Speaker
Description
Measuring the energy spectrum of cosmic electrons and positrons in the GeV - TeV energy range can provide crucial evidence for the existence of local sources, whether of astrophysical or exotic nature. Over the past years, measurements from different experiments have reported significant discrepancies, particularly at TeV energies, where uncertainties become more pronounced.
The latest Fermi-LAT measurement, published in 2017, relied on an electron+positron selection using supervised Machine Learning methods. While effective, these techniques are inherently model-dependent, as they require training on Monte Carlo simulations, making them susceptible to systematic uncertainties and biases.
In this work, we introduce a novel approach based on Unsupervised Learning techniques, which can autonomously identify patterns within experimental data. This method enables an almost model-independent selection, reducing potential biases and enhancing the robustness of the analysis. The selection is applied to data collected with the Fermi Large Area Telescope, demonstrating the potential of Unsupervised Learning in astroparticle physics.
| Collaboration(s) | Fermi-LAT Collaboration |
|---|
