airPLS(s,lambda, ⋯)

upper panel of Figure 5.8 shows the estimated baseline for the

d spectrum shown in Figure 5.5. The lower panel of Figure 5.8

he extracted peak spectrum. Compared with Figure 5.7, this

n generated an even worse result.

he upper panel, the estimated baseline for the spectra data shown in Figure 5.5

irPLS package. The lower panel, the estimated peak spectrum based on the

aseline using airPLS for the upper panel spectrum.

asymmetric least square smoother

mmetric least square smoother algorithm (ALSS) [Liland, et al.,

mploys the concept of residual weights. Suppose a spectrum is

by a vector s, an unknown baseline is denoted by a vector b,

weights are denoted by a vector w, the regularisation constant is

by ߣ. The objective function of the asymmetric least squared

is defined as below, where ܌ൌᇞᇞ܊ and W is a diagonal matrix