Filters








88,852 Hits in 6.3 sec

Towards Quantifying the Carbon Emissions of Differentially Private Machine Learning [article]

Rakshit Naidu, Harshita Diddee, Ajinkya Mulay, Aleti Vardhan, Krithika Ramesh, Ahmed Zamzam
2021 arXiv   pre-print
This paper investigates the impact of differential privacy on learning algorithms in terms of their carbon footprint due to either longer run-times or failed experiments.  ...  The cost of differential privacy is often a reduced model accuracy and a lowered convergence speed.  ...  Deep learning with differential privacy. Proceedings of the  ... 
arXiv:2107.06946v1 fatcat:7lfrlw3eynhcpl65jh2f4jgg2i

Defending against Reconstruction Attacks through Differentially Private Federated Learning for Classification of Heterogeneous Chest X-Ray Data [article]

Joceline Ziegler, Bjarne Pfitzner, Heinrich Schulz, Axel Saalbach, Bert Arnrich
2022 arXiv   pre-print
To the best of our knowledge, we are the first to directly compare the impact of differentially private training on two different neural network architectures, DenseNet121 and ResNet50.  ...  Both non-private baseline models achieved an area under the receiver operating characteristic curve (AUC) of 0.94 on the binary classification task of detecting the presence of a medical finding.  ...  Acknowledgments: Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) -Projektnummer 491466077  ... 
arXiv:2205.03168v2 fatcat:qppm5zvk6bcpjpwkytetq7bixq

Differential Privacy Has Disparate Impact on Model Accuracy [article]

Eugene Bagdasaryan, Vitaly Shmatikov
2019 arXiv   pre-print
We demonstrate that in the neural networks trained using differentially private stochastic gradient descent (DP-SGD), this cost is not borne equally: accuracy of DP models drops much more for the underrepresented  ...  The cost of differential privacy is a reduction in the model's accuracy.  ...  This research was supported in part by the NSF grants 1611770, 1704296, 1700832, and 1642120, the generosity of Eric and Wendy Schmidt by recommendation of the Schmidt Futures program, and a Google Faculty  ... 
arXiv:1905.12101v2 fatcat:7etrv7slejbhpohbrc2pwfhrry

SOCIAL FINANCE TAXONOMY IN TRANSITION TOWARDS A MORE SUSTAINABLE ECONOMY

Iryna Vasylchuk, Kateryna Slyusarenko, Inta Kotane
2019 SOCIETY INTEGRATION EDUCATION Proceedings of the International Scientific Conference  
for developing the classification of social impact investments and social enterprises.  ...  Based on the results of the study, the most relevant classification features of social enterprises and investments were developed and recommendations were made for integrating social finance taxonomy into  ...  Impact investments can be made by various methods, for example, through social models that create blended value and are small in size, by using a traditional model aligned with the theory of change (PRI  ... 
doi:10.17770/sie2019vol6.3835 fatcat:qzlgkcaqnjeb7fgfuu4dxivqmm

Evaluating Differentially Private Machine Learning in Practice [article]

Bargav Jayaraman, David Evans
2019 arXiv   pre-print
However, implementations of privacy-preserving machine learning often select large values of ϵ in order to get acceptable utility of the model, with little understanding of the impact of such choices on  ...  In this paper, we quantify the impact of these choices on privacy in experiments with logistic regression and neural network models.  ...  Finally, we would also like to thank Congzheng Song and Samuel Yeom for providing their implementation of inference attacks.  ... 
arXiv:1902.08874v4 fatcat:7ic6gclgfnhrrhs2wqn3mqfcwi

Mapping Philanthropic Foundations' Characteristics: Towards an International Integrative Framework of Foundation Types

Tobias Jung, Jenny Harrow, Diana Leat
2018 Nonprofit and Voluntary Sector Quarterly  
As philanthropic foundations take on increasingly prominent socio-political roles, the need for stronger conceptualizations of foundations as an organizational form is articulated widely across academic  ...  Building on institutional research's tradition of categorizing, classifying and typologizing organizational forms, our paper critically explores the different ways in which foundations have been cast and  ...  Based on the understanding that wealthier foundations will have different characteristics and behaviours compared to those that are less well-off, distinguishing between foundations in terms of size and  ... 
doi:10.1177/0899764018772135 fatcat:wsiabi4og5e2nksn3aiingxki4

Towards Better Generalization of Adaptive Gradient Methods

Yingxue Zhou, Belhal Karimi, Jinxing Yu, Zhiqiang Xu, Ping Li
2020 Neural Information Processing Systems  
To close this gap, we propose Stable Adaptive Gradient Descent (SAGD) for non-convex optimization which leverages differential privacy to boost the generalization performance of adaptive gradient methods  ...  We further conduct experiments on various popular deep learning tasks and models. Experimental results illustrate that SAGD is empirically competitive and often better than baselines.  ...  Acknowledgments and Disclosure of Funding We thank the anonymous Referees and Area Chair for their constructive comments. The work is supported by Baidu Research.  ... 
dblp:conf/nips/ZhouKYX020 fatcat:vviewupasrhlfaj3vq7d3tfhci

Toward Training at ImageNet Scale with Differential Privacy [article]

