1,494 Hits in 3.4 sec

Less is more: Selecting informative and diverse subsets with balancing constraints [article]

Srikumar Ramalingam, Daniel Glasner, Kaushal Patel, Raviteja Vemulapalli, Sadeep Jayasumana, Sanjiv Kumar
2021 arXiv   pre-print
Prior methods have exploited diversity and uncertainty in submodular objective functions for choosing subsets.  ...  In addition to these measures, we show that balancing constraints on predicted class labels and decision boundaries are beneficial.  ...  Acknowledgement We thank Gui Citovsky, Pranjal Awasthi, Chen Wang, Ramin Zabih, Aditya Menon, Hossein Mobahi, Dilip Krishnan, Xin Yu, and Sophia Domokos for valuable feedback.  ... 
arXiv:2104.12835v2 fatcat:ytcdgiepzreerkaj64hwofgcbq

Dynamic Thresholding for Online Distributed Data Selection [article]

Mariel A. Werner, Anastasios Angelopoulos, Stephen Bates, Michael I. Jordan
2022 arXiv   pre-print
Specifically, we develop algorithms for the online and distributed version of the problem, where data selection occurs in an uncoordinated fashion across multiple data streams.  ...  Finally, in learning tasks on ImageNet and MNIST, we show that our selection methods outperform random selection by 5-20%.  ...  Consequently, we design our value function to select class-balanced subsets from the data stream.  ... 
arXiv:2201.10547v2 fatcat:7rgjk6nxone6xa2jvplhnurgle

Submodularity in Data Subset Selection and Active Learning

Kai Wei, Rishabh K. Iyer, Jeff A. Bilmes
2015 International Conference on Machine Learning  
data subset selection framework.  ...  We show the connection of submodularity to the data likelihood functions for Naïve Bayes (NB) and Nearest Neighbor (NN) classifiers, and formulate the data subset selection problems for these classifiers  ...  IIS-1162606, the National Institutes of Health under award R01GM103544, and by a Google, a Microsoft, and an Intel research award. R.  ... 
dblp:conf/icml/WeiIB15 fatcat:uaicd43ffrcylpalzwxlnqs6gy

Streaming Submodular Maximization with Fairness Constraints [article]

Yanhao Wang and Francesco Fabbri and Michael Mathioudakis
2020 arXiv   pre-print
submodular function subject to a cardinality constraint – i.e., the size of the selected subset is restricted to be smaller than or equal to an input integer k.  ...  In this paper, we consider the problem with additional fairness constraints, which takes into account the group membership of data items and limits the number of items selected from each group to a given  ...  Yanhao Wang and Michael Mathioudakis have been supported by the MLDB project of Academy of Finland (decision number: 322046).  ... 
arXiv:2010.04412v1 fatcat:hzd5qcuwt5exzexb5lhxfnr27m

A submodular optimization approach to sentence set selection

Yusuke Shinohara
2014 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
In this paper, we propose a near-optimal method for selecting sentence sets along this approach. We first define our objective function, and show it to be a submodular function.  ...  The problem of designing phonetically-balanced sentence sets has been studied extensively in the past.  ...  In other words, there is no data like more and balanced data.  ... 
doi:10.1109/icassp.2014.6854375 dblp:conf/icassp/Shinohara14 fatcat:eir7a4pnw5cs7niozj37bx6yvu

GLISTER: Generalization based Data Subset Selection for Efficient and Robust Learning [article]

Krishnateja Killamsetty, Durga Sivasubramanian, Ganesh Ramakrishnan, Rishabh Iyer
2021 arXiv   pre-print
In this work, we introduce Glister, a GeneraLIzation based data Subset selecTion for Efficient and Robust learning framework.  ...  We then show that for a rich class of loss functions including cross-entropy, hinge-loss, squared-loss, and logistic-loss, the inner discrete data selection is an instance of (weakly) submodular optimization  ...  with a submodular data subset selection framework to label a subset of data points to train a classifier.  ... 
arXiv:2012.10630v4 fatcat:mvsyt6nvejbqla4zblqie3slja

Attackability Characterization of Adversarial Evasion Attack on Discrete Data

Yutong Wang, Yufei Han, Hongyan Bao, Yun Shen, Fenglong Ma, Jin Li, Xiangliang Zhang
2020 Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining  
We characterize the attackability of a targeted classifier on discrete data in evasion attack by bridging the attackability measurement and the regularity of the targeted classifier.  ...  Our study is inspired by the weak submodularity theory.  ...  The celebrated work by [13] sets up an equivalence between the smoothness and strongly concavity of log-likelihood objective functions and its weak submodularity in a feature subset selection problem  ... 
doi:10.1145/3394486.3403194 fatcat:u7gh7otodffkvhxcoe75rw67ki

Fairness in Streaming Submodular Maximization: Algorithms and Hardness [article]

