Reliable Model Selection without Reference Values by Utilizing Model Diversity with Prediction Similarity

Local Modeling by Adapting Source Calibration Models to Analyte Shifted Target Domain Samples Without Reference Values

Spectral multivariate calibration aims to derive models characterizing mathematical relationships between sample analyte amounts and corresponding spectral responses. These models are effective at predicting target domain sample analyte amounts when target samples are within the analyte and spectral calibration source domain. Models fail when target samples shift (analyte amounts and/or spectra) from the original calibration domain model. A total recalibration solution requires acquisition of new sample reference values and spectra. However, obtaining enough reference values to distinguish the target domain may be challenging or expensive. A simpler approach adapts the original model to the target domain using target sample spectra without analyte reference values (unlabeled). Analytical chemists have developed several machine learning algorithms using unlabeled regression domain adaptation processes. Unfortunately, prediction accuracy declines for these methods depending on how much the target domain analyte distribution has shifted from the calibration distribution, and regression transfer learning methods are instead needed. Regression domain adaptation and transfer learning are often referred to as model updating in analytical chemistry, but regression domain adaptation only applies to spectral shifts. The regression transfer learning method presented in this paper named null augmentation regression constant analyte (NARCA) leverages unlabeled repeat spectra of a single target sample to update an original calibration model to the shifted target domain sample. With sample repeat spectra, the analyte amount can be assumed constant or nearly constant for NARCA and because models are formed for one sample, NARCA operates as a local modeling method. The performance of NARCA as a regression transfer learning method is evaluated using five near-infrared data sets.

Semi-Supervised Deep Learning-Based Multi-component Spectral Calibration Modeling for UV-vis and Near-Infrared Spectroscopy without Information Loss

Spectral analysis is an important method for characterizing and identifying chemical species. However, quantitative spectral analysis of multiple chemical properties in the real world has always been a challenging problem due to the strong correlation, massive noise, and serious information overlapping of the spectral features. Here, we present a new semi-supervised spectral calibration method based on information lossless decoupling of spectral features named NICEM. To realize the separation and extraction of key latent features, the method uses the flow-based model non-linear independent component estimation (NICE) to learn the sample distribution. The spectral data information is transformed into independent latent variables obeying Gaussian distribution by the reversible structure of deep network without information loss, so as to find the essential properties and realize the feature nonlinear decomposition. Moreover, the association between the input latent feature variables and attributes is evaluated by the maximum mutual information coefficient to eliminate the adverse effects of irrelevant information in the latent variable space and mine key information. Since the latent variables are independent in each dimension, the NICEM method is easier to establish an accurate semi-supervised multi-component calibration model even for high overlapping and complex spectral data. The applicability of the proposed spectral modeling method is demonstrated by using three ultraviolet-visible and near-infrared spectral data sets with 15 physical and chemical properties including diesel fuels, corn, and multi-metal ions solution. Results show that the proposed NICEM method has the highest determination coefficient (R2) and significantly improves extrapolation compared with the seven state-of-the-art methods. The proposed method is intuitive because it obviates complex feature engineering and prior knowledge and is a promising spectral calibration tool for quantitative analysis in other spectroscopy applications.

Physicochemical Responsive Integrated Similarity Measure (PRISM) for a Comprehensive Quantitative Perspective of Sample Similarity Dynamically Assessed with NIR Spectra

Determining sample similarity underlies many foundational principles in analytical chemistry. For example, calibration models are unsuitable to predict outliers. Calibration transfer methods assume a moderate degree of sample and measurement dissimilarities between a calibration set and target prediction samples. Classification approaches link target sample similarities to groups of similar class samples. Although similarity is ubiquitous in analytical chemistry and everyday life, quantifying sample similarity is without a straightforward solution, especially when target domain samples are unlabeled and the only known features are measurable, such as spectra (the focus of this paper). The process proposed to assess sample similarity integrates spectral similarity information with contextual considerations among source analyte contents, model, and analyte predictions. This hybrid approach named the physicochemical responsive integrated similarity measure (PRISM) amplifies hidden-but-essential physicochemical properties encoded within respective spectra. PRISM is tested on four near-infrared (NIR) data sets for four diverse application areas to show efficacy. These applications are the assessment of prediction reliability and model updating for model generalizability, outlier detection, and basic matrix matching evaluation. Discussion is provided on adapting PRISM to classification problems. Results indicate that PRISM collects large amounts of similarity information and effectively integrates it to produce a quantitative similarity evaluation between the target sample and a source domain. The approach is also useful for biological samples with additional physiochemical variations. While PRISM is dynamically tested on NIR data, parts of PRISM were previously applied to other data types, and PRISM should be applicable to other measurement systems perturbed by matrix effects.

Domain Adaptation for In-Line Allergen Classification of Agri-Food Powders Using Near-Infrared Spectroscopy

The addition of incorrect agri-food powders to a production line due to human error is a large safety concern in food and drink manufacturing, owing to incorporation of allergens in the final product. This work combines near-infrared spectroscopy with machine-learning models for early detection of this problem. Specifically, domain adaptation is used to transfer models from spectra acquired under stationary conditions to moving samples, thereby minimizing the volume of labelled data required to collect on a production line. Two deep-learning domain-adaptation methodologies are used: domain-adversarial neural networks and semisupervised generative adversarial neural networks. Overall, accuracy of up to 96.0% was achieved using no labelled data from the target domain moving spectra, and up to 99.68% was achieved when incorporating a single labelled data instance for each material into model training. Using both domain-adaptation methodologies together achieved the highest prediction accuracies on average, as did combining measurements from two near-infrared spectroscopy sensors with different wavelength ranges. Ensemble methods were used to further increase model accuracy and provide quantification of model uncertainty, and a feature-permutation method was used for global interpretability of the models.

Calibration Model Updating to Novel Sample and Measurement Conditions without Reference Values

Updating a calibration model formed in original (primary) sample and spectral measurement conditions to predict analyte values in novel (secondary) conditions is an essential activity in analytical chemistry in order to avoid a complete recalibration. Established model updating methods require sample analyte reference values for a small set of secondary domain samples (labeled data) to be used in updating processes. Because obtaining reference values is time consuming and is the costly part of any calibration, methods are needed that do not require labeled secondary samples, thereby allowing on demand model updating. This paper compares model updating methods with and without labeled secondary samples. A hybrid model updating approach is also developed and evaluated. Unfortunately, a major impediment to adapting a model without secondary analyte reference values has been model selection. Because multiple tuning parameters are commonly involved in model updating methods, thousands of models are formed, making model selection complex. A recently developed framework is evaluated for automatic model selection of several two to three tuning parameter-based model updating methods without secondary analyte reference values (labels). The model selection method is based on model diversity and prediction similarity (MDPS) of the unlabeled samples to be predicted. The new secondary samples to be predicted can be used to form the updated models and again to select the final predicting models. Because models are formed and selected on demand to directly predict target samples, complicated cross-validation processes are not needed. Four near-infrared data sets covering 40 model updating situations are evaluated showing that MDPS can select reliable updated models outperforming or rivaling prediction errors from total recalibrations with secondary reference values.