Sparse Non-negative Matrix Factorization and its Application in Overlapped Chromatograms Separation

A new NMF algorithm has been proposed for the deconvolution of overlapping chromatograms of chemical mixture. Most of the NMF algorithms used so far for chromatogram separation do not converge to a stable limit point. To get same results for all the runs, instead of random initialization, three different initialization methods have been used namely, ALS-NMF (robust initialization), NNDSVD based initialization and EFA based initializations. To improve the convergence, a new sNMF algorithm with modified multiplicative update (ML-sNMF) has been proposed in this work for overlapped chromatogram separation. The algorithm has been validated with the help of simulated partially, severely overlapped and embedded chromatograms. The proposed ML-sNMF algorithm has also been validated with the help of experimental overlapping chromatograms obtained using Gas Chromatography –Flame Ionization Detector (GCFID) for the chemical mixture of acetone and acrolein.