v0.3.0: Supervised Baseline Approach

What's Changed

• fix: pseudo label generation for fttp longer than time series by @tilman151 in #46
• feat: add test step to supervised approach by @tilman151 in #47

Full Changelog: 0.2.0...0.3.0

v0.2.0: More Approaches

What's Changed

• feat: add cnn dann by @tilman151 in #25
• fix: add best model checkpoint to lstm dann by @tilman151 in #26
• refactor: unify supervised pre-training by @tilman151 in #27
• feat: add pseudo label approach by @tilman151 in #28
• feat: add dynamic conditional adaption by @tilman151 in #29
• refactor: decouple eval losses by @tilman151 in #30
• feat: make approach-unspecific hparams configurable by @tilman151 in #31
• feat: make RUL score mode configurable by @tilman151 in #32
• feat: make model hyperparameters loggable by @tilman151 in #33
• refactor: unify construction args for CNN and RNN models by @tilman151 in #34
• feat: degraded-only evaluation mode by @tilman151 in #35
• fix: make pseudo labels ready for inductive adaption by @tilman151 in #36
• fix: additional inductive pseudo label error by @tilman151 in #37
• tests: disable lightning logging in unit tests by @tilman151 in #38
• feat: add dilation to CNN extractor by @tilman151 in #39
• feat: add stride to CNN extractor by @tilman151 in #40
• fix: deactivate label transformation in latent align by default by @tilman151 in #41
• fix: make default max RUL of adarul None but behave as 1 by @tilman151 in #42
• fix/make pseudo labels work for normed rul by @tilman151 in #43
• fix: add safeguards for conditional adaption loss by @tilman151 in #44
• docs: update and include example notebooks by @tilman151 in #45

Full Changelog: 0.1.0...0.2.0

v0.1.0: Initial Release

What's Changed

• feat: add abstract base class for approaches by @tilman151 in #1
• fix: remove manual feature extractor by @tilman151 in #3
• feat: add feature extractor models by @tilman151 in #2
• feat: add head model by @tilman151 in #4
• feat: add losses by @tilman151 in #5
• feat: switch adaption losses over to torchmetrics by @tilman151 in #7
• fix: put dropout layers in right positions for rnns by @tilman151 in #8
• feat: make set_model open for override by @tilman151 in #9
• feat: create domain labels inside dann loss by @tilman151 in #10
• fix: rul score backward by @tilman151 in #11
• fix: make adaption metrics movable between devices by @tilman151 in #12
• fix: diverging devices in adaption metrics by @tilman151 in #13
• feat: add dropout option to fc head by @tilman151 in #15
• fix: remove dropout from cnn input by @tilman151 in #16
• fix: remove cpu only pytorch by @tilman151 in #18
• feat: enable checkpointing by @tilman151 in #19
• feat: add gradient weight to GRL by @tilman151 in #20
• fix: introduce epsilon for pairwise euclidean by @tilman151 in #22
• feat: add lstm dann by @tilman151 in #14
• feat: add consistency dann by @tilman151 in #17
• feat: add adarul by @tilman151 in #21
• feat: add tbigru by @tilman151 in #23
• feat: add latent alignment approach by @tilman151 in #6
• docs: update intro and readme by @tilman151 in #24

New Contributors

• @tilman151 made their first contribution in #1

Full Changelog: https://github.com/tilman151/rul-adapt/commits/0.1.0