Adjusting the Outputs of a Classifier to New a Priori Probabilities: A Simple Procedure
Open accessThe DOI matches its Crossref record; 26 of the 26 references it cites were found in a registry or an index.
Abstract
It sometimes happens (for instance in case control studies) that a classifier is trained on a data set that does not reflect the true a priori probabilities of the target classes on real-world data. This may have a negative effect on the classification accuracy obtained on the real-world data set, especially when the classifier's decisions are based on the a posteriori probabilities of class membership. Indeed, in this case, the trained classifier provides estimates of the a posteriori probabilities that are not valid for this real-world data set (they rely on the a priori probabilities of the training set). Applying the classifier as is (without correcting its outputs with respect to these new conditions) on this new data set may thus be suboptimal. In this note, we present a simple iterative procedure for adjusting the outputs of the trained classifier with respect to these new a priori probabilities without having to refit the model, even when these probabilities are not known in advance. As a by-product, estimates of the new a priori probabilities are also obtained. This iterative algorithm is a straightforward instance of the expectation-maximization (EM) algorithm and is shown to maximize the likelihood of the new data. Thereafter, we discuss a statistical test that can be applied to decide if the a priori class probabilities have changed from the training set to the real-world data. The procedure is illustrated on different classification problems involving a multilayer neural network, and comparisons with a standard procedure for a priori probability estimation are provided. Our original method, based on the EM algorithm, is shown to be superior to the standard one for a priori probability estimation. Experimental results also indicate that the classifier with adjusted outputs always performs better than the original one in terms of classification accuracy, when the a priori probability conditions differ from the training set to the real-world data. The gain in classification accuracy can be significant.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“However, if wellcalibrated fraud probabilities are needed, then the posterior probabilities can be adjusted (Saerens et al., 2002).”
From Data engineering for fraud detection (2021) · cites this paper as Saerens et al., 2002
Citations
Cites
Maximum Likelihood from Incomplete Data Via the EM Algorithm
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977) · Journal of the Royal Statistical Society Series B: Statistical Methodology · doi:10.1111/j.2517-6161.1977.tb01600.x · Cited by 50,444
Probability, Random Variables, and Stochastic Processes.
Abrahams, J., & Papoulis, A. (1984) · Journal of the American Statistical Association · doi:10.2307/2288754 · Cited by 16,226
UCI Repository of machine learning databases
Blake, C. L. (1998) · Medical Entomology and Zoology · Cited by 10,582
The EM Algorithm and Extensions
Kushary, D. (1998) · Technometrics · doi:10.1080/00401706.1998.10485534 · Cited by 5,417
An Introduction to the Theory of Statistics
(1911) · Nature · doi:10.1038/087345a0 · Cited by 4,394
Probability, Random Variables, and Stochastic Processes
Papoulis, A., & Hoffman, J. G. (1967) · Physics Today · doi:10.1063/1.3034123 · Cited by 4,107
Journal of the Royal Statistical Society (B)
Fildes, R. (1993) · International Journal of Forecasting · doi:10.1016/0169-2070(93)90088-5 · Cited by 3,437
Text classification from labeled and unlabeled documents using EM
Nigam, K., Mccallum, A. K., Thrun, S., & Mitchell, T. (2000) · Machine Learning · doi:10.1023/a:1007692713085 · Cited by 2,736 · cited 1×
“For instance, Nigam et al. (2000) use the EM algorithm to learn classiers that take advantage of both labeled and unlabeled data.”
MethodDiscriminant Analysis and Statistical Pattern Recognition.
