Filters








431 Hits in 0.92 sec

Byzantine-Resilient SGD in High Dimensions on Heterogeneous Data [article]

Deepesh Data, Suhas Diggavi
2020 arXiv   pre-print
We consider the heterogeneous data model, where different workers may have different local datasets, and we do not make any probabilistic assumptions on data generation.  ...  In order to be able to apply their filtering procedure in our heterogeneous data setting where workers compute stochastic gradients, we derive a new matrix concentration result, which may be of independent  ...  data.  ... 
arXiv:2005.07866v1 fatcat:ixlbiuqipfcztchkwaewrej77q

Analysis of Count Data by Transmuted Geometric Distribution [article]

Subrata Chakraborty, Deepesh Bhati
2016 arXiv   pre-print
We have demonstrate two applications of (TGD) in modeling real life count data.  ...  Data Analysis For the purpose of illustration, in this section, we consider following two data sets with details as follows: i.  ...  Let the complete-data be constituted with observed set of values y = (y 1 , · · · , y n ) and the hypothetical data set x = (x 1 , · · · , x n ), where the observations y i 's are distributed with random  ... 
arXiv:1610.07123v1 fatcat:cgwxkxbl55a33jigdrkamzdxja

Data Encoding for Byzantine-Resilient Distributed Optimization [article]

Deepesh Data, Linqi Song, Suhas Diggavi
2020 arXiv   pre-print
PGD is typically used in the data-parallel setting, where data is partitioned across different samples, whereas, CD is used in the model-parallelism setting, where data is partitioned across the parameter  ...  We develop a sparse encoding scheme which enables computationally efficient data encoding and decoding.  ...  Acknowledgements The work of Deepesh Data and Suhas Diggavi was partially supported by the Army Research Laboratory under Cooperative Agreement W911NF-17-2-0196, by the UC-NL grant LFR-18-548554, and by  ... 
arXiv:1907.02664v2 fatcat:ctptgj7gpfelraygvetdxrpy4m

Flexible Accuracy for Differential Privacy [article]

Aman Bansal, Rahul Chunduru, Deepesh Data, Manoj Prabhakaran
2021 arXiv   pre-print
Differential Privacy (DP) has become a gold standard in privacy-preserving data analysis.  ...  In particular, we illustrate an application to differentially private histograms, which in turn yields mechanisms for revealing the support of a dataset or the extremal values in the data.  ...  Acknowledgements The work of Deepesh Data was supported in part by NSF grants #1740047, #2007714, and UC-NL grant LFR-18-548554.  ... 
arXiv:2110.09580v1 fatcat:6hrekblz5bg2bfutupbwp5gcmi

Secure Computation of Randomized Functions: Further Results [article]

Deepesh Data, Vinod M. Prabhakaran
2017 arXiv   pre-print
Recently, Data (ISIT 2016) studied the remaining two cases (first, when privacy conditions are against both the users; and second, when privacy condition is only against Alice) and obtained single-letter  ...  INTRODUCTION Two-user secure computation allows mutually distrusting users to jointly perform computation of their private data interactively, in such a way that no individual learns anything beyond the  ...  Data [13] studied two variations of the same problem (one with privacy against both the users, and other with privacy only against Alice) with a probability distribution on inputs and gave rate-optimal  ... 
arXiv:1705.07081v1 fatcat:zuxpilcqk5e3bni2gs3b6bfrty

Successive Refinement of Privacy [article]

Antonious M. Girgis, Deepesh Data, Kamalika Chaudhuri, Christina Fragouli, Suhas Diggavi
2020 arXiv   pre-print
We call this setting successive refinement of privacy, as it provides hierarchical access to the raw data with different privacy levels.  ...  itself or to a function or statistic computed directly on the data.  ...  This approach enables efficient hierarchical access to the data (for example, when analysts have different levels of authorized access).  ... 
arXiv:2005.11651v1 fatcat:eirqcetcwzcpdab4o7flyq4vhi

Byzantine-Resilient High-Dimensional Federated Learning [article]

Deepesh Data, Suhas Diggavi
2020 arXiv   pre-print
probabilistic assumptions on data generation.  ...  We provide convergence analyses for strongly-convex and non-convex smooth objectives in the heterogeneous data setting, where different clients may have different local datasets, and we do not make any  ...  ., mobiles devices, organizations, etc.) collaboratively learn a machine learning model, where the training process is facilitated by the data held by the participating clients (without data centralization  ... 
arXiv:2006.13041v2 fatcat:ff7m23x625dqvlyv5kfpbqbjxq

A discrete probability model suitable for both symmetric and asymmetric count data

