Filters








216,965 Hits in 7.3 sec

On Second-Order Group Influence Functions for Black-Box Predictions [article]

Samyadeep Basu, Xuchen You, Soheil Feizi
2020 arXiv   pre-print
In this paper, we address this issue and propose second-order influence functions for identifying influential groups in test-time predictions.  ...  We also show that second-order influence functions could be used with optimization techniques to improve the selection of the most influential group for a test-sample.  ...  Acknowledgements This project was supported in part by NSF CAREER AWARD 1942230, HR001119S0026-GARD-FP-052, AWS Machine Learning Research Award, a sponsorship from Capital One, and Simons Fellowship on  ... 
arXiv:1911.00418v2 fatcat:bhbwvdqemnbtnb76ek2jtexpna

What will it take to generate fairness-preserving explanations? [article]

Jessica Dai, Sohini Upadhyay, Stephen H. Bach, Himabindu Lakkaraju
2021 arXiv   pre-print
However, the link between the fairness of the black-box model and the behavior of explanations for the black-box is unclear.  ...  We focus on explanations applied to tabular datasets, suggesting that explanations do not necessarily preserve the fairness properties of the black-box algorithm.  ...  Acknowledgements We would like to thank the anonymous reviewers for their insightful feedback. This work is supported in part by the NSF award #IIS-2008461, and Google.  ... 
arXiv:2106.13346v1 fatcat:r7oqytmvczdqnes3p4xgqh3rha

An interpretable semi-supervised classifier using two different strategies for amended self-labeling [article]

Isel Grau, Dipankar Sengupta, Maria M. Garcia Lorenzo, Ann Nowe
2020 arXiv   pre-print
Two different approaches for amending the self-labeling process are explored: a first one based on the confidence of the black box and the latter one based on measures from Rough Set Theory.  ...  In this paper, we report on an extended experimental study presenting an interpretable self-labeling grey-box classifier that uses a black box to estimate the missing class labels and a white box to explain  ...  The first strategy uses the confidence of the predictions made by the base black box and the second one focuses on the possible inconsistency of the enlarged dataset.  ... 
arXiv:2001.09502v2 fatcat:y2a6j54pxrhgbo5u432nksnsea

Payoff-based learning explains the decline in cooperation in public goods games

M. N. Burton-Chellew, H. H. Nax, S. A. West
2015 Proceedings of the Royal Society of London. Biological Sciences  
However, an alternative explanation for the empirical observations is that individuals are mistaken, but learn, during the game, how to improve their personal payoff.  ...  The results of such games have been used to argue that people are pro-social, and that humans are uniquely altruistic, willingly sacrificing their own welfare in order to benefit others.  ...  We thank Jay Biernaskie, Innes Cuthill, Claire El Mouden, Nichola Raihani and two anonymous referees for comments.  ... 
doi:10.1098/rspb.2014.2678 pmid:25589609 pmcid:PMC4309006 fatcat:u6trasnxwfgzrjcuf5pxkrutne

Look at the Variance! Efficient Black-box Explanations with Sobol-based Sensitivity Analysis [article]

Thomas Fel, Remi Cadene, Mathieu Chalvidal, Matthieu Cord, David Vigouroux, Thomas Serre
2021 arXiv   pre-print
Importantly, we show that the proposed method leads to favorable scores on standard benchmarks for vision (and language models) while drastically reducing the computing time compared to other black-box  ...  prediction through the lens of variance.  ...  The computing hardware was supported in part by NIH Office of the Director grant #S10OD025181 via the Center for Computation and Visualization.  ... 
arXiv:2111.04138v1 fatcat:p4jv36ankncipm7fmbnnypwtqm

Speech Understanding Through Syntactic and Semantic Analysis

Donald E. Walker
1973 International Joint Conference on Artificial Intelligence  
We already have begun work on a second ver-Stanford Research Institute is participating in a sion that will use a new parser, now under development, major program of research on the analysis of continuous  ...  The goal is the development of a new system is still in the process of construction, alspeech understanding system capable of engaging a human though the parser is far enough along for us to present operator  ...  ACKNOWLEDGEMENTS The following people have been involved in the research on speech understanding at SRI reflected in this paper:  ... 
dblp:conf/ijcai/Walker73 fatcat:p2eu3ehbtvhurksd5bft7wxw64

Fifty shades of black

Julien Leprince, Clayton Miller, Mario Frei, Henrik Madsen, Wim Zeiler
2021 Proceedings of the 8th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation  
The rapid growth of machine learning (black-box) techniques and computing capacity has started to transform many research domains, including building performance analysis.  ...  These insights present a starting point for further work towards highly scalable models yielding new characterizations of residential buildings.  ...  ACKNOWLEDGMENTS This work is funded by the Dutch Research Council (NWO), in the context of the call for Energy System Integration & Big Data (ESI-bida).  ... 
doi:10.1145/3486611.3491120 fatcat:mz4xd2jui5aw7fzxmylyl6fj5i

Auditing Black-box Models for Indirect Influence [article]

