958 Hits in 7.6 sec

Unbiased Learning to Rank via Propensity Ratio Scoring [article]

Nan Wang, Xuanhui Wang, Hongning Wang
2020 arXiv   pre-print
We then derive a new weighting scheme called Propensity Ratio Scoring (PRS) that takes a holistic treatment on both clicks and non-clicks.  ...  Though effective, it only corrects the bias introduced by treating clicked documents as relevant, but cannot handle the bias caused by treating unclicked ones as irrelevant.  ...  Simply treating click/non-click as relevance/irrelevance proves to distort LTR training [20, 21, 34] .  ... 
arXiv:2005.08480v1 fatcat:igk7ki6cpngm7c34zn5hqesprm

The Irrelevant Sound Phenomenon Revisited: What Role for Working Memory Capacity?

C. Philip Beaman
2004 Journal of Experimental Psychology. Learning, Memory and Cognition  
High-span individuals (as measured by the operation span [OSPAN] technique) are less likely than lowspan individuals to notice their own names in an unattended auditory stream (A. R. A. Conway, N.  ...  Low-OSPAN participants did, however, make more semantically related intrusion errors from the irrelevant sound stream in a free recall test (Experiment 4).  ...  A further experiment separating the high-and low-span groups by means of comparing only the top-and bottom-scoring quartiles will act as a check on this possibility.  ... 
doi:10.1037/0278-7393.30.5.1106 pmid:15355139 fatcat:huractrkzvgcxbypgzc2nbt2se

Counterfactual Learning to Rank using Heterogeneous Treatment Effect Estimation [article]

Mucun Tian, Chun Guo, Vito Ostuni, Zhen Zhu
2020 arXiv   pre-print
inverse propensity score (IPS) to debias LTR algorithms on the whole data set.  ...  To unbiasedly learn to rank, existing counterfactual frameworks first estimate the propensity (probability) of missing clicks with intervention data from a small portion of search traffic, and then use  ...  For irrelevant documents, we expect corrected CTR to be as low as 0 and not over-estimate the false CTR.  ... 
arXiv:2007.09798v1 fatcat:ulm6it6x4ncsxj4q7ptebjj4dy

Evaluating online ad campaigns in a pipeline

David Chan, Rong Ge, Ori Gershony, Tim Hesterberg, Diane Lambert
2010 Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining - KDD '10  
Simulations based on realistic scenarios show that the resulting estimates are more robust to selection bias than traditional alternatives, such as regression modeling or propensity scoring.  ...  Doubly robust estimation protects against the selection bias that is inherent in observational data, and a nonparametric test that is based on irrelevant outcomes provides further defense.  ...  Inverse Propensity Weighting Propensity score weighting is an alternative to propensity score matching. It can be motivated by a simple analogy.  ... 
doi:10.1145/1835804.1835809 dblp:conf/kdd/ChanGGHL10 fatcat:dmuycevz2vb2ded5kzpwnuxfdy

Unbiased Learning-to-Rank with Biased Feedback

Thorsten Joachims, Adith Swaminathan, Tobias Schnabel
2017 Proceedings of the Tenth ACM International Conference on Web Search and Data Mining - WSDM '17  
For example, position bias in search rankings strongly influences how many clicks a result receives, so that directly using click data as a training signal in Learning-to-Rank (LTR) methods yields sub-optimal  ...  Using this framework, we derive a Propensity-Weighted Ranking SVM for discriminative learning from implicit feedback, where click models take the role of the propensity estimator.  ...  ACKNOWLEDGMENTS This work was supported in part through NSF Awards IIS-1247637, IIS-1513692, IIS-1615706, and a gift from Bloomberg.  ... 
doi:10.1145/3018661.3018699 dblp:conf/wsdm/JoachimsSS17 fatcat:bvcuscpk2vbyro3rhboyhzo3di

Position Bias Estimation for Unbiased Learning to Rank in Personal Search

Xuanhui Wang, Nadav Golbandi, Michael Bendersky, Donald Metzler, Marc Najork
2018 Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining - WSDM '18  
We propose a regressionbased Expectation-Maximization (EM) algorithm that is based on a position bias click model and that can handle highly sparse clicks in personal search.  ...  A well-known challenge in learning from click data is its inherent bias and most notably position bias.  ...  Similarly, we train a third lambdaMART ranking function to optimize the IPW Prec using EM estimated propensity scores (denoted as EMCorrected).  ... 
doi:10.1145/3159652.3159732 dblp:conf/wsdm/WangGBMN18 fatcat:shszbvbxpbeydit6ik5f5isn7q

Doubly-Robust Estimation for Correcting Position-Bias in Click Feedback for Unbiased Learning to Rank [article]

Harrie Oosterhuis
2022 arXiv   pre-print
The prevalent approach to unbiased click-based learning-to-rank (LTR) is based on counterfactual inverse-propensity-scoring (IPS) estimation.  ...  As a solution, our estimator uses the expected treatment per rank, instead of the actual treatment that existing DR estimators use.  ...  [48] proposed a ratio-propensity-scoring (RPS) estimator that weights pairs of clicked and non-clicked items by their ratio between the propensities.  ... 
arXiv:2203.17118v2 fatcat:6z74x6ukynestnqneb6hgge6xm

Creativity and sensory gating indexed by the P50: Selective versus leaky sensory gating in divergent thinkers and creative achievers

Darya L. Zabelina, Daniel O'Leary, Narun Pornpattananangkul, Robin Nusslock, Mark Beeman
2015 Neuropsychologia  
in a regression.  ...  Finally, the ERP effect was specific to the P50neither divergent thinking nor creative achievement were related to later components, such as the N100 and P200.  ...  Gavin and Ken A. Paller for their valuable feedback regarding the manuscript.  ... 
doi:10.1016/j.neuropsychologia.2015.01.034 pmid:25623426 fatcat:wxdmj3qtobbh7lzpvwiskf3qza

