Skip to content

GitLab

  • Menu
Projects Groups Snippets
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Sign in
  • N NIFTy
  • Project information
    • Project information
    • Activity
    • Labels
    • Members
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributors
    • Graph
    • Compare
  • Issues 17
    • Issues 17
    • List
    • Boards
    • Service Desk
    • Milestones
  • Merge requests 12
    • Merge requests 12
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Deployments
    • Deployments
    • Releases
  • Packages & Registries
    • Packages & Registries
    • Container Registry
  • Analytics
    • Analytics
    • Value stream
    • CI/CD
    • Repository
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar

On Thursday, 7th July from 1 to 3 pm there will be a maintenance with a short downtime of GitLab.

  • ift
  • NIFTy
  • Issues
  • #185

Closed
Open
Created Aug 22, 2017 by Martin Reinecke@mtrOwner

Binbounds calculation in PowerSpace needs fixing

Currently, the initialization of a PowerSpace has several problems:

  • when specifying logarithmic=False, the code behaves differently from the code before the PowerSpace reorganization two months ago (this is because I'm now interpreting the parameter logarithmic=False to mean "please use linear binning", instead of "please use natural binning", which was its original meaning).
  • the "magic" formulae for computing nbins and binbounds are buggy; this was not really obvious before the reorganization, because they were not used in most circumstances).
  • the algorithm tries to derive binbounds from insufficient user information (e.g. if the user specifies the number of bins, the code simply guesses the position of the first and last binbounds, and the guess is not necessarily intuitive).

I would strongly favor an approach where the user has to specify a set of parameters that defines the binbounds absolutely unambiguously. E.g.:

  • by providing a binbounds array
  • by providing a type of spacing (linear or logarithmic), a number of bins, and the first and last binbound
  • by providing absolutely nothing, which leads to natural binning

There are very few cases where it could make sense for the code to guess the binbounds; the only situations that I can imagine are 1D RGSpaces and LMSpaces with linear binning. Here, there is an obvious choice for the number of bins and the placing of the binbounds.

Would it be OK if I introduced the necessary additional constraints?

To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information
Assignee
Assign to
Time tracking