Ies are processed. Six inducing points are applied to each FITC and VFE. Experiment 2: Impacts of s f on prediction accuracy and uncertainty. l is set for the 2 optimised value. s f varies from 0.1 by way of to 30.0. n is set to 0.five and 1.five, respectively. NO data from both cities are processed. Six inducing points are applied to both FITC and VFE. Experiment 3: Impacts of l on prediction accuracy and uncertainty. s f is set to the 2 optimised worth. l varies from 0.1 through to 30.0. n is set to 0.five and 1.5, respectively. NO information from both cities are processed. Six inducing points are applied to both FITC and VFE.RMSE =iNum (yi – yi )2 =1 , Num(34)where yi would be the ground truth value and yi represents predicted meant. Num could be the sample number in testing set. Figures 5 and 6 show the outcomes from Experiment 1. To create the results a lot more two distinguishable, the horizontal axes from the figures are set to log(n ). We can see from 2 is smaller, GPs perform the ideal in general, whilst the functionality Figure 5 that when n two of FITC and VFE varies. We can also observe that as n keeps escalating, the RMSE becomes really substantial for all methods/pollutants. Similar final results could be observed from Figure six also. Each comply with our theoretical conclusions, in spite of the truth that the 2 Neumann series is used to approximate the matrix inverse. We also notice that n features a two ) reaches extra significant impact on Sheffield data as RMSE increases ealier immediately after log(Isopropamide supplier nAtmosphere 2021, 12,11 ofzero. From Figure 6b,c, we also see that the uncertainty bounds of Sheffield information are 2 higher after log(n ) reaches zero. We assume the explanation is that Sheffield information are normally much less periodical than Pershawar information (see Figure 2), which influences the performance in the models.Pesh-NO-GP Pesh-NO-VFE Pesh-NO-FITC Shef-NO-GP Shef-NO-VFE Shef-NO-FITCPesh-NO -GPPesh-NO two -VFE2.2.Pesh-NO -FITCShef-NO2 -GP Shef-NO -VFEShef-NO2 -FITC1.1.0.0.—-(a)3Pesh-SO -GP2 2(b)Pesh-PM2.5 two.five 2.-GP -VFE -FITCPesh-SO -VFE2.Pesh-SO -FITC Shef-SO -GP2 22.Pesh-PM Pesh-PM Shef-PMShef-SO -VFE2.five 2.5 two.-GP -VFE -FITCShef-SO -FITCShef-PM Shef-PM1.1.0.0.—-(c)(d)2 Figure 5. Connection of n with four pollutants prediction RMSE: (a) NO, (b) NO2 , (c) SO2 , (d) PM2.five .eight six 4 two 0 -2 -2 -1 0 1 2 3Pesh_NO_GP Pesh_NO_VFE Pesh_NO_FITC Shef_NO_GP Shef_NO_VFE Shef_NO_FITC8 7 6 five 4 3 two 1 0 -1 -2 -2 -1 0 1 2 3Pesh-NO two -GP Pesh-NO 2 -VFE Pesh-NO 2 -FITC Shef-NO2 -GP Shef-NO2 -VFE Shef-NO2 -FITC(a)eight 7 6 five 4 four 3Pesh-SO2 -GP(b)eight 7 6Pesh-PM 2.five -GP Pesh-PM 2.five -VFE Pesh-PM two.5 -FITC Shef-PM 2.five -GP Shef-PM 2.five -VFE Shef-PM 2.five -FITCPesh-SO2 -VFE Pesh-SO2 -FITC Shef-SO 2 -GP Shef-SO two -VFE Shef-SO 2 -FITC1 0 -1 -2 -2 -1 0 12 1 0 -1 four -2 -1 0 1 2 3(c)(d)2 Figure 6. Connection of n with pollutants prediction uncertainty bound: (a) NO, (b) NO2 , (c) SO2 , (d) PM2.five .Atmosphere 2021, 12,12 of4.three. Impacts of Noise Level on ELBO and UBML Figure 7 shows the outcomes from Experiment two. In accordance with our theoretical final results, the effect of s f on the uncertainty need to grow to be higher as s f increases. That is verified by the outcomes shown in Figure 7b,d. Our theoretical benefits also suggest that the variation of s f wouldn’t impact the prediction accuracy. We can see from Figure 7a,c that when s f is smaller sized, it does influence the prediction accuracy, but when it exceeds a Okadaic acid ammonium salt Epigenetics certain worth, the impacts develop into negligible. Considering the Neumann series approximation, we would say that the experimental results comply with all the theoretical.
