Filters








36,785 Hits in 7.4 sec

Is the Best Better? Bayesian Statistical Model Comparison for Natural Language Processing [article]

Piotr Szymański, Kyle Gorman
2020 arXiv   pre-print
Recent work raises concerns about the use of standard splits to compare natural language processing models.  ...  We propose a Bayesian statistical model comparison technique which uses k-fold cross-validation across multiple data sets to estimate the likelihood that one model will outperform the other, or that the  ...  Acknowledgments We would like to thank Steve Bedrick for previous work on this topic.  ... 
arXiv:2010.03088v1 fatcat:lzjhrfxse5dgdcojgcqzux4ezy

Is the Best Better? Bayesian Statistical Model Comparison for Natural Language Processing

Piotr Szymański, Kyle Gorman
2020 Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)   unpublished
Recent work raises concerns about the use of standard splits to compare natural language processing models.  ...  We propose a Bayesian statistical model comparison technique which uses k-fold cross-validation across multiple data sets to estimate the likelihood that one model will outperform the other, or that the  ...  Acknowledgments We would like to thank Steve Bedrick for previous work on this topic.  ... 
doi:10.18653/v1/2020.emnlp-main.172 fatcat:qrqhn6omxbcbhk4j4xidd3c4rq

When is Deep Learning the Best Approach to Knowledge Tracing?

Theophile Gervet, Ken Koedinger, Jeff Schneider, Tom Mitchell
2020 Zenodo  
The advent of increasingly large scale datasets has turned deep learning models for learner performance prediction into competitive alternatives to classical Markov process and logistic regression models  ...  Markov process methods, like Bayesian Knowledge Tracing, lag behind other approaches.  ...  Dan Bindman for enlightening discussions about how to evaluate and visualize the performance of the alternative models considered here.  ... 
doi:10.5281/zenodo.4143614 fatcat:tsxvcwwa6vfgpa2waomm3dg3cu

Learning discrete Bayesian network parameters from continuous data streams: What is the best strategy?

Parot Ratnapinda, Marek J. Druzdzel
2015 Journal of Applied Logic  
While the differences in speed between incremental algorithms are not large (online EM is slightly slower), for all but small data sets online EM tends to be more accurate than incremental EM.  ...  We compare three approaches to learning numerical parameters of discrete Bayesian networks from continuous data streams: (1) the EM algorithm applied to all data, (2) the EM algorithm applied to data increments  ...  Interestingly, in an experiment reported by Liang and Klein [8] , consisting of four different unsupervised learning tasks in the domain of natural language processing, the online EM algorithm performed  ... 
doi:10.1016/j.jal.2015.03.007 fatcat:vj2cqlbgqjdmbgodagovkife4m

Two Better Books, But No Extraordinary Ones

RAPHAEL HANSON
1977 Contemporary Psychology  
Ferguson alone presents a measurement model; none presents a process model.  ...  The main feature of bayesian logic that permits it to deal with a limitless stream of information is that its core is a logic for judgment making or informa- tion processing rather than a logic for decision  ... 
doi:10.1037/016075 fatcat:wsz322cjfjgs7ai4q3keiazbey

Is This Really Relevant? A Guide to Best Practice Gaze-based Relevance Prediction Research

Melanie Heck, Paulina Sonntag, Christian Becker
2021 Adjunct Proceedings of the 29th ACM Conference on User Modeling, Adaptation and Personalization  
The insights may serve as a guide to establish best practices for the design and evaluation of relevance prediction models, thus allowing for better comparability of future work.  ...  As eye tracking is becoming feasible on commodity devices, it provides a powerful tool for inferring users' perceived relevance of objects.  ...  The insight gained from our literature review shall help to establish best practices for the design and evaluation of relevance prediction models, thus allowing for better comparability of future work.  ... 
doi:10.1145/3450614.3464476 fatcat:qztximjbo5fuzlk6mikztl2lsy

MaskGAN: Better Text Generation via Filling in the______ [article]

William Fedus, Ian Goodfellow, Andrew M. Dai
2018 arXiv   pre-print
GANs were originally designed to output differentiable values, so discrete language generation is challenging for them.  ...  Neural text generation models are often autoregressive language models or seq2seq models.  ...  the first year residents who humored us listening and commenting on almost every conceivable variation of this core idea.  ... 
arXiv:1801.07736v3 fatcat:lloc3o5isbg65mqjc6glunn7oa

Bimanual proprioception: are two hands better than one?

