During this summer, I got to the opportunity to work in Dr. David Caratelli’s neutrino physics lab as a part of the UCSB Physics REU program, funded by the NSF. I had two main projects, both associated with the MicroBooNE detector based at the Fermi National Accelerator Laboratory (FermiLab). I collaborated with people at FermiLab in addition to those in our group at UCSB. My graduate mentor, Michaelia Fang, also supported me throughout the summer.

My first project was assessing the effect of systematic uncertainties between simulated and experimental data on a graph neural network (GNN) predictions. The specific model we use for our analysis is called NuGraph2, and by the end of the summer I able to confidently show that the model is robust to detector modeling uncertainties. NuGraph2 is a rather mature machine learning (ML) tool, but my second project allowed me to work on prototyping my own model. We are interested in whether a point cloud model such as PointNeXt is able to infer the kinematics that distinguish between $\pi^0$ and $\eta$ particle decays. I was able to do some preliminary hyperparameter tuning to get preliminary results, but some biases in the training data may be skewing the performance metrics.

Part of the program were also opportunities to share my work with other students and faculty. We had a series of practice presentations leading up to an REU symposium where I gave a 15 minute overview of my work to students and faculty in the physics department. I also took part in an optional poster presentation (link to my poster) where I connected with other summer research students and shared my work in a more casual environment. Additionally, I presented my results on NuGraph2 to the MicroBooNE collaboration, ensuring that my work is known to others working with the analysis. When the final report and symposium presentation are finalized and posted to the UCSB Physics REU website, they will be found (here).