Countering Position Bias in Instructor Interventions in MOOC Discussion Forums

Muthu Kumar Chandrasekaran, Min-Yen Kan
2018 Proceedings of the 5th Workshop on Natural Language Processing Techniques for Educational Applications  
In a large scale dataset, we conclusively show that instructor interventions exhibit strong position bias, as measured by the position where the thread appeared on the user interface at the time of intervention  ...  To correct for this imbalance, they used class weights on examples, estimated as the ratio of intervened to non-intervened threads.  ...  Ratio), defined as the ratio of # of intervened to non-intervened threads. by removing posts after the first instructor post.  ... 
doi:10.18653/v1/w18-3720 dblp:conf/acl-tea/ChandrasekaranK18 fatcat:mjwnot6bjvcf7cdnve75nll3iu

Effects of Response Format on Psychometric Properties and Fairness of a Matrices Test: Multiple Choice vs. Free Response

Sonja Breuer, Thomas Scherndl, Tuulia M. Ortner
2020 Frontiers in Education  
The purpose of the current study was to examine effects of response format on psychometric properties and fairness of a matrices test according to examinee's sex, risk propensity, and test anxiety.  ...  As examinee characteristics may affect performance in Reasoning tests, concern about maintaining fairness is expressed.  ...  The corrected average number of pumps (0-128), which is the average of clicks per unexploded balloon, served as a score for risk propensity.  ... 
doi:10.3389/feduc.2020.00015 fatcat:jbmjh5hi6jas7m7sbgncsei5we

An efficient framework for online advertising effectiveness measurement and comparison

Pengyuan Wang, Yechao Liu, Marsha Meytlis, Han-Yun Tsao, Jian Yang, Pei Huang
2014 Proceedings of the 7th ACM international conference on Web search and data mining - WSDM '14  
One solid methodology for a fair comparison is to apply inverse propensity weighting with doubly robust estimation to the observational data.  ...  The choice of features, models and feature selection scheme are validated with irrelevant conversion test.  ...  The correction terms are products of the estimators from success model and the propensity score model.  ... 
doi:10.1145/2556195.2556235 dblp:conf/wsdm/WangLMTYH14 fatcat:ek7wuyzjgnftdiboepphwg3qwm

Intervention Harvesting for Context-Dependent Examination-Bias Estimation [article]

Zhichong Fang, Aman Agarwal, Thorsten Joachims
2018 arXiv   pre-print
Accurate estimates of examination bias are crucial for unbiased learning-to-rank from implicit feedback in search engines and recommender systems, since they enable the use of Inverse Propensity Score  ...  To overcome this limitation, we propose a Contextual Position-Based Model (CPBM) where the examination bias may also depend on a context vector describing the query and the user.  ...  Acknowledgments This research was supported in part by NSF Awards IIS-1513692 and IIS-1615706, a gift from Amazon, and a gift from Workday.  ... 
arXiv:1811.01802v2 fatcat:v32ctc4bdfbcxdbuilupkrfw7y

Chronic effects of cannabis on sensory gating

Samantha J. Broyd, Lisa-marie Greenwood, Rodney J. Croft, Anna Dalecki, Juanita Todd, Patricia T. Michie, Stuart J. Johnstone, Nadia Solowij
2013 International Journal of Psychophysiology  
Twenty controls and 21 regular cannabis users completed a P50 paired-click (S1 and S2) paradigm with an inter-pair interval of 9. s.  ...  Twenty controls and 21 regular cannabis users completed a P50 paired-click (S1 and S2) paradigm with an inter-pair interval of 9 seconds.  ...  Sensory gating in cannabis users and non-user controls Grand mean ERP waveforms to the first (S1) and second (S2) click at Cz are presented for cannabis users and controls in Figure 1 and mean (SD) P50  ... 
doi:10.1016/j.ijpsycho.2013.04.015 pmid:23628289 fatcat:gsihhbrj5bc37jqhzhszqpuey4

Large-scale Validation of Counterfactual Learning Methods: A Test-Bed [article]

Damien Lefortier, Adith Swaminathan, Xiaotao Gu, Thorsten Joachims, Maarten de Rijke
2017 arXiv   pre-print
The ability to perform effective off-policy learning would revolutionize the process of building better interactive systems, such as search engines and recommendation systems for e-commerce, computational  ...  In particular, we consider the problem of filling a banner ad with an aggregate of multiple products the user may want to purchase.  ...  Importantly, non-clicked examples were sub-sampled aggressively to reduce the dataset size and we kept only a random 10% sub-sample of them.  ... 
arXiv:1612.00367v2 fatcat:nsulmff7yvdhvgpikutypbmrgu

Pathological risk-propensity typifies Mafia members' cognitive profile

Gerardo Salvato, Maria Laura Fiorina, Gabriele De Maio, Elisa Francescon, Daniela Ovadia, Luisa Bernardinelli, Amedeo Santosuosso, Eraldo Paulesu, Gabriella Bottini
2020 Scientific Reports  
We found that OC members were more likely to show pathological risk-propensity than non-OC prisoners. We interpret this finding as the result of the internal dynamics of Mafia groups.  ...  Here we investigated the frontal lobe cognitive functions of 50 OC prisoners from the Mafia and 50 non-OC prisoners based on the performance of 50 non-prisoner controls.  ...  We considered this variable as a covariate in the subsequent analyses. As a second step, we standardized the OC and non-OC participants' test scores based on the non-prisoner control group's values.  ... 
doi:10.1038/s41598-020-65486-z pmid:32444792 fatcat:ss43gvrqvjdezle26rsbtdolz4
« Previous Showing results 1 — 15 out of 958 results