191,406 Hits in 4.9 sec

Guaranteed Validity for Empirical Approaches to Adaptive Data Analysis [article]

Ryan Rogers, Aaron Roth, Adam Smith, Nathan Srebro, Om Thakkar, Blake Woodworth
2020 arXiv   pre-print
Our main contribution is to design a framework for providing valid, instance-specific confidence intervals for point estimates that can be generated by heuristics.  ...  We design a general framework for answering adaptive statistical queries that focuses on providing explicit confidence intervals along with point estimates.  ...  Acknowledgements The authors would like to thank Omer Tamuz for helpful comments regarding a conjecture that existed in a prior version of this work.  ... 
arXiv:1906.09231v2 fatcat:suzno2narbg73moropqjv4vxry

The Everlasting Database: Statistical Validity at a Fair Price [article]

Blake Woodworth, Vitaly Feldman, Saharon Rosset, Nathan Srebro
2019 arXiv   pre-print
Crucially, we guarantee statistical validity without any assumptions on how the queries are generated.  ...  The problem of handling adaptivity in data analysis, intentional or not, permeates a variety of fields, including test-set overfitting in ML challenges and the accumulation of invalid scientific discoveries  ...  Preserving statistical validity in adaptive data analysis. CoRR, abs/1411.2664, 2014.  ... 
arXiv:1803.04307v3 fatcat:u2uhicg2uzht5atwjnpvadcvjq

Robust domain-adaptive discriminant analysis

Wouter Kouw, Marco Loog
2021 Pattern Recognition Letters  
a non-adaptive classifier without having to rely on the validity of strong assumptions.  ...  Consider a domain-adaptive supervised learning setting, where a classifier learns from labeled data in a source domain and unlabeled data in a target domain to predict the corresponding target labels.  ...  Since there is no labeled target data available for validation, we set the regularization parameter to a small value, namely 0.01.  ... 
doi:10.1016/j.patrec.2021.05.005 fatcat:dtquptrsrvcupbtq37s5xk4w4u


S. Üreten, M. Eisenmann, T. Nelius, E. Garrelts, D. Krause, S. Matthiesen
2020 Proceedings of the Design Society: DESIGN Conference  
AbstractThe requirements on validity for studies in design research are very high.  ...  Resulting main challenges are to find a suitable task, to operationalise the variables and to deal with a high analysis effort per participant.  ...  Acknowledgements We would like to thank all studies' participants, data collectors and analysers, who enabled a data basis rich enough to be analysed systematically and thus contributed to empirical design  ... 
doi:10.1017/dsd.2020.280 fatcat:4lcli2udj5b3zlffac6ps524gm

A Logical Approach for Empirical Risk Minimization in Machine Learning for Data Stratification

Taiwo, O. O., Awodele O., Kuyoro, S. O.
2017 Research Journal of Mathematics and Computer Science  
Hence the logical approach to determine the empirical risk minimization function for in machine learning for CML data stratification is shown in figure 1 .  ...  Methodology A logical approach for Empirical Risk Minimization in machine learning for Data Stratification is developed to aid a better grouping of Chronic Myeloid Leukemia disease dataset.  ... 
doi:10.28933/rjmcs-2017-10-3002 fatcat:nidkvqzadbhxfbhqzkyutiu5fe

Understanding and Comparing Approaches for Performance Engineering of Self-adaptive Systems Based on Queuing Networks

Davide Arcelli
2021 Journal of Ubiquitous Systems and Pervasive Networks  
Enabling self-adaptation within hardware/software systems is a complex task, mainly due to environment uncertainty that has to be faced while the system is providing its functionalities.  ...  Besides, non-functional goals that have to be met by the system may be introduced, defining Quality-of-Service (QoS) requirements which drive the adaptation.  ...  The author would like to thank Dott. Davide Di Ruscio for his precious suggestions concerning paper presentation.  ... 
doi:10.5383/juspn.14.02.004 fatcat:aj24mmmfcrfolk3mu5hpzrflza

PAC Prediction Sets for Meta-Learning [article]

Sangdon Park and Edgar Dobriban and Insup Lee and Osbert Bastani
2022 arXiv   pre-print
We study this problem in the context of meta learning, where the goal is to quickly adapt a predictor to new tasks.  ...  In particular, our prediction sets satisfy the PAC guarantee while having smaller size compared to other baselines that also satisfy this guarantee.  ...  A limitation of the current approach is that it requires enough calibration datapoints to satisfy the PAC guarantee, see [3] for an analysis.  ... 
arXiv:2207.02440v1 fatcat:baxqc6r7dbbfpbiorgjmbj3g5i

Preserving Statistical Validity in Adaptive Data Analysis [article]

