About
The DNNLikelihood framework is presented and its main features discussed. Such framework encodes the experimental likelihood functions in deep neural networks and allows for a lightweight and platform-independent distribution of physics results through the ONNX model format. The procedure retains the full experimental information and does not rely neither on Gaussian approximation nor on dimensionality reduction and is applicable to both binned and un-binned likelihood functions. The distributed DNNLikelihood could be adopted for various use cases, such as re-sampling through Markov Chain MC techniques, combinations with other likelihood functions with proper treatments of correlations and re-interpretation with different statistical approaches. The presentation will be followed by a hands-on tutorial for illustrating the whole procedure with a pseudo-experiment corresponding to a realist LHC search for new physics..
0:00:00 Slide 1
0:03:26 Slide 2
0:06:52 Slide 3
0:10:17 Slide 4
0:13:43 Slide 5
0:17:08 Slide 6
0:20:34 Slide 7
0:23:59 Slide 8
0:27:25 Slide 9
0:30:50 Slide 10
0:34:16 Slide 11
The talk
Credits and licence
- Institution
- CERN
- Department
- LPCC Workshops
- Speaker
- Andrea Coccaro
- Year
- 2021
- Language
- English
- Subjects
- Physics
- Source
- CERN
- Licence
- CERN, as CERN states it. CERN's terms
- Attribution
- © CERN. CERN provides its audiovisual media free of charge for educational and informational use, credited to CERN; this lecture is shown in CERN's own player.
- Who can watch
- CERN gives no rating. Invisible Network shows it to Tweens profiles and older, judged from what the source says about it. How courses are placed