The IceCube Neutrino Observatory measures high energy atmospheric neutrinos with high statistics. These atmospheric neutrinos originate from cosmic ray interactions in the atmosphere, mainly from the decay of pions and kaons. However, the rate of the measured neutrinos is affected by seasonal temperature variations in the Stratosphere, which are expected to increase with the particle's energy. Here, we present an analysis to determine the spectral dependence of the seasonal neutrino flux from IceCube's atmospheric muon neutrino data. The seasonal energy spectra are obtained by the Dortmund Spectrum Estimation Algorithm (DSEA+), a novel approach to spectrum unfolding, in which the energy distribution is estimated from measured quantities with machine learning algorithms. Preliminary results based on 10% of the data measured between 2011 and 2020 will be discussed.