Hawkins, D. M., & McLachlan, G. J. (1993) · Journal of the American Statistical Association · doi:10.2307/2290355 · Cited by 2,633
Mixture models : inference and applications to clustering
McLachlan, G. J., & Basford, K. E. (1988) · M. Dekker eBooks · Cited by 2,094
Cited by
Items 201–300 of 362
Regularization-based methods for ordinal quantification
Bunse, M., Moreo, A., Sebastiani, F., & Senz, M. (2024) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-024-01067-2 · Cited by 7
“The Saerens-Latinne-Decaestecker (SLD) method (Saerens et al. 2002), also known as “EM-based quantification”, follows an iterative expectation maximization approach, which (i) leverages Bayes’ theorem in the E-step, and (ii) updates the prevalence estimates in the M-step.”
Enhancing Autonomous Driving Safety with Collision Scenario Integration
Wang, Z., Lan, S., Sun, X., Chang, N., Li, Z., Yu, Z., & Álvarez, J. M. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246713 · Cited by 12
“To balance training, we sample batches proportionally from both datasets and apply adaptive loss weights [26], [27] to mitigate data imbalance and domain gaps.”
Building a Spanish/Catalan health records corpus with very sparse protected information labelled
Herrera, S. M., & Borras, J. T. (2018) · Cited by 3
Accurately Quantifying under Score Variability
Maletzke, A. G., Reis, D. D., Hassan, W., & Batista, G. (2021) · Proceedings · doi:10.1109/icdm51629.2021.00149 · Cited by 3
Kontos, K., & Bontempi, G. (2009) · Dépôt institutionnel de l'Université libre de Bruxelles (Université Libre de Bruxelles) · Cited by 4
Miller, D. J., Pal, S., & Wang, Y. (2008) · Neurocomputing · doi:10.1016/j.neucom.2008.03.018 · Cited by 3
Transductive Methods for Distributed Ensemble Classification
Miller, D. J., & Pal, S. (2006) · Conference on Information Sciences and Systems · doi:10.1109/ciss.2006.286392 · Cited by 3
“[18] developed a transductive method to effectively correct mismatched class priors, considering the case of a single, standalone classifier.”
BackgroundMethodInfluentialSECS: Efficient Deep Stream Processing via Class Skew DichotomyPreprint
Feng, B., Wan, K., Yang, S., & Ding, Y. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1809.06691 · Cited by 6
“Rescaling is a topic in statistics [48].”
Sentiment Quantification of User-Generated Content
Sebastiani, F. (2017) · Springer eBooks · doi:10.1007/978-1-4614-7163-9_110170-1 · Cited by 3
Minimising quantifier variance under prior probability shiftPreprint
Tasche, D. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.08209 · Cited by 16
“[17] revived the interest in the ML estimator for the unknown prior class probabilities in the test by specifying the associated ‘expectation maximisation’ (EM) algorithm.”
MethodHandling New Class in Online Label Shift
Qian, Y.-Y., Bai, Y., Zhang, Z., Zhao, P., & Zhou, Z. (2025) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2025.3583138 · Cited by 3
“Unsupervised Stream with Online Label Shift: Label shift, as a common type of distribution change, has been extensively studied in the context of “one-step” adaptation [32], [33], where one aims to adapt the model from the source to the target distribution.”
Factorizable Joint Shift in Multinomial Classification
Tasche, D. (2022) · Machine Learning and Knowledge Extraction · doi:10.3390/make4030038 · Cited by 6
“[4] and Theorem 2 of Elkan [5] from prior probability shift to general dataset shift.”
BackgroundLightweight Conditional Model Extrapolation for Streaming Data under Class-Prior Shift
Tomaszewska, P., & Lampert, C. H. (2022) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr56361.2022.9956195 · Cited by 2
“The core concept to achieve both properties simultaneously is a form of classifier adaptation [22], [23], that we discuss in the following.”
