Robust domain-adaptive discriminant analysis

Kouw, Wouter M.; Loog, Marco

doi:10.1016/j.patrec.2021.05.005

Cited by 2 publications

(10 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are approaches for handling distributional shifts but with a clearly different objective (Kouw & Loog, 2018; McNamara & Balcan, 2017; Pan & Yang, 2010). First, there is model retraining (Cui et al., 2018).…”

Section: Related Workmentioning

confidence: 99%

“…Formally, we address a distributional shift of the type where , but . This form of distributional shift is known in the literature as covariate shift (Kouw & Loog, 2018). In fact, distributional shifts in the form of covariate shifts are common at Aker and across OM practice.…”

Section: Model Development1mentioning

confidence: 99%

“…We consider other baselines from the machine learning literature (Kouw & Loog, 2018; McNamara & Balcan, 2017; Pan & Yang, 2010) that can—in principle—also handle distributional shifts, namely, model retraining and transfer learning with fine‐tuning. However, we emphasize that the aforementioned baselines focus on a different setup called supervised domain adaptation (and not un supervised domain adaptation, as in our decision problem).…”

Section: Robustness Checksmentioning

confidence: 99%

“…While there is increasing traction of predictive analytics in operations management (OM) (e.g., Baardman et al., 2018; Bastani et al., 2022; Mišić & Perakis, 2020; Olsen & Tomlin, 2020), there is, however, scarce advice on how distributional shifts can be addressed. Existing approaches to address distributional shifts make use of model retraining (Cui et al., 2018) or transfer learning via fine‐tuning (Bastani, 2021; Kouw & Loog, 2018; McNamara & Balcan, 2017; Pan & Yang, 2010). These methods are effective in make‐to‐stock manufacturing (e.g., fast‐moving consumer goods) characterized by low variety and high volumes.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Addressing distributional shifts in operations management: The case of order fulfillment in customized production

Senoner

Kratzwald

Kuzmanovic

et al. 2023

Production and Operations Management

View full text Add to dashboard Cite

To meet order fulfillment targets, manufacturers seek to optimize production schedules. Machine learning can support this objective by predicting throughput times on production lines given order specifications. However, this is challenging when manufacturers produce customized products because customization often leads to changes in the probability distribution of operational data—so‐called distributional shifts. Distributional shifts can harm the performance of predictive models when deployed to future customer orders with new specifications. The literature provides limited advice on how such distributional shifts can be addressed in operations management. Here, we propose a data‐driven approach based on adversarial learning, which allows us to account for distributional shifts in manufacturing settings with high degrees of product customization. We empirically validate our proposed approach using real‐world data from a job shop production that supplies large metal components to an oil platform construction yard. Across an extensive series of numerical experiments, we find that our adversarial learning approach outperforms common baselines. Overall, this paper shows how production managers can improve their decision making under distributional shifts.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Model Development1mentioning

confidence: 99%

Section: Robustness Checksmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Addressing distributional shifts in operations management: The case of order fulfillment in customized production

Senoner

Kratzwald

Kuzmanovic

et al. 2023

Production and Operations Management

View full text Add to dashboard Cite

show abstract

“…Various approaches exist to adapt a model to new current secondary conditions and maintain accurate predictions . In machine learning terminology, these methods would be categorized as transfer learning. − Transfer learning is a broad-based term used for dealing with situations where the primary and secondary conditions differ. The difference can be any combination of dissimilarities between spectral shape and location changes − for X differences and the analyte and interferent distributions (concentration ranges) for Y disparities.…”

mentioning

confidence: 99%

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

Spiers

Kalivas

2021

Anal. Chem.

View full text Add to dashboard Cite

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.

show abstract

Robust domain-adaptive discriminant analysis

Cited by 2 publications

References 9 publications

Addressing distributional shifts in operations management: The case of order fulfillment in customized production

Addressing distributional shifts in operations management: The case of order fulfillment in customized production

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

Contact Info

Product

Resources

About