Search for the radiative leptonic decay $D^{+}\to \gamma e^{+}{\nu }_e$ using Deep Learning

Using 20.3$\mathrm{f}{\mathrm{b}}^{-1}$ of $e^{+}{e}^{-}$ annihilation data collected at a center-of-mass energy of 3.773$\mathrm{G}\mathrm{e}\mathrm{V}$ with the BESIII detector, we report an improved search for the radiative leptonic decay $D^{+}\to \gamma e^{+}{\nu }_e$. An upper limit on its partial branching fraction for photon energies $E_{\gamma }>10\mathrm{M}\mathrm{e}\mathrm{V}$ was determined to be $1.2\times {10}^{-5}$ at 90% confidence level; this excludes most current theoretical predictions. A sophisticated deep learning approach, which includes thorough validation and is based on the Transformer architecture, was implemented to efficiently distinguish the signal from massive backgrounds.