To estimate DNAm in leukocyte subsets, we imputed immunophenotypic and DNAm measurements into the NNLS regression model.
 Nonstandard abbreviations: LDC, leukocyte differential count; DNAm, DNA methylation; CpG, CG dinucleotide; NNLS, nonnegative least-squares; cfDNA, cell-free circulating DNA.
These estimations are based on DNAm levels at the 3 CpGs for granulocytes (WDR20), lymphocytes (FYN), and monocytes (CENPA) in whole blood of the training set (n = 60) that were then implemented into the reverse approach of the nonnegative least-squares (NNLS) linear model.
Second, KNDLR is much faster than NNLS and SRC, especially on AR.
Comprehensive experiments on six different datasets demonstrate that proposed KNDLR outperforms existing LR method for classification and some other commonly used methods such as SVM, NNLS, SRC, and LRC, and our KNDLR is efficient.
Number of training 5 6 7 8 samples per class Our method 70.80 73.53 76.05 80.26 DLSR 50.48 50.93 52.80 53.77 CLSR 63.58 65.33 67.30 70.57 NNLS 68.90 71.02 73.88 76.83 ThSVM 70.42 72.20 74.93 79.46 KNN 59.30 61.76 63.85 67.49 SRC 56.18 58.44 60.13 63.09 LRC 68.90 71.22 73.78 77.71 NDLR 65.66 66.76 69.38 73.60 Number of training 9 10 samples per class Our method 81.33 82.36 DLSR 53.77 54.56 CLSR 71.00 71.40 NNLS 78.27 79.68 ThSVM 80.17 81.92 KNN 69.23 69.40 SRC 63.57 64.36 LRC 79.20 80.68 NDLR 73.57 74.56 Table 2: Accuracies (%) of different methods on the FERET database.
To exploit the Adecomp-SHOT on diffusion dataset and to take into account the ground truth properties of the diffusion, we imposed a real and nonnegative constraint by using the NNLS
algorithm and by introducing a change of variables.
Let us have an NNLS problem, an overdetermined or underdetermined system of linear equations
Like the simplex algorithm, the NNLS algorithm by Kong (2007) has a basis B consisting of a set of linearly independent columns of E used for solving a system of equations.
For the results presented in this report, the NNLS pressures were:
Similar to the scanning technique, the NNLS applied a least squares analysis to the DeltaQ relationship using the measured data.
Figure 4 shows the subimages of the true, blurred, PCG restored image after three iterations, and NNLS
improved subimages in this case.