Title: Machine learning based methods for extracting EFT parameters

Speaker: Dr. Suman Chatterjee, CERN

Abstract: Various extensions of the standard model of particle physics (SM) predict anomalous interactions at the weak scale. Effective field theory (EFT), a generalized extension of the SM, consists of all the possible operators of dimensions greater than four, satisfying the SM’s symmetries [1]. The EFT operators modify the production and decay kinematics of the particles involved in LHC collisions compared with those predicted by the SM. We have recently developed a tree boosting algorithm for collider measurements of multiple EFT-operator coefficients [2] [3]. The design of the discriminant exploits per-event information of the simulated data sets that encodes the predictions for different values of the coefficients. This “Boosted Information Tree” algorithm provides nearly optimal discrimination power order-by-order in the expansion in the EFT-operator coefficients and approaches the optimal likelihood ratio test statistic. Similar ideas have been persued by Chen et al. and Ambrosio et al. [4] [5], where the authors use neural networks. In this talk, I will discuss the algorithms and show their applications to different physics processes at LHC.

References:

[1] B. Grzadkowski, M. Iskrzynski, M. Misiak and J. Rosiek, Dimension-Six Terms in the Standard Model Lagrangian, JHEP10 (2010) 085, arXiv:1008.4884.

[2] S. Chatterjee, N. Frohner, L. Lechner, R. Schoefbeck, D. Schwarz, Tree boosting for learning EFT parameters, Comput. Phys. Commun. 277 (2022) 108385, arXiv:2107.10859.

[3] S. Chatterjee, S. Roshap, R. Schoefbeck, D. Schwarz, Learning the EFT likelihood with tree boosting, arXiv:2205.12976.

[4] S. Chen, A. Glioti, G. Panico, A. Wulzer, Parametrized classifiers for optimal EFT sensitivity, JHEP 05 (2021) 247.

[5] R. G. Ambrosio, J. Hoeve, M. Madigan, J. Rojo, V. Sanz, Unbinned multivariate observables for global SMEFT analyses from machine learning, arXiv: 2211.02058.