A method for transplanting cross-component infrared spectroscopy models based on transfer learning
A technology of infrared spectroscopy and transfer learning, applied in machine learning, computing models, character and pattern recognition, etc., can solve problems such as transplantation of different component models
Active Publication Date: 2021-09-14
ZHONGBEI UNIV
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] In order to overcome the deficiencies in the prior art, the present invention provides a cross-component infrared spectrum model transplantation method based on migration learning, which solves the problem that the existing infrared spectrum quantitative analysis model transplantation method is only for the same component between different instruments. Transplantation, unable to solve the problem of model transplantation between different components
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0053] The data source is the near-infrared spectral data set of 80 corn samples, the spectral scanning range is 1100-2498nm, the scanning interval is 2nm, and each sample contains 700 wavelength points. Two near-infrared spectrometers were used to scan all the corn samples. For the convenience of expression, the names of the two instruments were respectively named: m5 and mp5, such as figure 2 shown. In this dataset, there are four components: moisture, corn oil, protein, and fatty acids.
[0054] In this embodiment, 50 samples are randomly selected from the entire data set, and the corresponding protein components and the spectra obtained by the instrument m5 scan form the source component training set Randomly select 20 samples from the remaining 30 samples, and the corresponding corn oil components and the spectrum obtained by the instrument mp5 scan form the target component training set The remaining 10 samples, the corresponding corn oil components and the spectrum...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention relates to a transfer learning-based cross-component infrared spectrum model transplantation method, which collects the sample spectral data of the source component and the target component and performs principal component analysis, and selects the appropriate number of principal components according to the cumulative contribution rate, Carry out model transplantation, iteratively train multiple weak quantitative analysis models, and dynamically adjust the sample weights of source components and target components during each iteration. For the samples of the source component data set, if the prediction error of the previously constructed quantitative analysis model is large, the weight of the sample is reduced; for the samples of the target component data set, if the previous construction of the quantitative analysis model has a large prediction error. is large, increase the weight of the sample. Finally, for the target component sample to be predicted, the sample is sent to each weak quantitative analysis model for prediction, and the results of each weak quantitative analysis model are integrated and summarized by the weighted average method to obtain the prediction result of the strong quantitative analysis model .
Description
technical field [0001] The present invention relates to the technical field of infrared spectrum quantitative analysis models, more specifically, to a method for transplanting cross-component infrared spectrum models based on migration learning. Background technique [0002] With the rapid development of chemometrics, infrared spectroscopy has been widely used and promoted in many fields such as agriculture and industry due to its characteristics of non-destructive, rapid, pollution-free and low cost. However, in the actual application process, due to the different manufacturers and models of infrared spectrometers used, the established quantitative analysis model cannot be applied, and if a customized quantitative analysis model is established for each instrument, it will require a lot of investment Manpower, financial and material resources, obviously this idea is not advisable. Existing model transplantation methods such as slope-intercept method, direct correction metho...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/00G06K9/62G06N20/00
CPCG06F2218/12G06F18/2135
Inventor 陈媛媛李墅娜张瑞景宁王志斌
Owner ZHONGBEI UNIV