Speaker
Wenxing Zhang
Description
We propose a method to improve the efficiency of preselection in Higgs signal searches at CEPC.
For this propose we developed three machine learning algorithms including boosted decision tree algorithm, fully-connected neural networks and convolutional neural networks.
Among all these algorithms, we found the fully-connected neural networks gives the best prediction on Higgs signals.
Using such algorithms, we improve the signal strength of s-tagging events from cut-based result, $\mu_{ss} \sim 100$, to FCNN result $\mu_{ss} \sim 10$.
Primary author
Co-author
Prof.
Nakai Yuichiro