BackgroundMethods for Learning to Quantify
Esuli, A., Fabris, A., Moreo, A., & Sebastiani, F. (2023) · The information retrieval series · doi:10.1007/978-3-031-20467-8_4 · Cited by 4
Doubly Flexible Estimation under Label Shift
Lee, S.-H., Ma, Y., & Zhao, J. (2024) · Journal of the American Statistical Association · doi:10.1080/01621459.2024.2321653 · Cited by 20
“Saerens, Latinne, and Decaestecker (2002) proposed a simple Expectation-Maximization (EM) (Dempster, Laird, and Rubin 1977) procedure, named maximum likelihood label shift (MLLS), to estimate q Y ( y ) assuming access to a classifier that outputs the true conditional probabilities of the population…”
Sparse joint shift in multinomial classification
Tasche, D. (2024) · Foundations of Data Science · doi:10.3934/fods.2024012 · Cited by 7
“3), with G = F and Xi = 1Ci , this choice leads to a conditional (on F) confusion matrix method for the posterior probabilities in the target domain for general SJS and the original confusion matrix method (Saerens et al., 2001) for prior probability shift (case F = {∅,Ω}).”
BackgroundMethodRamachandran, S., & Rattani, A. (2023) · Neurocomputing · doi:10.1016/j.neucom.2025.130024 · Cited by 4
Improved risk minimization algorithms for technology-assisted review
Molinari, A., Esuli, A., & Sebastiani, F. (2023) · Intelligent Systems with Applications · doi:10.1016/j.iswa.2023.200209 · Cited by 5
Modeling Website Topic Cohesion at Scale to Improve Webpage Classification
Eswaran, D., Bennett, P. N., & Pfeiffer, J. J. (2015) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/2766462.2767834 · Cited by 2
“Here, we highlight previous work that has discussed adjusting posterior predictions when either the priors do not match [10] or attribute distributions vary [14].”
BackgroundRobustness Evaluation of Deep Learning Models Based on Local Prediction Consistency
Shi, Z., Zhong, C., Yokota, Y., Xia, W., & Sun, J. (2019) · International Conference on Machine Learning and Applications · doi:10.1109/icmla.2019.00224 · Cited by 2
“propose kernel mean matching to reweight training points [3].”
BackgroundMethodAn Approach to Identifying and Quantifying Bias in Biomedical Data
Kaluza, M. C. D. P., Jain, S., & Radivojac, P. (2022) · Pacific Symposium on Biocomputing (PSB) · doi:10.1142/9789811270611_0029 · Cited by 4
Long, S. S., Bruzzone, M., Mitropanopoulos, S., Kalamangalam, G. P., & Gunduz, A. (2023) · NeuroImage · doi:10.1016/j.neuroimage.2023.119961 · Cited by 2
Learning to Quantify: Estimating Class Prevalence via Supervised Learning
Moreo, A., & Sebastiani, F. (2019) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3331184.3331389 · Cited by 3
“(1) Introduction / Motivation (a) Solving quantification via “Classify and Count” (b) Concept drift and distribution drift [24, 31] (c) Vapnik’s principle (d) The “paradox of quantification” (2) Applications of quantification in machine learning, data mining, text mining, and NLP [14] (a) Sentiment quantification [11]…”
MethodPredictive Analytics for Fraud Detection
(2015) · Wiley eBooks · doi:10.1002/9781119146841.ch4 · Cited by 2
Graph-based approaches for semi-supervised and cross-domain sentiment analysis
Ponomareva, N. (2014) · Wolverhampton Intellectual Repository and E-Theses (University of Wolverhampton) · Cited by 2
Automatic Annotation of Spoken Language Using Out-of-Domain Resources and Domain Adaptation
Margolis, A. (2011) · Cited by 3
Constraint-Based, Transductive Learning for Distributed Ensemble Classification
Miller, D., Pal, S., & Wang, Y. (2006) · Machine learning for signal processing ... · doi:10.1109/mlsp.2006.275514 · Cited by 2
“This was not recognized in [15], where only local convergence was claimed.”
BackgroundMethodMortgage Prepayment Rate Estimation with Machine Learning
Saito, T. (2018) · Research Repository (Delft University of Technology) · Cited by 5
Domain Adaptation in Highly Imbalanced and Overlapping DatasetsPreprint
Ber, R. I., & Haramaty, T. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2005.03585 · Cited by 2
“In order to calibrate the classifier to the target set, and thus improve its classification quality, a known procedure was followed [12].”
