Release v0.7.0

[basnijholt]

(original issue on GitLab)

opened by Bas Nijholt (@basnijholt) at 2018-11-30T11:08:04.471Z

Since 0.6.0 we fixed the following issues:

• gitlab:#122: Remove public 'fname' learner attribute
• gitlab:#119: (Learner1D) add possibility to use the direct neighbors in the loss
• gitlab:#114: (LearnerND) allow any convex hull as domain
• gitlab:#121: How to handle NaN?
• gitlab:#107: Make BaseRunner an abstract base class
• gitlab:#112: (LearnerND) add iso-surface plot feature
• gitlab:#56: Improve plotting for learners
• gitlab:#118: widgets don't show up on adaptive.readthedocs.io
• gitlab:#91: Set up documentation
• gitlab:#62: AverageLearner math domain error
• gitlab:#113: make BalancingLearner work with the live_plot
• gitlab:#111: (LearnerND) flat simplices are sometimes added on the surface of the triangulation
• gitlab:#103: (BalancingLearner) make new balancinglearner that looks at the total loss rather than loss improvement
• gitlab:#110: LearnerND triangulation incomplete
• gitlab:#127: Typo in documentation foradaptive.learner.learner2D.uniform_loss(ip)
• gitlab:#126: (Learner1D) improve time complexity
• gitlab:#104: Learner1D could in some situations return -inf as loss improvement, which would make balancinglearner never choose to improve
• gitlab:#128: (LearnerND) fix plotting of scaled domains
• gitlab:#78: (LearnerND) scale y-values

and merged the following Merge Requests:

• gitlab:!131: Resolve "(Learner1D) add possibility to use the direct neighbors in the loss"
• gitlab:!137: adhere to PEP008 by using absolute imports
• gitlab:!135: test all the different loss functions in each test
• gitlab:!133: make 'fname' a parameter to 'save' and 'load' only
• gitlab:!136: build the Dockerimage used in CI
• gitlab:!134: change resolution_loss to a factory function
• gitlab:!118: add 'save' and 'load' to the learners and periodic saving to the Runner
• gitlab:!127: Resolve "(LearnerND) allow any convex hull as domain"
• gitlab:!130: save execution time on futures inside runners
• gitlab:!111: Resolve "Make BaseRunner an abstract base class"
• gitlab:!124: Resolve "(LearnerND) add iso-surface plot feature"
• gitlab:!108: exponentially decay message frequency in live_info
• gitlab:!129: add tutorials
• gitlab:!120: add documentation
• gitlab:!125: update to the latest miniver
• gitlab:!126: add check_whitespace
• gitlab:!123: add an option to plot a HoloMap with the BalancingLearner
• gitlab:!122: implement 'npoints' strategy for the 'BalancingLearner'
• gitlab:!119: (learnerND) no more (almost) flat simplices in the triangulation
• gitlab:!109: make a BalancingLearner strategy that compares the total loss rather than loss improvement
• gitlab:!117: Cache loss and display it in the live_info widget
• gitlab:!121: 2D: add loss that minimizes the area of the triangle in 3D
• gitlab:!139: Resolve "(Learner1D) improve time complexity"
• gitlab:!140: Resolve "(LearnerND) fix plotting of scaled domains"
• gitlab:!128: LearnerND scale output values before computing loss

When there is v0.7.0 we can set up the documentation for the latest release and master.

[basnijholt]

originally posted by Anton Akhmerov (@anton-akhmerov) at 2018-11-30T11:20:25.291Z on GitLab

A quick question before the release: are we consistent in naming XXX_loss and XXX_loss_function everywhere now?

[basnijholt]

originally posted by Anton Akhmerov (@anton-akhmerov) at 2018-11-30T11:20:54.958Z on GitLab

Other than that I'm on board with releasing.

[basnijholt]

originally posted by Bas Nijholt (@basnijholt) at 2018-11-30T11:21:58.025Z on GitLab

A quick question before the release: are we consistent in naming XXX_loss and XXX_loss_function everywhere now?

Yes, we are doing that :)

[basnijholt]

originally posted by Jorn Hoofwijk (@Jorn) at 2018-12-03T09:02:05.528Z on GitLab

I think indeed it would be a good time to do a new release, however I would suggest to also include https://gitlab.kwant-project.org/qt/adaptive/merge_requests/139. Since the introduction of the round in the loss computation, I noticed some (mayor) slowdown in the Learner1D, since the Python round() function appears to be quite a slow method :/. More about that in the MR