Philip Adler, Casey Falk, Sorelle A. Friedler, Gabriel Rybeck, Carlos Scheidegger, Brandon Smith, Suresh Venkatasubramanian
2016 arXiv   pre-print
Data-trained predictive models see widespread use, but for the most part they are used as black boxes which output a prediction or score.  ...  Our work focuses on the problem of indirect influence: how some features might indirectly influence outcomes via other, related features.  ...  Direct and Indirect Influence Much of the modern literature on black-box auditing (See Section II for details) focuses on what we call direct influence: how does a feature (or a group of features) directly  ... 
arXiv:1602.07043v2 fatcat:fnfcr6n46ra4tme3psuocpczqe

RelEx: A Model-Agnostic Relational Model Explainer [article]

Yue Zhang, David Defazio, Arti Ramesh
2020 arXiv   pre-print
In this work, we develop RelEx, a model-agnostic relational explainer to explain black-box relational models with only access to the outputs of the black-box.  ...  This is essential, as complex deep learning models with millions of parameters produce state of the art results, but it can be nearly impossible to explain their predictions.  ...  The first group finds important data points, which have high influence on learnt model behavior, including influence function (IF) [4] , and representer points [5] .  ... 
arXiv:2006.00305v1 fatcat:roqfurwfrng4tdq2akzovltawi

Auditing Black-Box Models for Indirect Influence

Philip Adler, Casey Falk, Sorelle A. Friedler, Gabriel Rybeck, Carlos Scheidegger, Brandon Smith, Suresh Venkatasubramanian
2016 2016 IEEE 16th International Conference on Data Mining (ICDM)  
Data-trained predictive models see widespread use, but for the most part they are used as black boxes which output a prediction or score.  ...  Our work focuses on the problem of indirect influence: how some features might indirectly influence outcomes via other, related features.  ...  Direct and Indirect Influence Much of the modern literature on black-box auditing (See Section II for details) focuses on what we call direct influence: how does a feature (or a group of features) directly  ... 
doi:10.1109/icdm.2016.0011 dblp:conf/icdm/AdlerFFRSSV16 fatcat:b3lvgpugkrgh5ohip6m6hpia3m

Auction optimization using regression trees and linear models as integer programs

Sicco Verwer, Yingqian Zhang, Qing Chuan Ye
2017 Artificial Intelligence  
We learn regression models from historical auctions, which are subsequently used to predict the expected value of orderings for new auctions.  ...  This new way of using learned models for optimization is promising. As the experimental results show, it significantly outperforms the black-box best-first search in nearly all settings.  ...  Second, we translate each of the item sets into constraints for both the black-box and white-box optimization solvers.  ... 
doi:10.1016/j.artint.2015.05.004 fatcat:kh67ekf245an3jva7gno7ovmoe

On The Importance Of Deep Learning Regularization Techniques In Knowledge Discovery

Ljubinka Sandjakoska, Atanas Hristov, Ana Madevska Bogdanova
2018 Zenodo  
The impact of regularization on knowledge discovery process is in the focus of this paper. In order to illustrate the effect of regularization in knowledge discovery, a case study is presented.  ...  Nowadays, in the era of complex data, the knowledge discovery process became one of the key challenges in the science.  ...  The knowledge in the black box of DNN is result of the learning process, allowed by activation function.  ... 
doi:10.5281/zenodo.1447167 fatcat:453e4d4rtbbf5anwyriirkn5vu

A Survey Of Methods For Explaining Black Box Models [article]

Riccardo Guidotti, Anna Monreale, Salvatore Ruggieri, Franco Turini, Dino Pedreschi, Fosca Giannotti
2018 arXiv   pre-print
The proposed classification of approaches to open black box models should also be useful for putting the many research open questions in perspective.  ...  Given a problem definition, a black box type, and a desired explanation this survey should help the researcher to find the proposals more useful for his own work.  ...  The first one is mostly focused on describing how black boxes work, while the second one is more interested in explaining the decisions even without understanding the details on how the opaque decision  ... 
arXiv:1802.01933v3 fatcat:n6ly5sqyjfhwjbcdx3h32cmiqi

Emotionality ratings and open-field behavior

Charles D. Corman, John Biondo
1969 Psychonomic Science  
On the fourth rating day, S was placed in the start box within the fjeld. Thirty seconds later the box was gently lifted from the open-field, and S was observed for 3 min.  ...  Within the limits of the procedures employed, no prediction of Ss' open-field behavior on the basis of emotionality rating could be made.  ...  On the fourth rating day, S was placed in the start box within the fjeld. Thirty seconds later the box was gently lifted from the open-field, and S was observed for 3 min.  ... 
doi:10.3758/bf03332740 fatcat:yiwbs34s5bgtroirgeypv4opau

Estimating extremely low probability of stochastic defect in extreme ultraviolet lithography from critical dimension distribution measurement

Hiroshi Fukuda, Yoshinori Momonoi, Kei Sakai
2019 Journal of Micro/Nanolithography  
We discuss a method for predicting stochastic defect probabilities from a histogram of feature sizes for patterns several orders of magnitude fewer than the number of features to inspect.  ...  The defect probabilities in the order between 10 −7 and 10 −5 were predicted from 10 5 measurement data for real EUV-exposed wafers, suggesting the effectiveness of the model and its potential for defect  ...  De Bisschop and IMEC for the sample preparation and for their support of this work.  ... 
doi:10.1117/1.jmm.18.2.024002 fatcat:m2sqqu4z2fgjvkpodqxeoae2k4
« Previous Showing results 1 — 15 out of 216,965 results