MethodSentiment Quantification of User-Generated Content
Sebastiani, F. (2018) · Springer eBooks · doi:10.1007/978-1-4939-7131-2_110170 · Cited by 4
Re-Assessing the" Classify and Count" Quantification Method
Moreo, A., & Sebastiani, F. (2020) · ISTI Open Portal · doi:10.1007/978-3-030-72240-1_6 · Cited by 14
Robust and efficient airplane cockpit video coding leveraging temporal redundancy
Mitrica, I., Fiandrotti, A., Ruellan, C., & Cagnazzo, M. (2024) · Multimedia Tools and Applications · doi:10.1007/s11042-024-18755-2 · Cited by 1
Adapting to Shifting Correlations with Unlabeled Data Calibration
Nguyen, M., Wang, A. Q., Kim, H., & Sabuncu, M. R. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-73021-4_14 · Cited by 5
Enhancing Test-Time Personalized Federated Learning via Dual Disentangled Factors
Deng, X., Chen, K., Zhang, Y. L., Yi, C., & Peng, Q. (2025) · International Conference on Advanced Algorithms and Control Engineering (ICAACE) · doi:10.1109/icaace65325.2025.11019778 · Cited by 1
“Additionally, we compare with EM [25], which uses expectation maximization to adjust label priors in an unsupervised manner.”
MetaQuant: A framework for metaheuristic based quantification
Serapião, A. B. D. S., Donyavi, Z., & Batista, G. (2025) · Applied Soft Computing · doi:10.1016/j.asoc.2025.113527 · Cited by 1
Kaluza, M. C. D. P., Tholeti, T., Chen, Y., Baeza-Yates, R. A., Predrag Radivojac, A., & Jain, S. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3711896.3736846 · Cited by 1
“1 Covariate-shift [43] and label-shift [15, 24, 40, 47] are two dominant selection bias models.”
M atch : A Maximum-Likelihood Approach for Classification under Label Shift
Donyavi, Z., Li, F., Zhang, Y., Silva, D. F., & Batista, G. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3711896.3737040 · Cited by 1
“The second is an update equation based on the Bayes rule [19] that operates over each instance’s posterior probabilities.”
InfluentialNormative Alignment of Recommender Systems via Internal Label Shift
Kruse, J., Lindskow, K., Andersen, M. R., Shimizu, R., McAuley, J., Mattei, P., & Frellsen, J. (2025) · ACM Conference on Recommender Systems (RecSys) · doi:10.1145/3705328.3759309 · Cited by 2
“This weight can be viewed as a reweighting term that adjusts the recommendations to align with the normative marginal distribution ˜ 𝑝 ( 𝑐 ) .”
InfluentialEfficient quantification on large-scale networks
Micheli, A., Moreo, A., Podda, M., Sebastiani, F., Simoni, W., & Tortorella, D. (2025) · Machine Learning · doi:10.1007/s10994-025-06915-w · Cited by 2
“We do so by adapting the Saerens-Latinne-Decaestecker method (Saerens et al., 2002) ( SLD ) to our setting.”
Towards Label Shift Adaptation for Robust IoT Device Identification
Azizi, S., Calderan, F. V., Okui, N., Nakahara, M., Kubota, A., Quiles, M. G., Batista, G., & Gharakheili, H. H. (2025) · ACM Workshop on CPS&IoT Security and Privacy (CPSIoTSec) · doi:10.1145/3733801.3764204 · Cited by 2
FMAPLS: Bayesian Label Shift Estimation Based on Dynamic Dirichlet Parameter Adaptation
Hu, J., & Barria, J. A. (2025) · IEEE Signal Processing Letters · doi:10.1109/lsp.2025.3622531 · Cited by 2
“Restrictions apply. in comparison with MLLS [3], BBSE [4], RLLS [5], and MAPLS [8].”