Jeremy D. Wong, Elizabeth T. Wilson, Dinant A. Kistemaker, Paul L. Gribble
2014 Journal of Neurophysiology  
Our results are consistent with the hypothesis that the nervous system both has knowledge of, and uses the limb with the best proprioceptive acuity for bimanual proprioception.  ...  Surprisingly, a Bayesian model that postulates optimal combination of sensory signals could not predict empirically observed bimanual acuity.  ...  Gribble), the Netherlands Organization for Scientific Research (to D. A. Kistemaker), a National Sciences and Engineering Council of Canada graduate scholarship (to E. T.  ... 
doi:10.1152/jn.00537.2013 pmid:24381030 pmcid:PMC4250236 fatcat:lwvavfkg25efxph2kdg2pvsi5m

Towards Better Confidence Estimation for Neural Models

Vishal Thanvantri Vasudevan, Abhinav Sethy, Alireza Roshan Ghias
2019 ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
In this work we focus on confidence modeling for neural network based text classification and sequence to sequence models in the context of Natural Language Understanding (NLU) tasks.  ...  For most applications, the confidence of a neural network model in it's output is computed as a function of the posterior probability, determined via a softmax layer.  ...  For the probability alignment metric, our confidence model performs better than the baseline in two of the four cases, however the difference is minimal in the other two cases.  ... 
doi:10.1109/icassp.2019.8683359 dblp:conf/icassp/VasudevanSG19 fatcat:73is6m6txfarfobzj634h4gvti

Experimenters: Here's to Better Design, Data Analysis, and Model Building

B. J. WINER
1979 Contemporary Psychology  
Statistical techniques are an im- portant adjunct to, not a replacement for, the natural skill of the experimenter.  ...  comparisons to near speakers such as linguistic apes and other cases, Curtiss believes that Genie has acquired lan- guage, but that it is right-hemisphere language (and cognition), not normal language.  ... 
doi:10.1037/018903 fatcat:36t6eohfgfbmrg6ydqdzsep77q

Towards a Better Understanding of Memory-Based Reasoning Systems [chapter]

John Rachlin, Simon Kasif, Steven Salzberg, David W. Aha
1994 Machine Learning Proceedings 1994  
competing for best performance (e.g., speech recognition and natural language parsing).  ...  ., by the case-based reasoning community) that instead of producing a description of the problem domain in terms of logical rules, functional descriptions, or a complex statistical model, it is possible  ...  Jantke for many helpful discussions, and the anonymous reviewers for their comments.  ... 
doi:10.1016/b978-1-55860-335-6.50037-4 dblp:conf/icml/RachlinKSA94 fatcat:73ksx6zefbblje6hhqwtkkc3qu

Regularizing translation models for better automatic image annotation

Feng Kang, Rong Jin, Joyce Y. Chai
2004 Proceedings of the Thirteenth ACM conference on Information and knowledge management - CIKM '04  
It views the process of annotating images as a process of translating the content from a 'visual language' to textual words.  ...  In the past, statistical machine translation models have been successfully applied to automatic image annotation task [8] .  ...  A machine translation model for automatic image annotation [8] views the process of annotating images as a process of translating information from a 'visual language' to textual words.  ... 
doi:10.1145/1031171.1031242 dblp:conf/cikm/KangJC04 fatcat:xoeurop3mbhdxkewsrptun5on4

Homo Heuristicus: Why Biased Minds Make Better Inferences

Gerd Gigerenzer, Henry Brighton
2009 Topics in Cognitive Science  
We review the major progress made so far: (a) the discovery of less-is-more effects; (b) the study of the ecological rationality of heuristics, which examines in which environments a given strategy succeeds  ...  or fails, and why; (c) an advancement from vague labels to computational models of heuristics; (d) the development of a systematic theory of heuristics that identifies their building blocks and the evolved  ...  Acknowledgments We are grateful to Julian Marewski, Shabnam Mousavi, Lael Schooler, and three anonymous reviewers for their helpful comments.  ... 
doi:10.1111/j.1756-8765.2008.01006.x pmid:25164802 fatcat:6ejkxm3djvb27fnaw7nmns2c2i

Homo heuristicus: Why Biased Minds Make Better Inferences [chapter]

Gerd Gigerenzer, Henry Brighton
2011 Heuristics  
We review the major progress made so far: (a) the discovery of less-is-more effects; (b) the study of the ecological rationality of heuristics, which examines in which environments a given strategy succeeds  ...  or fails, and why; (c) an advancement from vague labels to computational models of heuristics; (d) the development of a systematic theory of heuristics that identifies their building blocks and the evolved  ...  Acknowledgments We are grateful to Julian Marewski, Shabnam Mousavi, Lael Schooler, and three anonymous reviewers for their helpful comments.  ... 
doi:10.1093/acprof:oso/9780199744282.003.0001 fatcat:3ukk367rlnajnex6xkymvhx2je

Non-significant p-values? Strategies to understand and better determine the importance of effects and interactions in logistic regression

Zarina I Vakhitova, Clair L Alston-Knox
2018 PLoS ONE  
analysis of GLM is a suitable alternative, which is enhanced with prior knowledge about the direction of the effects; and Bayesian Model Averaging (BMA) is especially suited for new areas of research,  ...  In the context of generalized linear models (GLMs), interactions are automatically induced on the natural scale of the data.  ...  The relatively small sample size and the chosen parameterization will affect the model selection based on the changed deviance or other test statistics that are searching for the single best model, as  ... 
doi:10.1371/journal.pone.0205076 pmid:30475804 pmcid:PMC6261058 fatcat:p4fkzgxstzgdjpalnqsratq5ea
« Previous Showing results 1 — 15 out of 36,785 results