Deepesh Bhati, Subrata Chakraborty, Snober Lateef
2020 Filomat  
In this paper, an alternative discrete probability model, namely the discrete skew logistic distribution, suitable for both asymmetric and symmetric count data is proposed.  ...  Table 5 gives the statistical description of both the data sets and it is clear that coefficient of skewness for data set 1 is positive and for data set 2 is negative.  ...  Skewness Kurtosis Data set 1 -2 0 0 0.421 1 3 1.067 0.299 3.4626 Data set 2 -3 1 3 2.625 4 7 2.562 -0.381 2.5921 Table 6 : 6 Data fitting for dataset 1 parameter(mse) DSL(µ, p, q)  ... 
doi:10.2298/fil2008559b fatcat:x2d5dnxbnrh5vetqjbxr6oenra

Analysis of Count Data by Transmuted Geometric Distribution

Subrata Chakraborty, Deepesh Bhati
2019 Journal of Statistical Theory and Applications (JSTA)  
We have demonstrate two applications of (TGD) in modeling real life count data.  ...  DATA ANALYSIS For the purpose of illustration, in this section, we consider following two data sets with details as follows: Number of Fires in Greece (NTG) The data comprise of numbers of fires in district  ...  Insurance Claim Count (ICC) An insurance count data from Belgium in year 1993 is considered [13] and the data is as follows: 0(57178), 1(5617), 2(446), 3(50), 4(8).  ... 
doi:10.2991/jsta.d.191218.001 fatcat:wkyi55xcqvcxfmvx7ib4pwbu7i

A Uniform In Vitro Efficacy Dataset to Guide Antimicrobial Peptide Design

Deepesh Nagarajan, Tushar Nagarajan, Neha Nanajkar, Nagasuma Chandra
2019 Data  
Computational antimicrobial peptide design attempts typically use such data. However, utilizing heterogenous data aggregated from different sources presents significant drawbacks.  ...  We draw simple inferences from this data, and we discuss what characteristics are essential for antimicrobial peptide efficacy.  ...  Availability of CD data: Raw circular dichroism (CD) data has been made available in the supplementary information (Dataset S1).  ... 
doi:10.3390/data4010027 fatcat:liupru7dv5arbec4acahhpewvq

Renyi Differential Privacy of the Subsampled Shuffle Model in Distributed Learning [article]

Antonious M. Girgis, Deepesh Data, Suhas Diggavi
2021 arXiv   pre-print
We also demonstrate numerically significant improvement in privacy-learning performance operating point using real data sets.  ...  Motivated by stochastic optimization and the federated learning (FL) paradigm, we focus on the case where a small fraction of data samples are randomly sub-sampled in each round to participate in the learning  ...  This is because no client has access to enough data to build rich learning models locally and we do not want to directly share local data.  ... 
arXiv:2107.08763v1 fatcat:nqm5kksvdncqjeupmkszhib22u

On the Renyi Differential Privacy of the Shuffle Model [article]

Antonious M. Girgis, Deepesh Data, Suhas Diggavi, Ananda Theertha Suresh, Peter Kairouz
2021 arXiv   pre-print
Introduction Differential privacy (DP) [DMNS06] gives a principled and rigorous framework for data privacy by giving guarantees on the information leakage for individual data points from the output of  ...  Problem Formulation Let D = (d 1 , . . . , d n ) be a dataset consisting of n data points, where d i is a data point at the i'th client that takes values from a set X .  ... 
arXiv:2105.05180v1 fatcat:f7ncpfe34jcqff5k6q464xiham

On the Communication Complexity of Secure Computation [article]

Deepesh Data, Vinod M. Prabhakaran, Manoj M. Prabhakaran
2014 arXiv   pre-print
Our techniques include the use of a data processing inequality for residual information - i.e., the gap between mutual information and G\'acs-K\"orner common information, a new information inequality for  ...  Similarly, the terms in (17) are evaluated using the the distribution p X of the data X of Alice. The lower bound in (18) does not depend on the distributions p X and p Y of the data.  ...  In [WW08], Wolf and Wullschleger identified (among other things) the following important data processing inequality for residual information.  ... 
arXiv:1311.7584v2 fatcat:ksgnrmbq7zeu3ndj5v63bshcma

Interactive Secure Function Computation [article]

Deepesh Data, Gowtham R. Kurri, Jithin Ravi, Vinod M. Prabhakaran
2020 arXiv   pre-print
Data the input distribution and q Z1Z2|XY specifies the randomized function.  ...  By the data-processing inequality, this minimization can be restricted to W 's ranging over maximal independent sets of G [19] .  ... 
arXiv:1812.03838v3 fatcat:4sphlriuljb4dn6whi5oxinn7i

On coding for secure computing

Deepesh Data, Vinod M. Prabhakaran
2015 2015 IEEE International Symposium on Information Theory (ISIT)  
about other users' data other than what they can infer from their own data and the function value they compute.  ...  In information theoretically secure multiparty computation, several mutually distrusting users want to jointly compute a function of their private data such that users do not learn any additional information  ...  private data to each other.  ... 
doi:10.1109/isit.2015.7282954 dblp:conf/isit/DataP15 fatcat:o7l5jduqobaaxevgx4ea7xv2kq
« Previous Showing results 1 — 15 out of 431 results