Quantification via gaussian latent space representations
Pérez-Mon, O., del Coz, J. J., & González, P. (2026) · Neural Networks · doi:10.1016/j.neunet.2026.108886 · Cited by 3
Contrastive Learning for Domain Adaptation Under Domain Shift: A Critical Survey and Taxonomy
Nabi, S. T. (2026) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-026-10603-w · Cited by 4
Mitigating distribution shift via adaptive reweighting for robust visual question answering
Song, X., Wang, R., Xu, H., Yu, C., Wan, Z., Ma, T., Gao, C., & Sang, N. (2026) · Pattern Recognition · doi:10.1016/j.patcog.2026.113945 · Cited by 1
Jahangeer, A. S., & Pichai, S. (2026) · Future Internet · doi:10.3390/fi18090480 · Cited by 1
Miller, D. J., Raghuram, J., Kesidis, G., & Collins, C. M. (2012) · Neural Computation · doi:10.1162/neco_a_00284 · Cited by 3
“There has also been a significant amount of followup interest in this area, e.g. [7] [13] [15] [22] [24] [26] [28].”
BackgroundOrdinal Quantification Through Regularization
Bunse, M., Moreo, A., Sebastiani, F., & Senz, M. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-26419-1_3 · Cited by 5
“EM-based quantification, also known as the Saerens-Latinne-Decaestecker (SLD) method [33], follows an expectation maximization approach, which (i) leverages Bayes’ theorem in the E-step, and (ii) updates the prevalence estimates in the M-step.”
BackgroundMethodEsuli, A., Fabris, A., Moreo, A., & Sebastiani, F. (2023) · The information retrieval series · doi:10.1007/978-3-031-20467-8_1 · Cited by 1
A Noise-Adaptive Machine Learning Framework for Optimizing User Grouping in Dynamic IM-OFDMA Systems
Ay, F., Althunibat, S., Qaraqe, K. A., & Kurban, H. (2024) · IEEE Transactions on Communications · doi:10.1109/tcomm.2024.3455243 · Cited by 1
“In addressing label shift, [55] developed a novel iterative procedure that employs the expectation-maximization algorithm.”
Sequential label shift detection in classification data: An application to dengue fever
Evans, C., & G’Sell, M. (2024) · PLoS ONE · doi:10.1371/journal.pone.0310194 · Cited by 1
“Because it frequently arises in practice, we will focus on the label shift setting [10, 11], which has received recent attention in the machine learning literature [8, 9, 12, 13].”
BackgroundTong, Y., & Sundaram, S. (2024) · Allerton Conference on Communication, Control, and Computing · doi:10.1109/allerton63246.2024.10735314
“We use the idea of adjusting a classifier to a new priori probability from [24] and recursively apply it to the outcomes of the classifier.”
MethodAdapting to Generalized Online Label Shift by Invariant Representation Learning
Qian, Y.-Y., Wang, Y.-H., Zhang, Z., Jiang, Y., & Zhou, Z. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3690624.3709182 · Cited by 1
QuAcc: Using Quantification to Predict Classifier Accuracy Under Prior Probability Shift
Volpi, L., Moreo, A., & Sebastiani, F. (2025) · Intelligenza Artificiale · doi:10.1177/17248035251338347 · Cited by 1
“…trivial “ Classify and Count ” quanti fi cation method (that simply counts the number of times a class label has been attributed by the clas-si fi er), with those obtained by using quanti fi cation methods (described below) robust to PPS such as SLD (Saerens et al., 2002) or KDEy (Moreo et al., 2025).”
