This paper was presented researches for dimension reduction of sound speed profiles (SSP). In addition, we shown optimal information extraction and optimal compression for the SSP profiles. For the dimension reduction of SSP we used the principal component analysis, K-SVD, autoencoder and DictionaryLearning. For the information extraction was used following machine learning methods: boosting, logistic regression, random forest. These researches was conducted for the Barencevo sea SSP base.
43.30.Pc Ocean parameter estimation by acoustical methods; remote sensing; imaging, inversion, acoustic tomography
02.60.Ed Interpolation; curve fitting
$^1$Moscow State Aviation University (MAI)\
$^2$Andreyev Acoustics Institute
Science News of the Faculty of Physics, Lomonosov Moscow State University
This new information publication, which is intended to convey to the staff, students and graduate students, faculty colleagues and partners of the main achievements of scientists and scientific information on the events in the life of university physicists.