0.11.0

What's Changed

• Automate Python package build and upload on Pypi from Github CI by @relf in #104
• Fix FullFactorial when asked nb iof samples is small wrt x dimension by @relf in #105
• Make mixed-integer sampling methods available in Python by @relf in #106

Full Changelog: 0.10.1...0.11.0

0.10.1

• ego
  • Fix bug when best cost is in first iteration it is not selected (#102)
• egobox Python module
• Extend mixed-integer Py wrapper to ordered and enumeration variables (#102)
• Update notebooks (#102)

0.10.0

• gp, moe and egobox Python module:

  • Added Gaussian process sampling (#97)
  • Added string representation (#98)

• egobox Python module:

  • Change recombination enum to respect Python uppercase convention (#98)

• Notebooks and documentation updates (#97, #98, #99)

0.9.0

• ego:

  • Infill criterion is now a trait object in EgorSolver structure (#92)
  • Egor and EgorSolver API: methods taking argument of type Option now take argument of type T (#94)
  • EgorBuilder::min_within_mixed_space() is now EgorBuilder::min_within_mixint_space() (#96)
  • egobox-ego library doc updated (#95)

• egobox Python binding: Upgrade to PyO3 0.18 (#91)

0.8.2

ego : Fix Egor solver best iteration computation (#89)

0.8.1

• ego:
  • Make objective and constraints training in parallel (#86)
  • Lock mopta execution to allow concurrent computations (#84)
  • Fix and adjust infill criterion optimmization retries strategy (#87)
• moe:
  • Fix k-fold cross-validation (#85)

0.8.0

• ego:
  • Renaming XType, XSpec for consistency (#82)
  • Export history in optimization result (#81)
  • Use nb iter instead of nb eval, rename q_parallel as q_points (#79)
  • Warn when inf or nan detected during obj scaling computation (#78)
  • Parallelize constraint scales computations (#73)
  • Parallelize multistart optimizations (#76)
  • Handle GMM errors during MOE training (#75)
  • Handle possible errors from GMM clustering (#74)
  • Upgrade to argmin 0.8.0 (#72)
  • Add mopta08 test case as example (#71)
  • Fix scaling check for infinity (#70)
  • Use kriging surrogate by default (#69)

0.7.0

• gp:
  • Add analytic derivatives computations (#54, #55, #56, #58, #60). All derivatives available for all mean/correlation models are implemented.
  • Refactor MeanModel and CorrelationModel methods:
    • apply() renamed to value()
    • jac() renamed to jacobian()
  • Fix prediction computation when using linear regression (#52)
• ego:
  • Refactor Egor using argmin 0.7.0 solver framework EgorSolver can be used with argmin::Executor and benefit from observers and checkpointing features (#67)
  • Egor use kriging setting by default (i.e. one cluster with constant mean and squared exponential correlation model)
• Add notebook on Manuau Loa CO2 example to show GpMix/Gpx surrogate model usage (#62)
• Use xoshiro instead of isaac random generator (#63)
• Upgrade ndarray 0.15, linfa 0.6.1, PyO3 0.17 (#57, #64)

0.6.0

• gp: Kriging derivatives predictions are implemented (#44, #45), derivatives for Gp with linear regression are implemented (#47)
  • predict_derivatives: prediction of the output derivatives y wtr the input x
  • predict_variance_derivatives: prediction of the derivatives of the output variance wrt the input x
• moe: as above derivatives for smooth and hard predictions are implemented (#46)
• ego: when available derivatives are used to optimize the infill criterion with slsqp (#44)
• egobox Python binding: add GpMix/Gpx in Python egobox module, the Python binding of egobox-moe::Moe (#31)

0.5.0

• Enable Egor optimizer interruption with Ctrl+C (#30)
• API improvements: generalize the use of ArrayBase<...>, better linfa integration by implementing PredictInplace trait (#37)
• Minor performance improvement in mixture of experts clustering (#29)
• Documentation improvements: JOSS paper submission review (#34, #36, #38, #39, #40, #42)