Skip to content
GitLab
Projects Groups Topics Snippets
  • /
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Register
  • Sign in
  • adaptive adaptive
  • Project information
    • Project information
    • Activity
    • Labels
    • Members
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributor statistics
    • Graph
    • Compare revisions
  • Issues 36
    • Issues 36
    • List
    • Boards
    • Service Desk
    • Milestones
  • Merge requests 5
    • Merge requests 5
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Deployments
    • Deployments
    • Environments
    • Releases
  • Packages and registries
    • Packages and registries
    • Container Registry
  • Monitor
    • Monitor
    • Incidents
  • Analytics
    • Analytics
    • Value stream
    • CI/CD
    • Repository
  • Wiki
    • Wiki
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • Quantum TinkererQuantum Tinkerer
  • adaptiveadaptive
  • Merge requests
  • !68

Add LearnerND that does not interpolate the values of pending points

  • Review changes

  • Download
  • Email patches
  • Plain diff
Merged Jorn Hoofwijk requested to merge 52--dont-interpolate-in-2D into master May 27, 2018
  • Overview 233
  • Commits 129
  • Pipelines 65
  • Changes 7

it has some basic functionality now, there are still some things I am going to fix before we can actually use it:

TODO's:

  • add LearnerND
  • make a generic loss function that zooms in on deviations
  • make a better generic loss function (separate issue)
  • allow asking multiple points in one go
    • take pending points into account when choosing a new point
  • come up with a better choose_point_in_simplex (separate issue)
  • loss_per_simplex should only take a single simplex as input, not the entire triangulation
  • make Learner2D a child class of LearnerND (like a special case of the general case)
  • Performance
    • do not recompute triangulation so often
    • do not recompute losses for simplices of which you already know the loss
  • Triangulation: flip edges when it's beneficial
  • add tests for the LearnerND
  • add plot_slice functionality for 2D plots as well
  • add documentation (both user documentation as well as developer documentation) [currently docstrings and comments]
  • add 3d surface plot (separate issue)
  • add 3d crossection plot
  • scale Y-values to unit distance as well
  • allow for N-dimensional output (partially resolved, see #79 (closed))

Closes #52 (closed) and adds a LearnerND

Edited Jul 16, 2018 by Jorn Hoofwijk
Assignee
Assign to
Reviewers
Request review from
Time tracking
Source branch: 52--dont-interpolate-in-2D