Scenario Discovery with Active Learning: Unterschied zwischen den Versionen

Aus SDQ-Institutsseminar
Keine Bearbeitungszusammenfassung
Keine Bearbeitungszusammenfassung
Zeile 5: Zeile 5:
|betreuer=Vadim Arzamasov
|betreuer=Vadim Arzamasov
|termin=Institutsseminar/2020-05-08
|termin=Institutsseminar/2020-05-08
|kurzfassung=PRIM (Patient Rule Induction Method) is an algorithm used for discovering scenarios, by creating hyperboxes, that can be human-comprehensible. Yet PRIM alone usually requires relatively large datasets and computational simulations can be quite expensive. Consequently, one wants to obtain scenarios while reducing the number of simulations. It has been shown, that combining PRIM with ML models, can reduce the number of necessary simulation runs by around 75%.
|kurzfassung=PRIM (Patient Rule Induction Method) is an algorithm used for discovering scenarios, by creating hyperboxes in the input space. Yet PRIM alone usually requires large datasets and computational simulations can be expensive. Consequently, one wants to obtain scenarios while reducing the number of simulations. It has been shown, that combining PRIM with ML models, can reduce the number of necessary simulation runs by around 75%.
In this thesis, I analyze nine different  Active Learning (AL) sampling strategies together with several ML models, in order to find out if AL can systematically improve PRIM even further, and if a most beneficial combination of sampling method and intermediate ML model exists for this purpose.
In this thesis, I analyze nine different  Active Learning (AL) sampling strategies together with several ML models, in order to find out if AL can systematically improve PRIM even further, and if out of those strategies and models, a most beneficial combination of sampling method and intermediate ML model exists for this purpose.
}}
}}

Version vom 5. Mai 2020, 22:43 Uhr

Vortragende(r) Emmanouil Emmanouilidis
Vortragstyp Bachelorarbeit
Betreuer(in) Vadim Arzamasov
Termin Fr 8. Mai 2020
Vortragsmodus
Kurzfassung PRIM (Patient Rule Induction Method) is an algorithm used for discovering scenarios, by creating hyperboxes in the input space. Yet PRIM alone usually requires large datasets and computational simulations can be expensive. Consequently, one wants to obtain scenarios while reducing the number of simulations. It has been shown, that combining PRIM with ML models, can reduce the number of necessary simulation runs by around 75%.

In this thesis, I analyze nine different Active Learning (AL) sampling strategies together with several ML models, in order to find out if AL can systematically improve PRIM even further, and if out of those strategies and models, a most beneficial combination of sampling method and intermediate ML model exists for this purpose.