Filters








88 Hits in 3.6 sec

Non-Vacuous Generalization Bounds at the ImageNet Scale: A PAC-Bayesian Compression Approach [article]

Wenda Zhou and Victor Veitch and Morgane Austern and Ryan P. Adams and Peter Orbanz
2019 arXiv   pre-print
The bounds are complemented by empirical results that show an increase in overfitting implies an increase in the number of bits required to describe a trained network.  ...  As additional evidence connecting compression and generalization, we show that compressibility of models that tend to overfit is limited: We establish an absolute limit on expected compressibility as a  ...  A single cluster in each weight is given to be exactly zero and contains the pruned weights. The remaining clusters centers are learned using the ADAM optimizer over 1000 steps.  ... 
arXiv:1804.05862v3 fatcat:tgisbjtj7rbixf4yrgpjmbl3pu

Bayesian confusions surrounding simplicity and likelihood in perceptual organization

Peter A. van der Helm
2011 Acta Psychologica  
(which are used in Bayesian models of the Occamian simplicity principle) and objective probabilities (which are needed in Bayesian models of the Helmholtzian likelihood principle).  ...  Furthermore, Occamian counterparts of Bayesian priors and conditionals have led to another confusion, which seems to have been triggered by a dual role of regularity in perception.  ...  Compared to that, one coincidence can intuitively be said to occur in the category in Fig. 3b , two in the one in Fig. 3c , and three in the one in Fig. 3d .  ... 
doi:10.1016/j.actpsy.2011.09.007 pmid:21982531 fatcat:jrbn57sdvrhxvpqapkatqokmcy

A Philosophical Treatise of Universal Induction

Samuel Rathmanner, Marcus Hutter
2011 Entropy  
largely unknown and unappreciated in the wider scientific community.  ...  In the process we examine the major historical contributions that have led to the formulation of Solomonoff Induction as well as criticisms of Solomonoff and induction in general.  ...  The universal prior involves the use of Kolmogorov complexity which we argue is highly intuitive and does justice to both Occam and Epicurus.  ... 
doi:10.3390/e13061076 fatcat:2cpgyqohv5eqzifybuppy3wphi

Should we still believe in constrained supersymmetry?

Csaba Balázs, Andy Buckley, Daniel Carter, Benjamin Farmer, Martin White
2013 European Physical Journal C: Particles and Fields  
This is done in the Bayesian spirit where probability reflects a degree of belief in a proposition and Bayes' theorem tells us how to update it after acquiring new information.  ...  This reduction is largely due to substantial Occam factors induced by the LEP and LHC Higgs searches.  ...  Acknowledgements This research was funded in part by the ARC Centre of Excellence for Particle Physics at the Tera-scale, and in part by the Project of Knowledge Innovation Program (PKIP) of Chinese Academy  ... 
doi:10.1140/epjc/s10052-013-2563-y fatcat:xfenozdoffcyvpop6uolr3i7ci

A Dialogue with the Data: The Bayesian Foundations of Iterative Research in Qualitative Social Science

Tasha Fairfield, Andrew Charman
2019 Perspectives on Politics  
Bayesian probability not only fits naturally with how we intuitively move back and forth between theory and data, but also provides a framework for rational reasoning that mitigates confirmation bias and  ...  We advance efforts to explicate and improve inference in qualitative research that iterates between theory development, data collection, and data analysis, rather than proceeding linearly from hypothesizing  ...  Collier, David Collier, Justin Grimmer, Macartan Humphreys, Alan Jacobs, Jack Levy, James Mahoney, Jason Sharman, Hillel Soifer, and Elisabeth Wood for detailed comments and intellectual engagement.  ... 
doi:10.1017/s1537592718002177 fatcat:7tagbud2r5bhjfhvvoplazc5ue

Perceptual estimation obeys Occam's razor

Samuel J. Gershman, Yael Niv
2013 Frontiers in Psychology  
In a series of experiments, we tested this prediction by asking participants to estimate the number of colored circles on a computer screen, with the number of circles drawn from a color-specific distribution  ...  Theoretical models of unsupervised category learning postulate that humans "invent" categories to accommodate new patterns, but tend to group stimuli into a small number of categories.  ...  FUNDING This research was supported in part by the National Institute Of Mental Health of the National Institutes of Health under Award Number R01MH098861.  ... 
doi:10.3389/fpsyg.2013.00623 pmid:24137136 pmcid:PMC3780620 fatcat:n4obferqr5bbzduiiacbbvce6e

A genetic algorithm analysis of N* resonances inBp($gamma;,K+)$Lambda; reactions

D IRELAND
2004 Nuclear Physics A  
Our genetic algorithm method predicts that photon beam asymmetry and double polarization p(γ , K + )Λ measurements should provide the most sensitive information with respect to missing resonances.  ...  It is shown that, within the limitations of this tree-level analysis, a resonance in addition to the known set is required to obtain a reasonable fit.  ...  Acknowledgements This work was supported by the UK's Engineering and Physical Sciences Research Council, and the Fund for Scientific Research-Flanders.  ... 
doi:10.1016/s0375-9474(04)00705-5 fatcat:3s7rmm65u5benf7ela2y3ogy4a

