Constraining Effective Field Theories with Machine Learning

We present powerful new analysis techniques to constrain effective field theories at the LHC. By leveraging the structure of particle physics processes, we extract extra information from Monte Carlo simulations, which can be used to train neural network models that estimate the likelihood ratio. These methods scale well to processes with many observables and theory parameters, do not require any approximations of the parton shower or detector response, and can be evaluated in microseconds. We show that they allow us to put significantly stronger bounds on dimension-six operators than existing methods, demonstrating their potential to improve the precision of the LHC legacy constraints.

Figure 1

Schematic overview of the techniques presented in this Letter. (Parts of the figure are based on Ref. [31] and on an image created by Frank Krauss.)

Figure 2

Left: Estimated expected likelihood ratio based on a traditional doubly differential histogram analysis (orange dotted) and the new rascal technique (red dashed). We show a line in parameter space with particularly large difference between the methods. The gray dotted line marks the expected exclusion limit at 95% C.L. according to asymptotics. The vertical arrow shows how much more data the histogram approach requires to constrain the same parameter point with the same significance. The horizontal arrow demonstrates the increased physics reach of the machine-learning-based method. Right: Expected exclusion contours at 68% C.L. (innermost lines), 95% C.L., and 99.7% C.L. (outermost lines) based on the Neyman construction. In both panels, we assume 36 observed events and the SM to be true.