adaptive v0.7.3 Release Notes
Release Date: 2019-01-29 // over 5 years ago-
Implemented enhancements:
- โ Add a sequential executor #138
- โ Add tests for 1D interpolator learner #136
- โ Add integration learner #135
- ๐ Make the runner work with
asyncio
and inside Jupyter notebooks #133 - โ Add module for notebook integration and shortcuts for common executors #132
- โ Add homogeneous sampling learner #131
- โ Add a "balancing" learner #130
- Implement 2D and 3D learners #129
- โ Add a 0D averaging learner #128
- Write
interpolate
for the 1D learner such that it is more efficient #126 - Gracefully handle exceptions when evaluating the function to be learned #125
- ๐ Allow BalancingLearner to return arbitrary number of points from 'choose_points' #124
- 0๏ธโฃ Increase the default refresh rate for 'live_plot' #120
- โ remove default number of points to choose in
choose\_points
#118 - Consider using Gaussian process optimization as a learner #115
- ๐ Make
distributed.Client
work with automatic scaling of the cluster #104 - ๐ Improve plotting for learners #83
- ๐จ (refactor) learner.tell(x, None) might be renamed to learner.tell_pending(x) #73
- (feature) make interactive plots for learnerND plot_slice method #64
- 0๏ธโฃ (LearnerND) make default loss function better #63
- ๐ allow for N-d output #60
- โ add cross-section plot #58
- (BalancingLearner) make new balancinglearner that looks at the total loss rather than loss improvement #36
- (LearnerND) allow any convex hull as domain #25
- (Learner1D) add possibility to use the direct neighbors in the loss #20
๐ Fixed bugs:
- Distinguish actual loss and estimated loss #139
- Set the bounds in a smarter way #127
- some points get cluttered #86
- 2D learner specifies a 1D point causing 2D learner to fail #81
- Method 'Learner.tell' is ambiguous #80
- Learner1D fails with extremely narrow features #78
- AverageLearner math domain error #77
- (LearnerND) scale y-values #61
- Learner1D breaks if right bound is added before the left bound #45
- Learner1D's bound check algo in self.ask doesn't take self.data or self.pending_points #44
- Learner1D fails when function returns a list instead of a numpy.array #43
- Learner1D fails when a point (x, None) is added when x already exists #42
- Learner1D.ask breaks when adding points in some order #41
- Learner1D doesn't correctly set the interpolated loss when a point is added #40
- Learner1D could in some situations return -inf as loss improvement, which would make balancinglearner never choose to improve #35
- โ LearnerND fails for BalancingLearner test #34
- Learner2D suggests same point twice #30
- (LearnerND) if you stop the runner, and then try to continue, it fails. #23
Closed issues:
- โ Add Authors file and review license #137
- ๐ make the runner request points until it's using all cores #123
- Remove _choose_points #121
- ๐ Fix extremely long kernel restart times #119
- live plotting: add a universal visual cue that the goal is achieved. #117
- ipyparallel shouldn't be a dependency #114
- adaptive fails to discover features #113
- โ add tests for 2D learner #111
- DataSaver doesn't work with the BalancingLearner #110
- โ deleted issue #108
- removing optional dependencies #106
- ๐ Improve ipython event loop integration #105
- ๐ Use holoviews.TriMesh when it makes it to a release #103
- ๐พ save live plots into internal datastructure #101
- To-dos before making the repo public #100
- set the correct loss_improvement for the AverageLearner #95
- Ensure a minimum resolution #92
- ๐ change the error message in runner #91
- ๐ The ProcessPoolExecutor doesn't work on Windows #90
- 1D and 2D learner: stop interpolating function instead of the loss #87
- Discontinuities in zero should be detected and be approximated with some margin #85
- (minor bug) learner.choose_points gives wrong number of points in one very particular case #84
- 2D: if boundary point fails it will never be re-evaluated ... #82
- Learner2D + BalancingLearner too slow to use on many cores #79
- BalancingLearner.from_product doesn't work with the DataSaver #74
- ๐ง Follow-up from "WIP: Add LearnerND that does not interpolate the values of pending points" #70
- (triangulation) make method for finding initial simplex part of the triangulation class #68
- ๐จ (refactor) LearnerND._ask can be refactored to be so much more readable #67
- ๐ (LearnerND) make choose point in simplex better #62
- ๐ Make learnerND datastructures immutable where possible #54
- ๐ Rename LearnerND to TriangulatingLearner #51
- tell_many method #49
- ๐ Set up documentation #48
- ๐ DeprecationWarning: sorted_dict.iloc is deprecated. Use SortedDict.keys() instead. #47
- The example given in data_saver.py doesn't compile. #46
- What should learners do when fed the same point twice #39
- How should learners handle data that is outside of the domain #38
- โ No tests for the 'BalancingLearner' #37
- ๐ release 0.6.0 #33
- ๐ Make BaseRunner an abstract base class #32
- (BalancingLearner) loss is cached incorrectly #31
- LearnerND triangulation incomplete #29
- (LearnerND) flat simplices are sometimes added on the surface of the triangulation #28
- (LearnerND) add iso-surface plot feature #27
- ๐ make BalancingLearner work with the live_plot #26
- test_balancing_learner[Learner2D-ring_of_fire-learner_kwargs2] fails sometimes #24
- widgets don't show up on adaptive.readthedocs.io #21
- How to handle NaN? #18
- โ Remove public 'fname' learner attribute #17
- ๐ Release v0.7.0 #14
- (Learner1D) improve time complexity #13
- ๐ Typo in documentation for
adaptive.learner.learner2D.uniform\_loss\(ip\)
#12 - (LearnerND) fix plotting of scaled domains #11
- suggested points lie outside of domain #7
- DEVELOPMENT IS ON GITLAB: https://gitlab.kwant-project.org/qt/adaptive #5
๐ Merged pull requests:
- fix _get_data for the BalancingLearner #150 (basnijholt)