Speaker
Description
Monte Carlo simulations provide powerful approximations to experimental data when accurate modeling and parameter tuning are available. However, it is unrealistic to expect that a single parameterization can reproduce the full complexity of the data. A more plausible assumption is that different simulation setups describe different regions of observable space with varying degrees of accuracy.
In this work, I present an application of Latent Dirichlet Allocation (LDA) that exploits information from multiple simulation setups of known processes to learn the shapes of observable distributions directly from data. This approach provides a general framework to infer the latent topic structure underlying the observed data, which permits flexible and data-driven background modeling and uncertainty estimation.
I demonstrate this method in a specific application to multijet final states at collider experiments.
| I read the instructions above | Yes |
|---|