Marwa El Halabi, Slobodan Mitrović, Ashkan Norouzi-Fard, Jakab Tardos, Jakub Tarnawski
2020 arXiv   pre-print
Submodular maximization has become established as the method of choice for the task of selecting representative and diverse summaries of data.  ...  To this end, we develop the first streaming approximation algorithms for submodular maximization under fairness constraints, for both monotone and non-monotone functions.  ...  Acknowledgments and Disclosure of Funding  ... 
arXiv:2010.07431v2 fatcat:tzmdn4rnljcepa3nuew7m476tq

Near-Optimal Distributed Estimation for a Network of Sensing Units Operating Under Communication Constraints [article]

Abolfazl Hashemi, Osman Fatih Kilic, Haris Vikalo
2018 arXiv   pre-print
By leveraging the notion of weak submodularity, we develop an efficient greedy algorithm for the proposed formulation and show that the greedy algorithm achieves a constant factor approximation of the  ...  This problem is formulated as the maximization of a monotone objective function subject to a cardinality constraint.  ...  There, one is interested in the design of an optimal estimation scheme under communication constraints for a single unit (i.e., the fusion center) which collects sensor data.  ... 
arXiv:1807.07650v1 fatcat:axv7cuazcnbczijibegkb7kjva

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.  ...  Maximizing a submodular function subject to a constraint selects a subset that is non-redundant and thus not wasteful.  ... 
arXiv:2202.00132v1 fatcat:sp4b3ww3ajdxvfgigp7xw4f4yq

Submodularity in Action: From Machine Learning to Signal Processing Applications [article]

Ehsan Tohidi, Rouhollah Amiri, Mario Coutino, David Gesbert, Geert Leus, Amin Karbasi
2020 arXiv   pre-print
We introduce a variety of submodular-friendly applications, and elucidate the relation of submodularity to convexity and concavity which enables efficient optimization.  ...  With a mixture of theory and practice, we present different flavors of submodularity accompanying illustrative real-world case studies from modern SP and ML.  ...  In Fig. 6 , the R 2 z,S values for the selected subsets by different methods for sizes k ∈ {2, . . . , 8} is shown for two different data sets. VI.  ... 
arXiv:2006.09905v1 fatcat:ksn2bqbdczechktpa6ivcpwcau

Constrained Robust Submodular Partitioning

Shengjie Wang, Tianyi Zhou, Chandrashekhar Lavania, Jeff A. Bilmes
2021 Neural Information Processing Systems  
For example, when partitioning data for distributed training, we can add a constraint that the number of samples of each class is the same in each partition block, ensuring data balance.  ...  We study an extension of the robust submodular partition problem with additional constraints (e.g., cardinality, multiple matroids, and/or knapsack) on every block.  ...  We propose two classes of algorithms, Min-Block Greedy and Round-Robin Greedy based, and prove approximation bounds under various constraints.  ... 
dblp:conf/nips/WangZLB21 fatcat:k2afsth4kbbhhjul3x74rr35fa

Entropy rate superpixel segmentation

Ming-Yu Liu, Oncel Tuzel, Srikumar Ramalingam, Rama Chellappa
2011 CVPR 2011  
By exploiting submodular and monotonic properties of the objective function, we develop an efficient greedy algorithm.  ...  This objective function consists of two components: entropy rate of a random walk on a graph and a balancing term.  ...  Jacobs and Dr. Yuichi Taguchi for their valuable discussions and suggestions. This research was mostly conducted at MERL with support from MERL. Prof.  ... 
doi:10.1109/cvpr.2011.5995323 dblp:conf/cvpr/LiuTRC11 fatcat:xos7ilgefrdc3btgh7mdkzeuka

Data Subset Selection For Efficient Svm Training

Sara Mourad, Ahmed Tewfik, Haris Vikalo
2018 Zenodo  
Publication in the conference proceedings of EUSIPCO, Kos island, Greece, 2017  ...  In [30] , submodular maximization is used for data subset selection for two classification algorithms, Naive Bayes and Nearest Neighbors, by maximizing the likelihood of the entire data set under those  ...  DATA SUBSET SELECTION A.  ... 
doi:10.5281/zenodo.1160174 fatcat:qsex46vdf5dnxmibj6h4rkefnm

A Parallel Approximation Algorithm for Maximizing Submodular b-Matching [article]

S M Ferdous, Alex Pothen, Arif Khan, Ajay Panyala, Mahantesh Halappanavar
2021 arXiv   pre-print
We design new serial and parallel approximation algorithms for computing a maximum weight b-matching in an edge-weighted graph with a submodular objective function.  ...  We have applied the approximate submodular b-matching algorithm to assign tasks to processors in the computation of Fock matrices in quantum chemistry on parallel computers.  ...  We will show that a submodular objective with these b-Matching constraints implicitly encodes the desired load balance.  ... 
arXiv:2107.05793v1 fatcat:wz4s267exbgevdjzszq36moea4
« Previous Showing results 1 — 15 out of 1,494 results