domm99/experiments-2026-FGCS-dt-aggregates-for-adaptive-MLOps: 1.0.0

1.0.0 (2026-04-27)

Features

• add common code for training (4aca0a9)
• add deactivate method (4483ffb)
• add empty methods to be implemented in dt (599195d)
• add getter for active local dts (ecdb7d1)
• add handler for training (a91fe65)
• add last training time in inference event payload (c5419e1)
• add learning config file (be51454)
• add minimal main with patients activation scheduling (cfbf73b)
• add patient deactivation event scheduling (0049063)
• add scheduling of training and inferences for experiment TrainAfterTime (1a514a9)
• add seed to dt aggregate (6aa0ef8)
• add simulator start (b40f2df)
• add support for monitors (409ddd3)
• add training metrics export to csv (99f916c)
• better charts (0ae7c0b)
• DTAggregate: implement new data notification and train (a437ee6)
• export F1score, precision and recall (326e7e8)
• export more metrics on time when using adaptive policy (c0ac54e)
• implement DT aggregate skeleton (2fda026)
• implement DT skeleton (e23d63a)
• implement handler for inference on local dts (b79ca3f)
• implement handler for patient deactivation (7dd78a5)
• implement handler for training (eb4f631)
• implement inference for test on local DTs (5bd01aa)
• implement labelling (a52da1f)
• implement learning utils (58e935c)
• implement method to notify new model to local dts (1b9aefa)
• implement methods to register/unregister local dts (90f38f6)
• implement monitor for retraining on enough dts activated (f32a981)
• implement monitor for retraining on enough dts activated (87f76f4)
• implement monitor for retraining on performance degradation (e7544d9)
• implement notification of patient series to dt aggregate (c0f99fc)
• implement script to split dataset by patient (75f4984)
• implement simulator skeleton (8d0ec7f)
• implement timeseries forecasting on all data (a1fba5e)
• init dt aggregate for each dt (2385091)
• more learning configs (3ed8fc9)
• now using classification (7bb5ce3)
• passing last traing time to dt (417c91c)
• setup simulator to be a DES (09e7224)

Dependency updates

• deps: add codecarbon (f74335a)
• deps: add matplotlib (7a5a5e4)
• deps: add pandas (3deb2f6)
• deps: add seaborn (d4405d8)
• deps: add torch (0c755d5)

Bug Fixes

• add data export folder creation in main (920c292)
• adding timestamp column to DT data (a650857)
• comment out redudant monitor (aeb8a46)
• exporting test into one global csv (e158343)
• fix bootstrap months parameter (a634cb0)
• fix data updating from DT to DTAggregate (c19cf4f)
• fix events name (fe78c5b)
• fix inference scheduling (e3b1a86)
• fix last train time (087a96e)
• fix model setter to load fresh model from new weights (9c5fcc6)
• fix model setting (0b68bd6)
• fix patient_series type (1ee7876)
• fix test event timing (27dda2b)
• fix test set start and end indexes (2d01a7e)
• fix test set when there are not enough points (5580028)
• fix time horizon of the prediction and time interval for plotting (b8e6559)
• getting current mean and std for patients that becomes active before retraining (9018cfe)
• import create loaders function (4c5e3ac)
• init first model in dtaggregate as a LSTM and not None (aaac46f)
• init optimizer after the model (e24f117)
• moving also the model to the right device (6ec458e)
• moving also the model to the right device (2e9e115)
• passing also the seed to DT (74c302b)
• remove DTAggregate type to avoid circular dependencies (b67476e)
• remove useless init dts code (8510f23)
• remove val ratio (e11b1fe)
• removing deploy on dockerhub (a0d4531)
• safely ignoring patients with not enough data (5b05187)
• update project name (33a238f)
• using model state_dict and not model itself (fd2ca69)
• using normalized series also in testing (b6462f2)
• using weighted loss to handle class imbalance (b363e27)

General maintenance

• adding log prints (5d693e4)
• remove useless todo comment (c8dd3e0)
• updating some hyperparameters (a8e6d97)