Deep calibration transfer: Transferring deep learning models between infrared spectroscopy instruments

Puneet Mishra,Dário Passos

doi:10.1016/j.infrared.2021.103863

Puneet Mishra, Dário Passos

Open Access

https://doi.org/10.1016/j.infrared.2021.103863

Copy DOI

Abstract

Calibration transfer (CT) is required when a model developed on one instrument needs to be transferred and used on a new instrument. Several methods are available in the chemometrics domain to transfer the multivariate calibrations developed using modelling techniques such as partial least-square regression. However, recently deep learning (DL) models are gaining popularity to model spectral data. The traditional multivariate CT methods are not suitable to transfer a deep learning model which is based on neural networks architectures. Hence, this study presents the concept of deep calibration transfer (CT) for transferring a DL model made on one instrument onto a new instrument. The deep CT is based on the concept of transfer learning from the DL domain. To show it, two different CT cases are presented. The first case is the CT between benchtop FT-NIR (Fourier Transform Near Infrared) instruments, and the second case is the CT between handheld NIR (Near Infrared) instruments. In both the demonstrated cases, the transfer was performed standard-free i.e., no common standard samples were used to estimate any transfer function. The results showed that with deep CT, the DL models made on one instrument can be easily adapted and transferred to a new instrument. The main benefit of the deep CT is that it is a standard free approach and does not require any standard sample measurements. Such a standard free approach to transfer DL models between instruments can support a widespread sharing of chemometric DL models between the scientific practitioners.

Highlights

Chemometrics calibrations are of high-value, as they are the results of extensive sensor measurements and wet chemistry destructive refer ence analysis [1]
This study showed that, like chemometrics classic models, deep learning (DL) models have generalisation problems when directly used on a new instru ment
The DL models require a model transfer procedure to adjust to the new instrument on which they are intended to be deployed

Summary

Introduction

Chemometrics calibrations are of high-value, as they are the results of extensive sensor measurements and wet chemistry destructive refer ence analysis [1]. The user always ex pects that once the calibration is set up, the need for repeated wet chemistry and data analysis can be bypassed [5]. The expectations of the user are met, there are certain situations where the user may need to repeat the wet chemistry and data analysis. When the user decides to increase its research facilities by adding on a new similar instrument or when the user needs to replace an old/damaged instrument with a new one [6,7]. To be able to calibrate the new instrument, the user may need to repeat the calibra tion procedure previously done for the old instrument [6,7,8]

Objectives

Methods

Results

Conclusion