Setting the Accuracy of LO Gridpack Generation

Hi,

You are not generating data, but just the optimal grid.

So even with an optimal grid, I would expect that you have to generate data within X channel
and therefore that you do see that line (which can be missleading on what it means in the context of gridpack).

The precision of the grid is mainly related to the number of separation of the grid in each dimension, since the algorithm identifies the value which separate each axis of integration such that each bin get the same contribution to the integral.

Therefore at some point increasing the requested precision is actually irrelevant since this will only marginally change the position of the grid. In that case, you would need to not only increase the requested precision but also increase the size of each grid by increasing the number of bin by dimension.

This is not an option that we propose since this might not be trivial since this is likely to create issue with the training parameter. (the first iteration might not have enough point to correctly train and this can lead to issue). But if you are interested to study that I can probably give some guidance on the various file to edit.

Cheers,

Olivier

> On 24 Jan 2023, at 16:20, Christopher Young <email address hidden> wrote:
>
> New question #704527 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/704527
>
> Hi,
> When creating gridpacks for leading order Madgraph samples I think the accuracy is set to 0.01 Is it possible to increase this accuracy?
>
> The reason that I want to do this is because when the gridpack is used to create large samples of events it appears that further integration is performed as I see a line like;
> 10:32:39 INFO: need to improve 162 channels
> therefore I would have thought I could save time from this step if my gridpack was created with a higher accuracy?
>
> Many thanks,
> Chris.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.