Release Release 0.2.0 · pykale/pykale

#281: Add quantile binning for uncertainty estimation including error bound estimation and evaluation metrics
#360: Add multimodal neural network for multimodal data integration with AVMNIST example
#361: Add MOGONET for multiomics integration
#395: Create main.py, cross_validation for multisite_neuroimg_adapt example
#407: Add prototypical network
#437: Add mida_trainer pipeline
#458: Add example DrugBAN PyG
#460: Update cross-validation for transductive domain adaptation
#463: Add symmetric weights and connectome vizualization
#464: Implement AutoMIDAClassificationTrainer and MIDA enhancement
#471: Add multimodal tri-stream pre-training and fine-tuning for image and signal modalities
#475: Add feature importance interpretation to MOGONET example
#476: Allow custom param_grid for AutoMIDAClassificationTrainer
#477: Add Integrated Gradients interpretation method and minor refinements

#423: Update the default checkpoint (ckpt) resume and test file name to None
#444: Pin Torch to 2.0.0 to fix test OSError and update related packages
#445: Pin torch to 2.3.0
#453: Fix device allocation error and resolve type mismatches
#455: Fix device allocation error, resolve type mismatches, and update repo configurations
#456: Support Python 3.11 and stop support Python 3.8
#459: Upgrade minimum sklearn version from 0.23.2 to 1.6.1
#472: Fix auto-unwrapping in the multiomics example
#481: Support torch 2.6
#482: Set neurokit2<=0.2.11 in setup.py
#487: Fix circular import and refactor correlation analysis
#494: Update weight norm imports for DrugBAN
#511: Fix torch-numpy type mismatch

#383: Update keyword param for pretrained ResNet
#401: Upgrade to support PyTorch 2.0+ and Lightning 2.0+
#403: Refine examples via unifying device settings and output folder names
#409: Refactor ResNet
#417: Add resume and test checkpoint loading to main.py in example/cifas_cn…
#426: Refactor documentation and resolve module visibility issues
#429: Cache preload data
#438: Upgrade to support Numpy 2.0.0+
#441: Refactor merge modules losses distance and metrics to one
#449: Fix DANN torch2.0
#467: Unify API for mida_trainer and multi_domain_trainer, and improve mida_trainer test
#469: Refactor DomainNetSmallImage
#480: Make the output of interpretation in cancer more clean
#483: Refactor DrugBAN
#484: Reorganize kale.embed APIs
#492: Refactor uncertainty_quantile with BoxPlotter class
#497: Support Python 3.12
#498: Refactor Jaccard evaluation in uncertainty metrics
#502: Refactor uncertainty quantiles by implementing QuantileBinningAnalyzer
#505: Unify domain adapter loader
#508: Refactor Squeeze-and-Excitation Layers
#509: Stop support Python 3.9
#514: Refactor/embed.BaseCnn address issue #513
#520: Fix #518: [Refactor] Variable naming in CNN modules

Navigation Menu