国外的纯学术ppt里面有些动画很是生动.ppt

《国外的纯学术ppt里面有些动画很是生动.ppt》由会员分享，可在线阅读，更多相关《国外的纯学术ppt里面有些动画很是生动.ppt（77页珍藏版）》请在三一办公上搜索。

1、BackgroundHistory,0.order data,1870s,X,X,5.2,2.6,6.8,Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,BackgroundFluorescence,Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,Lightsource,Sample,Detector,Excitessample,Emittedfrom sample,BackgroundFluorescence,Backgrou

2、nd Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,BackgroundFluorescence,X,Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,BackgroundLF-NMR,Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,N,Detector,Magnetic Field,Radio signal,Sample,X,*)Low Field Nuclear

3、Magnetic Resonance,BackgroundLF-NMR*),Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,X,BackgroundLF-NMR:SLICING*),Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,*)Pedersen,Bro&Engelsen(2001)Magnetic Resonance in Food Science,202-209,Can be seen as an expansion o

4、f PCA from two-way data to multi-way data,BackgroundPARAFAC,Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,BackgroundPARAFAC,PARAFACI+J+K per factor,PCAI+J K per factor,PARAFACFluorescence,=,A,B,C,Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,PARAFACLF-NMR,=,A,

5、B,C,Background Spectroscopy PARAFACFluorescenceLF-NMR2nd orderSummary,IntroductionFluorescence,Make an automatic model for use with fluorescence dataNumber of factorsHandling scatter effectsPractical aspect:missing values in the landscape,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd

6、orderSummary,Fluorescence#of Factors,Cross-validation,Structure in loadings,Split-half analysis,Jack-knifing,A-priori knowledge,Ratios,Known samples,Bootstrapping,New automaticmethod,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,Rinnan&Bro,#of FactorsData,Fluorescence dat

7、a12 data set 10 samples in each3-5 fluorophores in each sampleAnalyzed by PARAFAC81 diagnostic toolsModels with increasing complexity,1,2,3,4,5,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,#of FactorsStructure in loadings,Rinnan&Bro,BackgroundFluorescence#of F

8、actors Light scatter NaNLF-NMR2nd orderSummary,#of FactorsRatio PCA/PARAFAC,PARAFAC,PCA,Number of components,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,#of FactorsCORCONDIA*),Compare PARAFAC and Tucker3,BackgroundFluorescence#of Factors Light scatter Practic

9、alLF-NMR2nd orderSummary,Number of components,*)Bro&Kiers(2003)J.of Chemometrics,274-286,#of FactorsUniqueness in landscapes,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,#of FactorsGood vs.Bad,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NM

10、R2nd orderSummary,#of FactorsGood vs.Bad,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,#of FactorsKnown samples,A-score0.500-2,Emission wavelength(nm),Excitation wavelength(nm),Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary

11、,F,F+2,Factors,#of FactorsKnown samples,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,81,24,1,0,Number of factors,Times estimated,#of FactorsResults,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,Resultat,Data set1234567891

12、01112,Good,Catechol,Rinnan&Bro,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,FluorescenceLight scatter,Excitation,Emission,2nd orderRayleigh,1st order Rayleigh,Raman,BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Rinnan&Ande

13、rsen(2004),Light scatterWhy a problem?,Rinnan&Andersen(2004),BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterWhy a problem?,X,X,Rinnan&Andersen(2004),BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSum

14、mary,Light scatterSample decomposition,Rinnan&Andersen(2004),BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterHandling scatter,Subtraction of standard,Cut off and insert missing,Weights,Modeling of Rayleigh,Rinnan&Andersen(2004),BackgroundFl

15、uorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterWhy more than one method!?,The data presented so far is a bit simple Sugar data,Excitation,Emission,1st order Rayleigh,Rinnan&Andersen(2004),BackgroundFluorescence#of Factors Light scatter Band of NaN Mod

16、eling NaNLF-NMR2nd orderSummary,Light scatterCutting off(Hard weights),Emission loadings,Excitation loadings,Rinnan&Andersen(2004),BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterWeighting-MILES,Emission loadings,Excitation loadings,Rinnan&

17、Andersen(2004),BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterBand of missing values,BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Rinnan&Andersen(2004),Band of missing valuesHard weights,Em

18、ission loadings,Excitation loadings,Rinnan&Andersen(2004),BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Band of missing valuesWeighting-MILES,Emission loadings,Excitation loadings,Rinnan&Andersen(2004),BackgroundFluorescence#of Factors Light scatter B

19、and of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterA novel method,The Rayleigh scatter width has to be estimated quite accuratelyThe band width of missing data should also be correctWhat about an automatic method of removing the Rayleigh scatter,that was not so prone to the estimation of the

20、 width of the Rayleigh scatter?Modeling the Rayleigh is the answer!,BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Rinnan,Booksh&Bro,Light scatterModeling Rayleigh,A Gauss-Lorentz curve fitting method,Rinnan,Booksh&Bro,BackgroundFluorescence#of Factors

21、 Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterModeling Rayleigh,Rinnan,Booksh&Bro,BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterModeling Rayleigh,Rinnan,Booksh&Bro,BackgroundFluorescence#of Factors Light scatt

22、er Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterFancy good,Rinnan,Booksh&Bro,BackgroundFluorescence#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,Light scatterWith constraints better,Emission loadings,Excitation loadings,Rinnan,Booksh&Bro,BackgroundFluorescen

23、ce#of Factors Light scatter Band of NaN Modeling NaNLF-NMR2nd orderSummary,FluorescenceMissing values,Can be treated with:Letting PARAFAC handle the missing valuesWeighting the missing area downNon-negativity constraintsInsertion of 0s into the matrix,Thygesen,Rinnan,Barsberg&Mller(2004):CehmoLab,71

