Yang Kang

Dr. Yang Kang

orcid

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 May 2021 in Operations Research
Verified
Journal Article
Closed access
JBJose Blanchet
Jose Blanchet
YKYang Kang
Yang Kang
Financial institutions make decisions according to a model of uncertainty. At the same time, regulators often evaluate the risk exposure of these institutions using a model of uncertainty, which is often different from the one used by the institutions. How can one incorporate both views into a single framework? This paper provides such a framework. It quantifies the impact of the misspecification inherent to the financial institution data-driven model via the introduction of an adversarial player. The adversary replaces the institution's generated scenarios by the regulator's scenarios subject to a budget constraint and a cost that measures the distance between the two sets of scenarios (using what in statistics is known as the Wasserstein distance). This paper also harnesses statistical theory to make inference about the size of the estimated error when the sample sizes (both of the institution and the regulator) are large. The framework is explained more broadly in the context of distributionally robust optimization (a class of perfect information games, in which decisions are taken against an adversary that perturbs a baseline distribution).

1 September 2016 in Landscape and Urban Planning
Verified
Journal Article
Closed access
Zeyuan QiuZeyuan Qiu
Zeyuan Qiu
MDMichael G. Dosskey
Michael G. Dosskey
YKYang Kang
Yang Kang

9 November 2020 in Sustainability
Verified
Journal Article
Open access
PDF file
LXLin Xie
Lin Xie
Zeyuan QiuZeyuan Qiu
Zeyuan Qiu
LYLiangzhi You
Liangzhi You
YKYang Kang
Yang Kang
Agrochemicals are overused in China. One strategy to reduce agrochemical use is to increase farm size because of the potential effect of economy of scale. Existing studies at a micro scale present mixed and often conflicting results on the relationship between agrochemical use and farm size. This study aimed to assess that relationship from a macro perspective using an aggregated panel dataset in 30 provinces in China from 2009 to 2016. The empirical results confirm the existence of both economy and diseconomy of scale effects on agrochemical use in China. The agrochemical application rates decreased as the proportion of farms between 0.667 and 2 ha increased. The diseconomy of scale existed when significantly larger farms, such as the farms larger than 3.34 ha, continued to emerge. Given the fact that 78.6% of farms are under 0.667 ha in China, our results suggest that the reduction strategy based on only expanding farm size might achieve some initial success in reducing agrochemical use, but the effect would fade away and be reversed as significantly large farms continue to emerge. These results have significant policy implications as China is proactively developing and implementing various policies and strategies to modernize its agriculture toward achieving its sustainability goals.

29 March 2019 in arXiv
Verified
Preprint
Open access
PDF file
JBJose Blanchet
Jose Blanchet
YKYang Kang
Yang Kang
FZFan Zhang
Fan Zhang
KMKarthyek Murthy
Karthyek Murthy
Recently, (Blanchet, Kang, and Murhy 2016, and Blanchet, and Kang 2017) showed that several machine learning algorithms, such as square-root Lasso, Support Vector Machines, and regularized logistic regression, among many others, can be represented exactly as distributionally robust optimization (DRO) problems. The distributional uncertainty is defined as a neighborhood centered at the empirical distribution. We propose a methodology which learns such neighborhood in a natural data-driven way. We show rigorously that our framework encompasses adaptive regularization as a particular case. Moreover, we demonstrate empirically that our proposed methodology is able to improve upon a wide range of popular machine learning estimators.

21 March 2019 in arXiv
Verified
Preprint
Open access
PDF file
JBJose Blanchet
Jose Blanchet
YKYang Kang
Yang Kang
We propose a novel method for semi-supervised learning (SSL) based on data-driven distributionally robust optimization (DRO) using optimal transport metrics. Our proposed method enhances generalization error by using the unlabeled data to restrict the support of the worst case distribution in our DRO formulation. We enable the implementation of our DRO formulation by proposing a stochastic gradient descent algorithm which allows to easily implement the training procedure. We demonstrate that our Semi-supervised DRO method is able to improve the generalization error over natural supervised procedures and state-of-the-art SSL estimators. Finally, we include a discussion on the large sample behavior of the optimal uncertainty region in the DRO formulation. Our discussion exposes important aspects such as the role of dimension reduction in SSL.

11 May 2017 in arXiv
Verified
Preprint
Open access
PDF file
JBJose Blanchet
Jose Blanchet
YKYang Kang
Yang Kang
Regularized estimators in the context of group variables have been applied successfully in model and feature selection in order to preserve interpretability. We formulate a Distributionally Robust Optimization (DRO) problem which recovers popular estimators, such as Group Square Root Lasso (GSRL). Our DRO formulation allows us to interpret GSRL as a game, in which we learn a regression parameter while an adversary chooses a perturbation of the data. We wish to pick the parameter to minimize the expected loss under any plausible model chosen by the adversary - who, on the other hand, wishes to increase the expected loss. The regularization parameter turns out to be precisely determined by the amount of perturbation on the training data allowed by the adversary. In this paper, we introduce a data-driven (statistical) criterion for the optimal choice of regularization, which we evaluate asymptotically, in closed form, as the size of the training set increases. Our easy-to-evaluate regularization formula is compared against cross-validation, showing good (sometimes superior) performance.