Filters








234 Hits in 4.8 sec

A Submodular Feature-Aware Framework for Label Subset Selection in Extreme Classification Problems

Elham J. Barezi, Ian D. Wood, Pascale Fung, Hamid R. Rabiee
2019 Proceedings of the 2019 Conference of the North  
Extreme classification is a classification task on an extremely large number of labels (tags).  ...  We propose a submodular maximization framework with linear cost to find informative labels which are most relevant to other labels yet least redundant with each other.  ...  Conclusion and Future Work We propose a novel approach for extreme multilabel classification that simplifies the problem by selecting an informative and easily modelled subset of labels and subsequently  ... 
doi:10.18653/v1/n19-1106 dblp:conf/naacl/BareziWFR19 fatcat:blohzjhla5cjtcrdvoz2ws3ati

Learning From Less Data: A Unified Data Subset Selection and Active Learning Framework for Computer Vision [article]

Vishal Kaushal, Rishabh Iyer, Suraj Kothawade, Rohan Mahadev, Khoshrav Doctor, Ganesh Ramakrishnan
2019 arXiv   pre-print
They can also help improve the efficiency of active learning in further reducing human labeling efforts by selecting a subset of the examples obtained using the conventional uncertainty sampling based  ...  In this work, we empirically demonstrate the effectiveness of two diversity models, namely the Facility-Location and Dispersion models for training-data subset selection and reducing labeling effort.  ...  The authors would like to thank Suyash Shetty, Anurag Sahoo, Narsimha Raju and Pankaj Singh for discussions and useful suggestions on the manuscripts.  ... 
arXiv:1901.01151v1 fatcat:qynsco4vcrhx3atvengh2qz2sa

SPOT: A framework for selection of prototypes using optimal transport [article]

Karthik S. Gurumoorthy and Pratik Jawanpuria and Bamdev Mishra
2021 arXiv   pre-print
In this work, we develop an optimal transport (OT) based framework to select informative prototypical examples that best represent a given target dataset.  ...  We model the prototype selection problem as learning a sparse (empirical) probability distribution having the minimum OT distance from the target distribution.  ...  To this end, we propose a novel framework for Selection of Prototypes using the Optimal Transport theory or the SPOT framework for searching a subset P from a source dataset X (i.e., P ⊂ X) that best represents  ... 
arXiv:2103.10159v2 fatcat:cjnshghnvffk5l3lfm53koiznu

Coverage optimized active learning for k - NN classifiers

Ajay J. Joshi, Fatih Porikli, Nikolaos Papanikolopoulos
2012 2012 IEEE International Conference on Robotics and Automation  
Using submodular function optimization, the proposed algorithm presents a nearoptimal selection strategy for an otherwise intractable problem.  ...  are labeled neighbors in a small neighborhood of each point.  ...  SELECTION FRAMEWORK In a 1 − N N classifier, for each data point, the label is obtained by finding the label of the nearest labeled point.  ... 
doi:10.1109/icra.2012.6225054 dblp:conf/icra/JoshiPP12 fatcat:o76xyljylngqxnd5lkidedzxom

Multi-class batch-mode active learning for image classification

Ajay J Joshi, Fatih Porikli, Nikolaos Papanikolopoulos
2010 2010 IEEE International Conference on Robotics and Automation  
In this paper we present a batch-mode active learning framework for multi-class image classification systems.  ...  Our framework addresses two important issues: i) it handles redundancy between different images which is crucial when batch-mode selection is performed; and ii) we pose batchselection as a submodular function  ...  In image classification problems, large amounts of human labeled training data are required for satisfactory performance.  ... 
doi:10.1109/robot.2010.5509293 dblp:conf/icra/JoshiPP10 fatcat:ebesfw4c2fbynb2yrtjr6pjnou

Submodularity In Machine Learning and Artificial Intelligence [article]

Jeff Bilmes
2022 arXiv   pre-print
and condensation, and data subset selection and feature selection.  ...  In this manuscript, we offer a gentle review of submodularity and supermodularity and their properties.  ...  This is the standard feature selection problem in machine learning for which there are many possible solutions. Rather than select features, we might wish to select a subset of the training data.  ... 
arXiv:2202.00132v1 fatcat:sp4b3ww3ajdxvfgigp7xw4f4yq

GraphReach: Position-Aware Graph Neural Network using Reachability Estimations [article]

Sunil Nishad and Shubhangi Agarwal and Arnab Bhattacharya and Sayan Ranu
2021 arXiv   pre-print
We show that this combinatorial anchor selection problem is NP-hard and, consequently, develop a greedy (1-1/e) approximation heuristic.  ...  In this paper, we develop GraphReach, a position-aware inductive GNN that captures the global positions of nodes through reachability estimations with respect to a set of anchor nodes.  ...  Rather, they are used as additional features along with node embeddings for node classification.  ... 
arXiv:2008.09657v4 fatcat:ovfr6orujzbjxb7wwnqacqy6ga

Batch Mode Active Learning for Multimedia Pattern Recognition

Shayok Chakraborty, Vineeth Balasubramanian, Sethuraman Panchanathan
2012 2012 IEEE International Symposium on Multimedia  
This has expanded the possibility of solving real world problems using computational learning frameworks.  ...  However, while gathering a large amount of data is cheap and easy, annotating them with class labels is an expensive process in terms of time, labor and human expertise.  ...  ACKNOWLEDGEMENTS My tenure at Arizona State University has been influenced and guided by a number of people to whom I am deeply indebted.  ... 
doi:10.1109/ism.2012.101 dblp:conf/ism/ChakrabortyBP12 fatcat:kvr4sjlulrcv5cdwtrapadskm4

