Details
-
Task
-
Status: Closed
-
Major
-
Resolution: Done
-
None
-
None
-
None
Description
diluca will work on the support and development of the DL1 tagger and training software for release 21 and will work in collaboration with the current DL1 team on the new training software being developed to improve training time and flexibility. He will continue to develop the flexible hyperparameter optimization software, enable random search and potentially bayesian methods, and ensure that such optimizations can be performed on batch systems. Andrea will also perform model comparison tests, and he will help in the training studies to understand and mitigate the effects of pileup. This work will be documented in an ATLAS internal note, twiki, or code repository README, as appropriate.
Local Supervisor: riuppa
Technical Supervisor: cpollard
Start Date: 2019-05-30