Jussi Keppo

Jussi Keppo

orcid

Research Focus & Metrics

Academic Career Length

12Years (2015-Present)

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

Representative Publications

View All Publications
16 May 2024 in International Journal of Computer Vision
Verified
Journal Article
Closed access
PDF file
MLMing Li
Ming Li
PZPan Zhou
Pan Zhou
JLJia-Wei Liu
Jia-Wei Liu
JKJussi Keppo
Jussi Keppo
MLMin Lin
Min Lin
XXXiangyu Xu
Xiangyu Xu

28 December 2023 in IEEE Transactions on Multimedia
Verified
Journal Article
Closed access
MLMing Li
Ming Li
HFHuazhu Fu
Huazhu Fu
SHShengfeng He
Shengfeng He
HFHehe Fan
Hehe Fan
JLJared Liu
Jared Liu
MSMike Zheng Shou
Mike Zheng Shou
Learning discriminative and robust representations is important for facial expression recognition (FER) due to subtly different emotional faces and their subjective annotations. Previous works usually address one representation solely because these two goals seem to be contradictory for optimization. Their performances inevitably suffer from challenges from the other representation. In this paper, by considering this problem from two novel perspectives, we demonstrate that discriminative and robust representations can be learned in a unified approach, i.e. DR-FER, and mutually benefit each other. Moreover, we make it with the supervision from only original annotations. Specifically, to learn discriminative representations, we propose performing masked image modeling (MIM) as an auxiliary task to force our network to discover expression-related facial areas. This is the first attempt to employ MIM to explore discriminative patterns in a self-supervised manner. To extract robust representations, we present a category-aware self-paced learning schedule to mine high-quality annotated ( easy ) expressions and incorrectly annotated ( hard ) counterparts. We further introduce a retrieval similarity-based relabeling strategy to correct hard expression annotations, exploiting them more effectively. By enhancing the discrimination ability of the FER classifier as a bridge, these two learning goals significantly strengthen each other. Extensive experiments on several popular benchmarks demonstrate the superior performance of our DR-FER. Moreover, thorough visualizations and extra experiments on manually annotation-corrupted datasets show that our approach successfully accomplishes learning both discriminative and robust representations simultaneously.

1 March 2019 in Journal of Economic Dynamics and Control
Verified
Journal Article
Closed access
MDMin Dai
Min Dai
SHShan Huang
Shan Huang
JKJussi Keppo
Jussi Keppo

1 April 2018 in Operations Research
Verified
Journal Article
Closed access
JNJie Ning
Jie Ning
VBVolodymyr Babich
Volodymyr Babich
JHJohn Handley
John Handley
JKJussi Keppo
Jussi Keppo
Managed print service (MPS) is a type of information technology infrastructure service that provides centralized management of companies’ printing device fleets. In this paper, we estimate the provider’s risk preference in MPS using a proprietary data set from Xerox Corporation. We adopt a structural approach in our empirical analysis by modeling the contracting and usage processes of MPS as a two-stage screening game and building econometric models based on the equilibrium contracts and print volumes. Our econometric models have a unique hierarchical structure that allows clustering of printers with the same contracts in the same company, thereby capturing the B2B nature of MPS. We find that Xerox exhibits risk aversion in MPS contracting and provide institutional details of Xerox’s commission holdback policy that may cause the observed risk aversion. In the counterfactual analysis, we demonstrate the significance of the provider’s risk aversion and the implications of the commission holdback policy on equilibrium contracts, the expected earnings of Xerox and customer companies, and their preferences for printer models. The E-companion is available at https://doi.org/10.1287/opre.2017.1673 .

1 July 2025 in Manufacturing & Service Operations Management
Verified
Journal Article
Closed access
ZCZhi Chen
Zhi Chen
JKJussi Keppo
Jussi Keppo
Problem definition: Many innovations today are data driven. To improve the algorithms of these products, firms make substantial investments in data collection. However, the data are limited for an individual firm, which caps the benefits of the algorithms. Data sharing among the firms can help alleviate this problem, but firms may be reluctant to share their proprietary data, because doing so may result in a loss of competitive advantage. Motivated by this tension, we ask the following two questions. First, when do firms voluntarily share their data and when are they reluctant to do so? Second, how does voluntary data sharing compare with the corresponding centralized data sharing (as if a government were to decide the level of data sharing to promote innovation)? Methodology/results: Using a game-theoretic model, we identify two key factors that determine the answers to the questions: (i) the relationship between firms’ data sets and (ii) the degree of uncertainty associated with the innovation. We find that firms voluntarily share data if their data sets are complements or if the uncertainty is high. Moreover, relative to the centralized data sharing, firms voluntarily share too little (respectively, much) data when their data sets are complements (respectively, redundant) and the uncertainty is moderate (respectively, high). Managerial implications: Our findings provide plausible explanations on data-sharing practices, for example, why firms in the autonomous vehicle industry voluntarily share their proprietary data. We also shed light on the antitrust issues associated with data sharing, where too much voluntary data sharing can reduce firms’ data collection incentives and stifle competition. Moreover, if the government considers subsidizing firms’ data collection efforts to accelerate innovation, cost subsidies are particularly effective when paired with mandatory data-sharing regulations. The government should exercise caution under voluntary data sharing because higher subsidies may not necessarily lead to higher innovation. Funding: Z. Chen’s research is supported by the Start-Up Grant from the National University of Singapore [Grant A-0003854-00-00]. Supplemental Material: The e-companion is available at https://doi.org/10.1287/msom.2023.0463 .

1 May 2023 in Management Science
Verified
Journal Article
Closed access
YJYanwei Jia
Yanwei Jia
JKJussi Keppo
Jussi Keppo
VSVille Satopää
Ville Satopää
Decision makers often ask experts to forecast a future state. Experts, however, can be biased. In the economics and psychology literature, one extensively studied behavioral bias is called herding. Under strong levels of herding, disclosure of public information may lower forecasting accuracy. This result, however, is derived only for point forecasts. In this paper, we consider experts’ probabilistic forecasts under herding, find a closed-form expression for the first two moments of a unique equilibrium forecast, and show that the experts report too similar locations and inflate the variance of their forecasts because of herding. Furthermore, we show that the negative externality of public information no longer holds. In addition to reacting to new information as expected, probabilistic forecasts contain more information about the experts’ full beliefs and interpersonal structure. This facilitates model estimation. To this end, we consider a one-shot setting with one forecast per expert and show that our model is identifiable up to an infinite number of solutions based on point forecasts but up to two solutions based on probabilistic forecasts. We then provide a Bayesian estimation procedure for these two solutions and apply it to economic forecasting data collected by the European Central Bank and the Federal Reserve Bank of Philadelphia. We find that, on average, the experts invest around 19% of their efforts into making similar forecasts. The level of herding shows an increasing trend from 1999 to 2007 but drops sharply during the financial crisis of 2007–2009 and then rises again until 2019.This paper was accepted by Yan Chen, behavioral economics and decision analysis.Supplemental Material: The electronic companion and data are available at https://doi.org/10.1287/mnsc.2022.4487 .