- Updates pfp
- NOTE! File Paths in pfp_runner are hardcoded. Error will occure. Change manually
- Top above will be fixed in the next days
- Implemented Simulation Driven Runner
- Old Runner is now Algorithm Driven Runner
- Implemented Use Case pfp (Maurice BA)
- Implemented inequality parameter constraint (sum(w*p) >= c)
- NOTE: TOP 1-3 are not finished
- rounding bug fixed in mrp
- reduced stochastic propability in mrp_sim (got to eval if promising)
- wording change: batch_* -> trial_*
- small change cma-es. Skipps initial now and generates {trial_size} arms directly
- turbo state updates correctly now
- reduced tolerance of turbo to force exploration
- increased penalty cost rate from 0.25 to 0.33 (reason: at 50D best value found at sl of 0.78)
- set surrogate model für turbo and gpei as experiment parameter (gsheet)
- identify best trial fixed
- Bug at CMA-ES fixed. Now finds minimum instead of maximum
- Brute Force with logs
- log every new best point
- MRP: reduced propability of delay and reduction (stochastic)
- MRP: bug fix at releases. Now it works as it should :)
- MRP: added function to get data of every class instance (csv)
- wording: configs -> experiment
- fixed solver and simulation
- configs folder will be created if not exists
- further minor changes
- wording changes at mrp_runner (inventory -> stock, demand -> orders)
- Implemented new BOMs etc. in Gsheet
- Sobol Runner
- Brute Force Runner
- MRP Solver OOP
- MRP Simulation OOP
- acquisition value implementiert
- feature importance hinzugefügt
- Algorithmus Runner Klasse erstellt, aktuelle Runner erben davon
- MRP stochastic_method via Sheets konfigurierbar
- MRP method discrete -> "Tail" hinzugefügt
- changelog file hinzugefügt (offensichtlich)
- laden der Sheets aus der main heraus, dazu als sysarg "load" eintragen
- Habe Funktion geschrieben zum checken der sysargs
- Genereller Aufbau der sysargs: main.py [experiment_id] [opt:replication] [opt:"load"]
- aber auch möglich main.py [opt:"load"] [experiment_id] [opt:replication]
- wenn num_init/n_init -1 dann num_init = 2*dim (definiert im Algorithmus Runner)
- cmaes implementiert (WIP und hartes theoretisches Defizit meinerseits, muss lernen was die Parametrisierungen bedeuten und diese ggf. implementieren...)
- Update Gsheet Felder u_stochastic_method, a_sigma0 (Hyperparameter cma-es)
- algo runner Funktion "get_technical_specs" für experiment json Informationen. sm and acqf im SubRunner definieren.
- *_stepwise files gelöscht
- requirements.txt upgedated (cma)
- Changes einfach chronologisch (neueste oben) eintragen mit Datum und Name (siehe oben)