A genetic algorithm analysis of resonances in p(γ,K+)Λ reactions

D.G Ireland, S Janssen, J Ryckebusch
2004 Nuclear Physics A  
Our genetic algorithm method predicts that photon beam asymmetry and double polarization p(γ,K^+)Λ measurements should provide the most sensitive information with respect to missing resonances.  ...  It is shown that, within the limitations of this tree-level analysis, a resonance in addition to the known set is required to obtain a reasonable fit.  ...  Acknowledgements This work was supported by the UK's Engineering and Physical Sciences Research Council, and the Fund for Scientific Research-Flanders.  ... 
doi:10.1016/j.nuclphysa.2004.05.007 fatcat:j6a427dzqbghle3kkxel2aqoq4

Bayesian analysis of multiple direct detection experiments [article]

Chiara Arina
2014 arXiv   pre-print
Lastly the Bayes' factor gives inconclusive evidence for an incompatibility between the data sets of XENON100 and the hints of detection.  ...  In particular we discuss the exclusion limit of XENON100 and the debated hints of detection under the hypothesis of a WIMP signal.  ...  The surface electron background is exponentially falling off in the energy window for WIMP detection, while the neutron and lead background are constant in the same energy range.  ... 
arXiv:1310.5718v2 fatcat:qibiw55nqbhvth2s4f6da7kafe

A Bayesian Nonparametric Approach to Testing for Dependence Between Random Variables

Sarah Filippi, Chris C. Holmes
2017 Bayesian Analysis  
Nonparametric and nonlinear measures of statistical dependence between pairs of random variables are important tools in modern data analysis.  ...  Pólya tree priors can accommodate known uncertainty in the form of the underlying sampling distribution and provides an explicit posterior probability measure of both dependence and independence.  ...  Acknowledgments We thank Moustafa Abdalla for sharing his insights into gene expression analysis and pathway enrichment, and in his support for the analysis in Section 4.2.  ... 
doi:10.1214/16-ba1027 fatcat:oyu2vfcn4fasdih5pnogh3tdru

A Machine Learning Tutorial for Operational Meteorology, Part I: Traditional Machine Learning [article]

Randy J. Chase, David R. Harrison, Amanda Burke, Gary M. Lackmann, Amy McGovern
2022 arXiv   pre-print
The following machine learning methods are demonstrated: linear regression; logistic regression; decision trees; random forest; gradient boosted decision trees; naive Bayes; and support vector machines  ...  Furthermore, all code (in the form of Jupyter notebooks and Google Colaboratory notebooks) used to make the examples in the paper is provided in an effort to catalyse the use of machine learning in meteorology  ...  ICER-2019758, supporting authors RJC, AM and AB.  ... 
arXiv:2204.07492v2 fatcat:yn4w5xuq7ndfpd4xo2lwxgyz4i

Generalized Categorization Axioms [article]

Jian Yu
2016 arXiv   pre-print
In order to generalize categorization axioms into general cases, categorization input and categorization output are reinterpreted by inner and outer category representation.  ...  Categorization axioms have been proposed to axiomatizing clustering results, which offers a hint of bridging the difference between human recognition system and machine learning through an intuitive observation  ...  In order to be consistent with the intuition, category similarity mapping and category dissimilarity mapping are usually supposed to be non negative in this section.  ... 
arXiv:1503.09082v11 fatcat:y3dozglpxvdavjijnmcowwa52u

A bayesian hierarchical mixture of experts approach to estimate speech quality

S. Iman Mossavat, Oliver Amft, Bert de Vries, Petko N. Petkov, W. Bastiaan Kleijn
2010 2010 Second International Workshop on Quality of Multimedia Experience (QoMEX)  
Despite using weaker modeling assumptions, we are still able to achieve comparable accuracy on predicting mean-opinion-scores with P.563.  ...  This paper demonstrates the potential of theoretically motivated learning methods in solving the problem of nonintrusive quality estimation for which the state-of-the-art is represented by ITU-T P.563  ...  The design of P.563 is intuitive and interpretable.  ... 
doi:10.1109/qomex.2010.5516203 fatcat:isf53vzbezenpc7yyi6gsrt5re

Bayesian analysis of multiple direct detection experiments

Chiara Arina
2014 Physics of the Dark Universe  
Lastly the Bayes' factor gives inconclusive evidence for an incompatibility between the data sets of XENON100 and the hints of detection.  ...  In particular we discuss the exclusion limit of XENON100 and the debated hints of detection under the hypothesis of a WIMP signal.  ...  This is a consequence of the predictiveness of models 1a and 1b: Occams' razor is at work, penalizing the other physics models for their excessive free parameters unsupported by the data.  ... 
doi:10.1016/j.dark.2014.03.003 fatcat:ty7lcnqabrgvhccou5wa3dywyu

Applying Occam's razor in modeling cognition: A Bayesian approach

In Jae Myung, Mark A. Pitt
1997 Psychonomic Bulletin & Review  
parameter space, and, most importantly, the functional form of the model (i.e., the way in which the parameters are combined in the model's equation).  ...  Application examples are presented and implications of the results for evaluating models of cognition are discussed.  ...  predicted data.  ... 
doi:10.3758/bf03210778 fatcat:pe52kyejsjathigjxxxf2wvqum
« Previous Showing results 1 — 15 out of 88 results