Alexey Kurakin, Shuang Song, Steve Chien, Roxana Geambasu, Andreas Terzis, Abhradeep Thakurta
2022 arXiv   pre-print
The model we use was pretrained on the Places365 data set as a starting point.  ...  Differential privacy (DP) is the de facto standard for training machine learning (ML) models, including neural networks, while ensuring the privacy of individual examples in the training set.  ...  In evaluating Resnet-50 and Resnet-18, we observe that model size impacts accuracy differently than one might expect in non-private training.  ... 
arXiv:2201.12328v2 fatcat:uynzcbrvsbfajneyxg5eff76wa

Adherence of private health system hospitals to dissemination of outcomes according to the Global Reporting Initiative (GRI) model

Celso Machado Junior, Robson Danúbio da Silva César, Maria Tereza Saraiva de Souza
2017 Einstein (São Paulo)  
Objective To verify if there is an analogy between the indicators of the Global Reporting Initiative adopted by hospitals in the private healthcare system.  ...  Results The organizations surveyed had a significant adherence of their economic, social and environmental indicators of the model proposed by the Global Reporting Initiative, showing an analogous field  ...  These indicators point to two attention groups: the first geared towards optimization of resources (energy and water), and the second, towards monitoring and mitigation of the impact of their activities  ... 
doi:10.1590/s1679-45082017gs3989 pmid:29091158 pmcid:PMC5823050 fatcat:kwd4hkiptfhtvdhzkkq5gmxmgu

Detection and Classification of Mammogram using Fusion Model of Multi-View Feature

2020 VOLUME-8 ISSUE-10, AUGUST 2019, REGULAR ISSUE  
The involvement of mammogram classification saves the doctor's and physician's time.  ...  In this paper we have focused on the survey of Convolutional Neural Network (CNN) methods for breast image classification in multiview features.  ...  Impact of various preparing dataset size on the relative execution parameters of the different adjusting strategies. [1] 2.  ... 
doi:10.35940/ijitee.h6160.069820 fatcat:yyzr5ahc4rhstfrn6vmtyuu67e

Integrally Private Model Selection For Decision Trees

Navoda Senavirathne, Vicenç Torra
2019 Computers & security  
Also, we provide recommendations for private model selection based on the accuracy and stability of the models along with the diversity of training data that can be used to generate the models. (N.  ...  One of the primary goals of such attacks is to infer information about the training data used to construct the models.  ...  Differential privacy ( Dwork, 2006 ) has contributed towards privacy preserving machine learning through different approaches such as the implementation of differentially private machine learning algorithms  ... 
doi:10.1016/j.cose.2019.01.006 fatcat:dfm5skbeg5dqrmy7msrhlzcq6i

Student and School-level Predictors of Pharmacy Residency Attainment

Kayley Lyons, Danielle A. Taylor, Lana M. Minshew, Jacqueline E. McLaughlin
2018 American Journal of Pharmaceutical Education  
Logistic multilevel modeling was used to examine the effects of select student and school level characteristics on pharmacy residency attainment, as indicated by students on the AACP Graduating Student  ...  Student and pharmacy school characteristics impact the likelihood of pharmacy residency attainment. Further research is needed to understand the mechanisms associated with these effects.  ...  ACKNOWLEDGMENTS The ideas expressed in this manuscript are those of the authors and do not represent the position or work of the AACP.  ... 
doi:10.5688/ajpe6220 pmid:29606710 pmcid:PMC5869752 fatcat:jfqi4mlokncsvnsss4g3mqvmna

Betting exclusively for private labels: Could it have negative consequences for retailers?

J.L. Ruiz-Real, J.C. Gázquez-Abad, I. Esteban-Millat, F.J. Martínez-López
2017 Spanish Journal of Marketing-ESIC  
Another differentiating feature is the methodology used. Estimation of the structural equation model permits the simultaneous estimation of the relationships between the variables.  ...  Findings -The image of stores that only offer their own brand is mainly configured by price consciousness and the attitude toward the private label.  ...  In light of this, we propose the following hypothesis with regard to PLs: H5. A favorable attitude toward private labels has a direct and positive impact on private label purchase intention.  ... 
doi:10.1016/j.sjme.2016.12.004 fatcat:c7kofvlbczczlklcva24652uui

Enhancing the Privacy of Federated Learning with Sketching [article]

Zaoxing Liu, Tian Li, Virginia Smith, Vyas Sekar
2019 arXiv   pre-print
We evaluate the feasibility of sketching-based federated learning with a prototype on three representative learning models.  ...  However, current methods still share model updates, which may contain private information (e.g., one's weight and height), during the training process.  ...  Depending on the size of the machine learning model, the computed local updates can be a vector of up to millions of numbers.  ... 
arXiv:1911.01812v1 fatcat:mquqgd2ykjepdidtf5bx4rkkpq

Towards Social Role-Based Interruptibility Management [article]

Christoph Anderson, Judith Simone Heinisch, Shohreh Deldari, Flora D. Salim, Sandra Ohly, Klaus David, Veljko Pejovic
2021 arXiv   pre-print
The two-stage interruptibility model achieves a F1 score of 0.70. Finally, we examine the influence of multi-device data on social role and interruptibility classification performances.  ...  We design and evaluate social role classification models based on spatio-temporal and application based features.  ...  Table 2 , shows the results of a paired t-test to investigate the impact of multi-device settings on interruptibility and social role classification.  ... 
arXiv:2106.04265v1 fatcat:a7tfm7xcirgobcxp5aueguhpt4
« Previous Showing results 1 — 15 out of 88,852 results