Discrete Energy Minimization, beyond Submodularity: Applications and Approximations [article]

Shai Bagon
2012 arXiv   pre-print
In this thesis I explore challenging discrete energy minimization problems that arise mainly in the context of computer vision tasks.  ...  Experiments show that these new methods yield good results for representative challenging problems.  ...  We show a remarkable improvement for ICM combined in our multiscale framework compared with a single-scale scheme.  ... 
arXiv:1210.7362v2 fatcat:6txhssbravatphb7tu6tedp6rm

Deep Metric Learning via Facility Location

Hyun Oh Song, Stefanie Jegelka, Vivek Rathod, Kevin Murphy
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
In this paper, we propose a new metric learning scheme, based on structured prediction, that is aware of the global structure of the embedding space, and which is designed to optimize a clustering quality  ...  However, current methods, all focus on a very local view of the data.  ...  Maximizing equation 4 with respect to subset S is NP-hard, but there is a well established worst case optimality bound of O 1 − 1 e for the greedy solution of the problem via submodularity [17] .  ... 
doi:10.1109/cvpr.2017.237 dblp:conf/cvpr/SongJR017 fatcat:wx64w3vpt5c45oy36ahxpsfdwa

Quality meets Diversity: A Model-Agnostic Framework for Computerized Adaptive Testing [article]

Haoyang Bi, Haiping Ma, Zhenya Huang, Yu Yin, Qi Liu, Enhong Chen, Yu Su, Shijin Wang
2021 arXiv   pre-print
In this paper, we study a novel model-agnostic CAT problem, where we aim to propose a flexible framework that can adapt to different cognitive models.  ...  It shows the advantage of tailoring a personalized testing procedure for each examinee, which selects questions step by step, depending on her performance.  ...  Starting with a machine learning model and a data selection strategy, at each step, AL framework selects a batch of unlabeled data to be annotated for supplementing the limited labeled data so that the  ... 
arXiv:2101.05986v1 fatcat:kn7yzy2ljbc5pb2bmmpalizdp4

Deep Metric Learning via Facility Location [article]

Hyun Oh Song, Stefanie Jegelka, Vivek Rathod, Kevin Murphy
2017 arXiv   pre-print
We propose a global metric learning scheme for optimizing the deep metric embedding with the learnable clustering function and the clustering metric (NMI) in a novel structured prediction framework.  ...  Learning the representation and the similarity metric in an end-to-end fashion with deep networks have demonstrated outstanding results for clustering and retrieval.  ...  Maximizing equation 4 with respect to subset S is NP-hard, but there is a well established worst case optimality bound of O 1 − 1 e for the greedy solution of the problem via submodularity [17] .  ... 
arXiv:1612.01213v2 fatcat:wntspogij5gvjiayxl5sbjfaty

Actively learning to infer social ties

Honglei Zhuang, Jie Tang, Wenbin Tang, Tiancheng Lou, Alvin Chin, Xia Wang
2012 Data mining and knowledge discovery  
For a relationship, we compute the number of emails for different communication types.  ...  It is also supported by a research funding from Nokia Research Center. Appendix: Feature definition In this section, we introduce how we define the attribute factor functions.  ...  We define the problem in a semi-supervised framework, and propose a Partially-Labeled Pairwise Factor Graph Model (PLP-FGM) for learning to infer the relationship semantics.  ... 
doi:10.1007/s10618-012-0274-x fatcat:eoao4kyepjdo5h4qvhybbzllza

Robust Counterfactual Explanations on Graph Neural Networks [article]

Mohit Bajaj, Lingyang Chu, Zi Yu Xue, Jian Pei, Lanjun Wang, Peter Cho-Ho Lam, Yong Zhang
2022 arXiv   pre-print
Massive deployment of Graph Neural Networks (GNNs) in high-stake applications generates a strong demand for explanations that are robust to noise and align well with human intuition.  ...  These explanations are not robust to noise because independently optimizing the correlation for a single input can easily overfit noise.  ...  The problem in Equation ( 4 ) can be proven to be a Submodular Cost Submodular Cover (SCSC) problem [18] (see Appendix D for proof) that is well known to be NP-hard [5] .  ... 
arXiv:2107.04086v3 fatcat:3r35ia6jdveuxfjncfapbjba5u

Information Theoretic Measures for Fairness-aware Feature Selection [article]

Sajad Khodadadian, Mohamed Nafea, AmirEmad Ghassami, Negar Kiyavash
2021 arXiv   pre-print
In this work, we develop a framework for fairness-aware feature selection which takes into account the correlation among the features and the decision outcome, and is based on information theoretic measures  ...  Finally, we design a fairness utility score for each feature (for feature selection) which quantifies how this feature influences accurate as well as nondiscriminatory decisions.  ...  Challenges and our contributions Our goal is to develop a framework for fairness-aware feature selection by computing a score for each feature which captures both its accuracy and discriminatory impacts  ... 
arXiv:2106.00772v2 fatcat:s423jrciivhplcqfs3tbi2vezy
« Previous Showing results 1 — 15 out of 234 results