Comparing Neural Net to Likelihood Techniques in HH to bbWW* Production for the Higgs Self Coupling

The Higgs self-interaction is a yet-to-be-confirmed prediction of the Standard Model (SM), making its discovery a key objective of future runs at the Large Hadron Collider (LHC). This coupling is measured at the LHC by the rate of Higgs boson pair production (HH), which is exceptionally low. So far, analyses have only been able to establish upper limits on the HH production cross section. In this talk, we present recent progress at CMS in improving these constraints by analyzing the HH to bbWW* decay mode, with a specific focus on events containing a final state of one lepton, two b-tagged jets, and two non-b-tagged jets. Our analysis strategy employs an initial event selection based on low-level kinematic variables, construction of higher-level variables to capture the distinctive kinematic signatures of both signal and background processes, choosing a discriminator, and performing a maximum likelihood fit. Two approaches were explored as discriminators: the use of a Log-Likelihood Ratio (LLR) of the kinematic variables and a Deep Neural Network (DNN) multi-classifier. We will present the expected results from both approaches, along with a supplementary study investigating the effectiveness of using LLRs as inputs to the DNN.