Cosmic rays reveal an invisible and energetic component of the Universe. Despite being discovered over a century ago, their origins and acceleration mechanisms remain largely unknown. Possible sources include supernova remnants, pulsars, active galactic nuclei, and even dark matter. Recent observations show an unexpected excess of high-energy electrons and positrons, suggesting nearby astrophysical sources or exotic origins.

The Dark Matter Particle Explorer (DAMPE), launched in 2015, was designed to investigate these mysteries by precisely measuring cosmic-ray particles. Its first results revealed a spectral break near 0.9 TeV in the electron-positron flux. Extending these measurements to higher energies is challenging due to the rarity of such particles.

In this work, we improve statistics by including side-entering events and applying deep learning techniques. Convolutional neural networks significantly enhance energy reconstruction and particle identification, enabling a reliable flux measurement from 100 GeV to 3 TeV, consistent with previous DAMPE results.