Applications of large scale kernel machines for real world human mood estimation

For many years the kernel methods were the primary tool for machine learning and computer vision. With their bad scalability for large dataset and the development of deep learning methods their usability decreased. In this work we show that the recent development of kernel approximation with random features can be used in real world applications. We build a mood estimation algorithm by utilizing multiple kernel learning approximated by random features. The algorithm is tested on popular large scale dataset and compared with state of the art methods.

• Krasimir Tonchev
• Nikolay Neshov
• Agata Manolova
• Teodora Sechkova

4th International Conference on Mathematics and Computers in Sciences and Industry, 2017.

https://www.iaras.org/iaras/filedownloads/ijsp/2017/003-0008(2017).pdf