Implementation of valid and stable algorithm of QL1-NMF for analyzing environmental elf magnetic signals
Motoaki Mouri, Ichi Takumi, Hiroshi Yasukawa · 2017
Previously, we developed two NMF algorithms named QL1-NMF using quasi-L1 norm for analyzing environmental ELF magnetic field measurements. When the data includes many outliers, the QL1-NMF algorithms returned better results than other BSS algorithms using L1 norm. However, the derivative of cost function in our first algorithm was not based on a monotonically increasing function. This problem decreased the validity of algorithm. Our second algorithm had problems about stability though it was based on a monotonically increasing derivative. The method therefore required improvement of validity and stability. In the work described in this paper, we newly introduced the update functions that were based on a monotonically increasing derivative. Computer simulation results and real data's results confirm the new algorithm works more stable than the previous one.