A new method for solving the complete inverse ill-posed problem for the Fredholm integral equation of the first kind using deep learning algorithms. The approach first aims to reconstruct charged-particle energy spectra from experimental attenuation curves and dose curves - a classic ill-posed inverse problem in experimental plasma and accelerator physics. Beyond beam diagnostics, the same formulation is in demand across a wide range of applied-physics and industrial problems where an unknown distribution must be recovered from sparse and noisy empirical datasets. This innovative technique has enabled the acquisition of many spectral energy distributions for electron beams from pulsed vacuum and gas-filled diodes, revealing complex spectral features such as anomalous energy electrons (AEE).
- Kozhevnikov, V., Kozyrev, A., Sorokin, D., Tarasenko, V., Beloplotov, D., Baksht, E., & Lomaev, M. (2026) Plasma, 9 (2), 18 (DOI: 10.3390/plasma9020018)
- Tarasenko, V.F., Kozhevnikov, V.Y., Kozyrev, A.V., Baksht, E.K., & Vorobyov, M.S. (2026) arXiv preprint (DOI: 10.48550/arXiv.2606.27889)
- Kozhevnikov, V., Kozyrev, A., Klepalova, E., Tarasenko, V., & Baksht, E. (2026) arXiv preprint (DOI: 10.48550/arXiv.2608.11628)
- Kozhevnikov, V.Yu., Tarasenko, V.F., Vorobyov, M.S., Baksht, E.Kh., & Kozyrev, A.V. (2026) arXiv preprint (DOI: 10.48550/arXiv.2609.34350)