Integrating Weight with Ensemble to Handle Changes in Class Distribution
Limsetto, N., & Waiyamai, K. (2014) · Lecture notes in computer science · doi:10.1007/978-3-319-08979-9_8 · Cited by 1
Minimax strategies for training classifiers under unknown priors
Aláiz-Rodríguez, R., & Cid‐Sueiro, J. (2003) · Machine learning for signal processing ... · doi:10.1109/nnsp.2002.1030036 · Cited by 1
On the Complexity of Learning a Class Ratio from Unlabeled Data
Fish, B., & Reyzin, L. (2020) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.12013 · Cited by 1
“This problem may also arise when attempting to correct for differences between training and testing distributions (Du Plessis & Sugiyama, 2014; Saerens et al., 2002).”
BackgroundOptimizing class priors to improve the detection of social signals in audio data
Gosztolya, G. (2021) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2021.104541 · Cited by 3
Ensembles of Classifiers and Quantifiers with Data Fusion for Quantification Learning
Serapião, A. B. D. S., Donyavi, Z., & Batista, G. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-45275-8_1 · Cited by 1
Learning to Abstain via Curve Optimization.Preprint
Alexandari, A. M., Shrikumar, A., & Kundaje, A. B. (2018) · arXiv (Cornell University) · Cited by 1
A Label Proportions Estimation Technique for Adversarial Domain Adaptation in Text ClassificationPreprint
Chen, Z., Singla, K., Atkins, D. C., Imel, Z. E., & Narayanan, S. S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2003.07444 · Cited by 1
“It is worth mentioning that given accurate label shift estimate, we can also improve the classification accuracy by the prior probability adjustment [16] re-weighting the class importance in C.”
BackgroundBaesens, B., Rösch, D., & Scheule, H. (2016) · doi:10.1002/9781119449560.ch8 · Cited by 1
Profit‐Driven Analytical Techniques
Verbeke, W., Baesens, B., & Bravo, C. (2017) · Wiley eBooks · doi:10.1002/9781119443179.ch5 · Cited by 1
Dealing with Unknown Priors in Supervised Classification
Saerens, M., Souchon, N., Renders, J.-M., & Decaestecker, C. (2004)
Self-citationAláiz-Rodríguez, R., & Cid‐Sueiro, J. (2002) · Lecture notes in computer science · doi:10.1007/3-540-46084-5_66
Performance Bounds for Pairwise Entity ResolutionPreprint
Barnes, M., Miller, K., & Dubrawski, A. W. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1509.03302
“We refer the reader to state-of-the-art results for class prior estimation [10 , 11].”
MethodDecision-Making with Unknown Priors in Supervised Classication
Marco, S., & Decaestecker, C. (2010) · Cited by 2
Self-citationMonaural Speech Segregation in Reverberant Environments
Jin, Z. (2010) · OhioLink ETD Center (Ohio Library and Information Network)
Counting on Uncertainty: Obtaining Fish Counts from Machine Learning Decisions
Boom, B. (2016) · Intelligent systems reference library · doi:10.1007/978-3-319-30208-9_15
非定常環境下での学習:共変量シフト適応,クラスバランス変化適応,変化検知
将, 杉., 誠, 山., ドゥ・プレシ, マ., & リウ, ソ. (2014) · 日本統計学会誌 · doi:10.11329/jjssj.44.113
Semi-supervised sufficient dimension reduction under class-prior change
Kawakubo, H., & Sugiyama, M. (2016) · Conference on Technologies and Applications of Artificial Intelligence (TAAI) · doi:10.1109/taai.2016.7880182
“For example, the method based on the expectation-maximization algorithm [15], distribution matching under the Kullback-Leibler divergence [16], the Pearson divergence [16], the í µí°¿ 2 -distance [17], the maximum mean discrepancy [18], and the energy distance (ED) [19].”
MethodDensity-Difference EstimationPreprint
Sugiyama, M., Kanamori, T., Suzuki, T., du Plessis, M. C., Liu, S., & Takeuchi, I. (2012) · arXiv (Cornell University)
Unsupervised RecalibrationPreprint
Ziegler, A., & Czyż, P. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1908.09157
“2 of Saerens et al. (2001)).”
