on my radar!

It looks like the data is focused on the MaxCut problem. Are there plans to extend it to other popular graph problems? In my approach, I more or less followed the methodology from the paper, but I also generated data for Vertex Cover, Graph Partition and Stable Set. More problems are planned, not necessarily based on graphs. I am currently evaluating the prediction quality of Kernel Density Estimation for different values of a bandwidth parameter.

rsln-s commented 2 years ago

The data is indeed focused on MaxCut for now. Note that the dataset is quite large! We are indeed very interested in adding new datasets for other problems (including beyond graphs) and welcome contributions.