Optimization of process parameters using weighted convex loss functions

Chang, Yen‐Chang; Liu, Ching‐Ti; Hung, Wen-Liang

doi:10.1016/j.ejor.2008.04.013

Cited by 5 publications

(2 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, weighing misclassification of inputs belonging to minorities can heavily help in soothing existing biases in data (Phan et al, 2017). Similarly, using a convex loss function with weighted penalties to differentiate between quality variables helps to find the best parameters (Chang et al, 2009). We follow suit and utilize WLFs to harshly penalize the most egregious mispredictions from a human's semantic similarity perspective.…”

Section: Related Workmentioning

confidence: 99%

Not all Failure Modes are Created Equal: Training Deep Neural Networks for Explicable (Mis)Classification

Olmo¹,

Sengupta²,

Kambhampati³

2020

Preprint

View full text Add to dashboard Cite

Deep Neural Networks are often brittle on image classification tasks and known to misclassify inputs. While these misclassifications may be inevitable, all failure modes cannot be considered equal. Certain misclassifications (eg. classifying the image of a dog to an airplane) can create surprise and result in the loss of human trust in the system. Even worse, certain errors (eg. a person misclassified as a primate) can have societal impacts. Thus, in this work, we aim to reduce inexplicable errors. To address this challenge, we first discuss how to obtain the class-level semantics that captures the human's expectation (M h ) regarding which classes are semantically close vs. ones that are far away. We show that for data-sets like CIFAR-10 and CIFAR-100, class-level semantics can be obtained by leveraging human subject studies (significantly inexpensive compared to existing works) and, whenever possible, by utilizing publicly available human-curated knowledge. Second, we propose the use of Weighted Loss Functions (WLFs) to penalize misclassifications by the weight of their inexplicability. Finally, we show that training (or even fine-tuning) existing classifiers with the two proposed methods lead to Deep Neural Networks that have (1) comparable top-1 accuracy, an important metric in operational contexts, (2) more explicable failure modes and (3) require significantly less cost in teams of additional human labels compared to existing work.* Indicates equal contribution. Names ordered alphabetically.

show abstract

Section: Related Workmentioning

confidence: 99%

Not all Failure Modes are Created Equal: Training Deep Neural Networks for Explicable (Mis)Classification

Olmo¹,

Sengupta²,

Kambhampati³

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…30,31 However, some aspects have been questioned by some academic scholars, and Taguchi’s on-line quality feedback control system is not sufficiently rigorous. 32,33 Researchers are attempting to improve it, but their focus is mainly on the quality loss function and not on the on-line quality feedback control system. 34,35 A few scholars have identified problems related to the distributional assumption of a product’s quality characteristic value and the estimation of the rate of nonconformance in Taguchi’s on-line quality feedback control system; these problems greatly affect the effectiveness of the system.…”

Section: Introductionmentioning

confidence: 99%

Optimization of Taguchi’s on-line quality feedback control system

Zhang

Chen

et al. 2016

Proceedings of the Institution of Mechanical Engineers, Part B:

View full text Add to dashboard Cite

A feedback control system for quality characteristics is one of the most crucial components in Taguchi’s on-line quality management. In this article, three problems in Taguchi’s on-line quality feedback control system are discussed. Furthermore, countermeasures for improving the system are proposed. First, the quality loss of products that are out of control is reestimated by approximating the probability density of their quality characteristic value to a linear distribution. Second, the quality loss of products that are under control is reestimated by approximating the probability density of their quality characteristic value to a normal distribution. Third, the decreased profit resulting from a process shutdown, which is excluded in Taguchi’s on-line quality feedback control system, is considered in calculating the adjustment cost. The management cost and quality loss constitute the total quality cost. In the improved on-line quality feedback control system, the optimal measurement interval and optimal administrative boundary are calculated by minimizing the total quality cost using an iterative method. An example of the adjustment of the administrative boundary and the measurement interval for manufacturing automobile parts is presented to illustrate the effectiveness of the improved on-line system.

show abstract

A new desirability function-based method for correlated multiple response optimization

Salmasnia

Bashiri

2014

Int J Adv Manuf Technol

View full text Add to dashboard Cite

Optimization of process parameters using weighted convex loss functions

Cited by 5 publications

References 14 publications

Not all Failure Modes are Created Equal: Training Deep Neural Networks for Explicable (Mis)Classification

Not all Failure Modes are Created Equal: Training Deep Neural Networks for Explicable (Mis)Classification

Optimization of Taguchi’s on-line quality feedback control system

A new desirability function-based method for correlated multiple response optimization

Contact Info

Product

Resources

About