Tutorial: Supervised Learning for Prevalence Estimation
Moreo, A., & Sebastiani, F. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-27629-4_3
Towards a New Understanding of the Training of Neural Networks with Mislabeled Training DataPreprint
Gish, H., Silovský, J., Sung, M.-L., Siu, M.-H., Hartmann, W., & Jiang, Z. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1909.09136
EM with Bias-Corrected Calibration is Hard-To-Beat at Label Shift AdaptationPreprint
Alexandari, A. M., Kundaje, A. B., & Shrikumar, A. (2019) · arXiv (Cornell University)
Towards a New Understanding of the Training of Neural Networks with Mislabeled Training Data
Gish, H., Silovský, J., Sung, M.-L., Siu, M.-H., Hartmann, W., & Jiang, Z. (2020) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp40776.2020.9053014
Extending Adversarial Attacks to Produce Adversarial Class Probability DistributionsPreprint
Vadillo, J., Santana, R., & Lozano, J. A. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2004.06383 · Cited by 2
“Indeed, it has been shown that such drifts can degrade the performance of the classifiers (Lipton et al., 2018; Vucetic and Obradovic, 2001; Saerens et al., 2002), or imply ethical issues when such changes cause the predictions of the models to be biased or unfair (Vucetic and Obradovic, 2001; Biswas and Mukherjee, 20…”
BackgroundInfluentialSequential changepoint detection for label shift in classificationPreprint
Evans, C., & G’Sell, M. (2020) · arXiv (Cornell University)
A semi-supervised approach to growing classification trees
Santhiappan, S., & Ravindran, B. (2020) · Joint International Conference on Data Science & Management of Data (CODS-COMAD) · doi:10.1145/3430984.3431009
“An EM-based approach [20] that alternates between estimating the class proportions from per-instance predictions and correcting the predictions using the estimated class proportions as priors.”
MethodInference with generalizable classifier predictionsPreprint
Evans, C., Weinberg, Z. Y., Puthenveedu, M. A., & G’Sell, M. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.07623 · Cited by 2
“Fortunately, there are a variety of methods for detecting and correcting for label shift, which allow predictions to be easily adjusted in this setting (Saerens et al., 2002; Storkey, 2009; Lipton et al., 2018; Garg et al., 2020).”
MethodInfluentialDomain Adaptation for Autoencoder-Based End-to-End Communication Over Wireless Channels.Preprint
Raghuram, J., Zeng, Y., Martí, D. G., Jha, S., Banerjee, S., Widmer, J., & Ortiz, R. R. (2021) · arXiv (Cornell University)
A Symmetric Loss Perspective of Reliable Machine Learning
Charoenphakdee, N., Lee, J., & Sugiyama, M. (2026) · Fields Institute communications · doi:10.1007/978-3-032-03844-9_5 · Cited by 1
Evaluation of Quantification Algorithms
Esuli, A., Fabris, A., Moreo, A., & Sebastiani, F. (2023) · The information retrieval series · doi:10.1007/978-3-031-20467-8_3
Esuli, A., Fabris, A., Moreo, A., & Sebastiani, F. (2023) · The information retrieval series · doi:10.1007/978-3-031-20467-8_5
Esuli, A., Fabris, A., Moreo, A., & Sebastiani, F. (2023) · The information retrieval series · doi:10.1007/978-3-031-20467-8_7
Esuli, A., Fabris, A., Moreo, A., & Sebastiani, F. (2023) · The information retrieval series · doi:10.1007/978-3-031-20467-8_6
Applications of Quantification
Esuli, A., Fabris, A., Moreo, A., & Sebastiani, F. (2023) · The information retrieval series · doi:10.1007/978-3-031-20467-8_2
Interpreting Deep Text Quantification Models
Bang, Y., Khaleel, M., & Tavanapong, W. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-39821-6_25
Dynamic Quantification With Constrained Error Under Unknown General Dataset Shift
Denham, B., Lai, E. M.-K., Sinha, R., & Naeem, M. A. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2023.3349286
“The Fisher information of EM’s maximum-likelihood estimates can produce confidence intervals [5], [7], [21] that adequately serve as prediction intervals [21].”