Cynthia Dwork and Vitaly Feldman and Moritz Hardt and Toniann Pitassi and Omer Reingold and Aaron Roth
2016 arXiv   pre-print
In this work we initiate a principled study of how to guarantee the validity of statistical inference in adaptive data analysis.  ...  A great deal of effort has been devoted to reducing the risk of spurious scientific discoveries, from the use of sophisticated validation techniques, to deep statistical methods for controlling the false  ...  Acknowledgements We would like to thank Sanjeev Arora, Nina Balcan, Avrim Blum, Dean Foster, Michael Kearns, Jon Kleinberg, Sasha Rakhlin, and Jon Ullman for enlightening discussions and helpful comments  ... 
arXiv:1411.2664v3 fatcat:q2oslwlcx5co7fnjdybn4n5pse

Claims and Evidence for Architecture-Based Self-adaptation: A Systematic Literature Review [chapter]

Danny Weyns, Tanvir Ahmad
2013 Lecture Notes in Computer Science  
studies used for data collection.  ...  Architecture-based self-adaptation is a promising approach to tackle these challenges.  ...  However, reflecting on the results and analysis of our study, we conclude that there are opportunities for improving coherence in research to move the field forward.  ... 
doi:10.1007/978-3-642-39031-9_22 fatcat:svpmyzvjvfbsdi5jqsa3i2ktki

On the Fundamental Tautology of Validating Data-Driven Models and Simulations [chapter]

John Michopoulos, Sam Lambrakos
2005 Lecture Notes in Computer Science  
This paper discusses the fundamental irrelevance of conventional validation procedures with respect to data-driven models and simulations.  ...  need for a critical evaluation of paradigms underlying Qualification, Validation and Verification (QV&V).  ...  Darema for her support of our DDDAS-related activities. Partial support from NRL's 6.1 core-funding is also greatly acknowledged.  ... 
doi:10.1007/11428848_95 fatcat:xy2h6y24zzhadpse7cec3ot45i

Target Robust Discriminant Analysis [article]

Wouter M. Kouw, Marco Loog
2021 arXiv   pre-print
We construct robust parameter estimators for discriminant analysis that guarantee performance improvements of the adaptive classifier over the non-adaptive source classifier.  ...  In practice, the data distribution at test time often differs, to a smaller or larger extent, from that of the original training data.  ...  Since no labelled target data is available for validation, the regularization parameter was set to 0.01 for logistic and 0.01n for quadratic losses.  ... 
arXiv:1806.09463v2 fatcat:iivrruu3xraxzku7wzpvhqsbde

Preserving Statistical Validity in Adaptive Data Analysis

Cynthia Dwork, Vitaly Feldman, Moritz Hardt, Toniann Pitassi, Omer Reingold, Aaron Leon Roth
2015 Proceedings of the Forty-Seventh Annual ACM on Symposium on Theory of Computing - STOC '15  
In this work we initiate a principled study of how to guarantee the validity of statistical inference in adaptive data analysis.  ...  A great deal of effort has been devoted to reducing the risk of spurious scientific discoveries, from the use of sophisticated validation techniques, to deep statistical methods for controlling the false  ...  Acknowledgements We would like to particularly thank Jon Ullman for many enlightening discussions about this work.  ... 
doi:10.1145/2746539.2746580 dblp:conf/stoc/DworkFHPRR15 fatcat:4bwgkro545gonpjd72m7pvacre

Distribution-Free Federated Learning with Conformal Predictions [article]

Charles Lu, Jayasheree Kalpathy-Cramer
2022 arXiv   pre-print
In this paper, we propose to address these challenges by incorporating an adaptive conformal framework into federated learning to ensure distribution-free prediction sets that provide coverage guarantees  ...  Empirical results on the MedMNIST medical imaging benchmark demonstrate our federated method provides tighter coverage over local conformal predictions on 6 different medical imaging datasets for 2D and  ...  Ethical approval was not required as confirmed by the license attached with the open access data.  ... 
arXiv:2110.07661v2 fatcat:rcyyd74vlnfojfoqp3rgxzkblm

A Triple-A supply chain measurement model: validation and analysis

Juan A. Marin-Garcia, Rafaela Alfalla-Luque, Jose A.D. Machuca
2018 International Journal of Physical Distribution & Logistics Management  
scale for future empirical research and industrial applications.  ...  Design/methodology/approach -Following a literature review, Triple-A SC variables (agility, alignment and adaptability) are conceptualized and a list of possible items is created for their measurement.  ...  The findings provide guidance for empirical research with parsimonious data.  ... 
doi:10.1108/ijpdlm-06-2018-0233 fatcat:73tjj3z3krh47if7ifbd6lmpoe

Adaptive Kalman Filtering

S.D. Brown, S.C. Rutan
1985 Journal of Research of the National Bureau of Standards  
Modifications to the filter involve allowing the filter to adapt the measurement model to the experimental data through matching the theoretical and observed covariance of the filter innovations sequence  ...  When valid models are available, many methods work well, but when models used in the estimation are in error, most methods fail.  ...  Approaches using factor analysis [7] have Bracketed figures indicate literature refrencrs. been developed for situations where the model is unknown, but these approaches are generally limited to very  ... 
doi:10.6028/jres.090.032 pmid:34566171 pmcid:PMC6644984 fatcat:pty5ponyrfaj5mrguqdq6wczja
« Previous Showing results 1 — 15 out of 191,406 results