24、,p.97-106,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,Missing valuesAlternatives,Missingvalues,Zeros,Signal/Data area,Thygesen,Rinnan,Barsberg&Mller(2004):CehmoLab,71,p.97-106,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,Missing valuesResult

25、s,Thygesen,Rinnan,Barsberg&Mller(2004):CehmoLab,71,p.97-106,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,Missing valuesResults-Landscapes,None,Weighted,Non-Negativity,Zeros,Thygesen,Rinnan,Barsberg&Mller(2004):CehmoLab,71,p.97-106,BackgroundFluorescence#of Factors Light

26、scatter NaNLF-NMR2nd orderSummary,Missing valuesResults-Excitation,None,Weighted,Non-Negativity,Zeros,Thygesen,Rinnan,Barsberg&Mller(2004):CehmoLab,71,p.97-106,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,Missing valuesResults-Emission,None,Weighted,Non-Negativity,Zeros,

27、Thygesen,Rinnan,Barsberg&Mller(2004):CehmoLab,71,p.97-106,BackgroundFluorescence#of Factors Light scatter NaNLF-NMR2nd orderSummary,IntroductionLF-NMR,Practical aspects of PARAFAC and LF-NMR(SLICING)Correction of baselineClassificationPrediction(semi 2nd order prediction),BackgroundFluorescenceLF-NM

28、R Correction Classification Regression2nd orderSummary,LF-NMRCorrection,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Andersen&Rinnan(2002):LWT,35,p.687-696,CorrectionResidual,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Anders

29、en&Rinnan(2002):LWT,35,p.687-696,CorrectionFactors,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Andersen&Rinnan(2002):LWT,35,p.687-696,CorrectionMethod,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Andersen&Rinnan(2002):LWT,35,

30、p.687-696,CorrectionCorrected data,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Andersen&Rinnan(2002):LWT,35,p.687-696,CorrectionResult-Factors,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Andersen&Rinnan(2002):LWT,35,p.687-69

31、6,LF-NMRClassification,Sensory data,LF-NMR,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Povlsen,Rinnan,van den Berg,Andersen&Thybo(2003):LWT,36,p.423-432,LF-NMRClassification,PCA and Sensory,PARAFAC and LF-NMR,BackgroundFluorescenceLF-NMR Correction Classificatio

32、n Regression2nd orderSummary,Povlsen,Rinnan,van den Berg,Andersen&Thybo(2003):LWT,36,p.423-432,LF-NMRRegression,PLS,SLICING,BackgroundFluorescenceLF-NMR Correction Classification Regression2nd orderSummary,Rinnan(2003):IJAS,15,p.393-402,LF-NMRRegression,1-Q2,BackgroundFluorescenceLF-NMR Correction C

33、lassification Regression2nd orderSummary,Rinnan(2003):IJAS,15,p.393-402,IntroductionSecond order prediction,Measured(R+G+B),Predicted(Yellow),Calibration set,New samples,Error,BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Rinnan,Riu&Bro(2004),Second order prediction

34、Unfolding fluorescence,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Second order predictionPLS vs PARAFAC,Unfold-PLS,Samples,Excitation,The data-EEM,Samples,Different excitations,Emission,Emission,Emission,PARAFAC,Samples,Emission,Emission,Exci

35、tation,Samples+Excitation Emission,Samples+Excitation+Emission,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Second order predictionData sets,Calibration set,New samples,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Altern

36、atives UncertaintySummary,Second order predictionPLS vs PARAFAC,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Second order predictionData sets with interferents,Calibration set,New samples,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd ord

37、er Advantage Alternatives UncertaintySummary,Second order predictionPLS vs PARAFAC,Measured,Predicted,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Second order predictionSecond order advantage,PARAFAC,PLS,Catechol,Hydroquinone,Rinnan,Riu&Bro(20

38、04),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,EEM,New samples,EEM,Calibration set,Second order predictionSecond order advantage,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Second order predictionAlternativ

39、es,=,A,B,C,Calibration,New samples,0,A,B,C,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,SOP Example,All simulated data3 or 4 analytes in calibration set3 interferentsDifferent kind of overlap between analytes and interferentsFour different nois

40、e levels7,4,3 and 2 samples in the calibration setOne or several samples in the test set10 different noise additions 10 replicates,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,SOP Ex:Results,Analyzed by ANOVA and PCATwo very bad methodsTwo good

41、 methods,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,New samples,Second order predictionResults,Calibration,0,A,B,C,Best,2.best,A,A,Problematic,Rinnan,Riu&Bro(2004),BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummar

42、y,Second order predictionUncertainty in prediction,Multivariate,Univariate,Bro,Rinnan&Faber(2004)ChemoLab,BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Second order predictionUncertainty in prediction,Reference Error:0+Model Error:0.84=Total Error:0.84,Reference Err

43、or:0.84+Model Error:0=Total Error:0.84,Bro,Rinnan&Faber(2004)ChemoLab,BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives UncertaintySummary,Second order predictionUncertainty in prediction,Bro,Rinnan&Faber(2004)ChemoLab,BackgroundFluorescenceLF-NMR2nd order Advantage Alternatives Uncertain

44、tySummary,Application of PARAFACSummary,FluorescenceNumber of factorsHandling RayleighPractical aspects:missing valuesLF-NMRBaseline correctionClassificationPredictionSecond Order PredictionBest methodEstimate uncertainty,BackgroundFluorescenceLF-NMR2nd orderSummary,Thanks are due to,Rasmus BroVibeke T.SvenssonFrans van den BergJordi RiuKarl S.BookshCharlotte M.AndersenAll at Food Technology,Co-authors:Lisbeth G.ThygesenKlaas FaberSren BarsbergJens K.S.MllerHenrik J.AndersenAnette K.Thybo,