BackgroundMethodInfluentialEfficient data integration under prior probability shift
Huang, M., Qin, J., & Huang, C. (2024) · Biometrics · doi:10.1093/biomtc/ujae035 · Cited by 1
Estimation of class priors for improving classification accuracy during deployment
McIntosh, B., Daba, N., & Mahalanobis, A. (2024) · Elsevier eBooks · doi:10.1016/b978-0-44-319077-3.00010-9
Adapting Classifiers to Changing Class Priors During Deployment
Daba, N., McIntosh, B., & Mahalanobis, A. (2024) · Communications in computer and information science · doi:10.1007/978-3-031-74640-6_35
Evaluating Sentiment Quantification Methods in Brazilian Portuguese Corpora
Silva, L. N. D. S., Silva, D. F., & Caseli, H. D. M. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-79035-5_16
Cuevas, F. G., & Schmid, H. (2025) · Lecture notes in computer science · doi:10.1007/978-981-97-8702-9_32
Combining adaptive local aggregation average and test-time energy adaptation for federated learning
Liao, J., Yi, C., Chen, K., & Peng, Q. (2025) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-025-02580-6
“EM [25] is a simple iterative procedure for adjusting the outputs of the trained classifier with respect to new priori probabilities without having to refit the model.”
Jacknis, N. J., & Maddala, K. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5240826
Class-Aware Sinkhorn-DRO for Few-Shot Domain Adaptation
Chen, T. Y. (2025) · Communications in computer and information science · doi:10.1007/978-981-95-0988-1_13
An Adaptive Transfer Learning Perspective on Classification in Nonstationary Environments
Reeve, H. W. J. (2025) · Statistical Science · doi:10.1214/25-sts1002
Castro, L., Rueda, M. D. M., Sánchez-Cantalejo, C., Ferri, R., & Cabrera‐León, A. (2024) · Figshare · doi:10.6084/m9.figshare.26681113
Metadata
- Authors
- Saerens, M., Latinne, P., & Decaestecker, C.
- Affiliations
- Université Libre de Bruxelles (BE)
- Venue
- Neural Computation
- Publisher
- MIT Press
- Published
- 2002-01-01
- Volume
- 14
- Issue
- 1
- Pages
- 21-41
- ISSN
- 0899-7667 (ISSN-L)
- 1530-888X
- OpenAlex
- W2141253686
- Cited by
- 361
- References
- 26
- Citations by year
- 2026: 57 · 2025: 28 · 2024: 26 · 2023: 24 · 2022: 23 · 2021: 32 · 2020: 27 · 2019: 19 · 2018: 15 · 2017: 9 · 2016: 17 · 2015: 11 · 2014: 9 · 2013: 4 · 2012: 6
- FWCI
- 4.14
- Type
- journal article
- Language
- en
- Open access
- green
- dipot.ulb.ac.be
- Licence
- other-oa
- Locations
- Neural Computation (publishedVersion)
- PubMed (publishedVersion)
- Digital Access to Libraries (Université catholique de Louvain (UCL), l'Université de Namur (UNamur) and the Université Saint-Louis (USL-B)) (submittedVersion)
- Dépôt institutionnel de l'Université libre de Bruxelles (Université Libre de Bruxelles) (submittedVersion, other-oa)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Imbalanced Data Classification Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
- Physical Sciences › Computer Science › Artificial Intelligence › Data Stream Mining Techniques
- Keywords
- posterior probability, expectation-maximization algorithm, likelihood maximization, multilayer neural networks, case-control study, classification accuracy
- MeSH terms
- Classification, Probability, Probability Theory
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.