222 Hits in 4.1 sec

Debiasing Model Updates for Improving Personalized Federated Training

Durmus Alp Emre Acar, Yue Zhao, Ruizhao Zhu, Ramon Matas Navarro, Matthew Mattina, Paul N. Whatmough, Venkatesh Saligrama
2021 International Conference on Machine Learning  
The trained global meta-model is then personalized locally by each device to meet its specific objective.  ...  Different from the conventional federated learning setting, training customized models for each device is hindered by both the inherent data biases of the various devices, as well as the requirements imposed  ...  We can infer that PFL is capable of debiasing meta-model updates at the server allowing for superior device personalization.  ... 
dblp:conf/icml/AcarZZNMWS21 fatcat:73ku5gefqzeizdbuk6uh6r6anu

Generating Data to Mitigate Spurious Correlations in Natural Language Inference Datasets [article]

Yuxiang Wu, Matt Gardner, Pontus Stenetorp, Pradeep Dasigi
2022 arXiv   pre-print
Results show that models trained on our debiased datasets generalise better than those trained on the original datasets in all settings.  ...  We propose to tackle this problem by generating a debiased version of a dataset, which can then be used to train a debiased, off-the-shelf model, by simply replacing its training data.  ...  Acknowledgments The authors would like to thank Max Bartolo, Alexis Ross, Doug Downey, Jesse Dodge, Pasquale Minervini, and Sebastian Riedel for their helpful discussion and feedback.  ... 
arXiv:2203.12942v1 fatcat:2rcxn3wwmfav5jzyg3tcuz3ie4

Algorithmic Fairness and Bias Mitigation for Clinical Machine Learning: Insights from Rapid COVID-19 Diagnosis by Adversarial Learning [article]

Jenny Yang, Andrew AS Soltan, Yang Yang, David A Clifton
2022 medRxiv   pre-print
For example, if one class is over-presented or errors/inconsistencies in practice are reflected in the training data, then a model can be biased by these.  ...  We trained our framework on a large, real-world COVID-19 dataset and demonstrated that adversarial training demonstrably improves outcome fairness (with respect to equalized odds), while still achieving  ...  gratitude to Jingyi Wang & Dr Jolene Atia at University Hospitals Birmingham NHS Foundation trust, Phillip Dickson at Bedfordshire Hospitals, and Paul Meredith at Portsmouth Hospitals University NHS Trust for  ... 
doi:10.1101/2022.01.13.22268948 fatcat:g65bmzfly5db7mrxuqgep4dhki

Federated Learning of Molecular Properties with Graph Neural Networks in a Heterogeneous Setting [article]

Wei Zhu, Jiebo Luo, Andrew White
2022 arXiv   pre-print
FLIT(+) can align the local training across heterogeneous clients by improving the performance for uncertain samples.  ...  Federated learning allows end-users to build a global model collaboratively while keeping the training data distributed over isolated clients.  ...  Client-side Updates For completeness, we describe typical training steps to update the GNN model for client side training.  ... 
arXiv:2109.07258v3 fatcat:e4dntvhf7jcsdmnbn67wum23xa

Real-Time Decentralized knowledge Transfer at the Edge [article]

Orpaz Goldstein, Mohammad Kachuee, Derek Shiell, Majid Sarrafzadeh
2021 arXiv   pre-print
We propose a method based on knowledge distillation for pairwise knowledge transfer pipelines from models trained on non-i.i.d. data and compare it to other popular knowledge transfer methods.  ...  Transferring knowledge in a selective decentralized approach enables models to retain their local insights, allowing for local flavors of a machine learning model.  ...  Additionally, while federated solutions allow for some consolidation of models, federation will retract the benefit of private local models or personalized individual models.  ... 
arXiv:2011.05961v4 fatcat:zjcpc5ipqzbkxpzyfrcegny7su

Personalized News Recommendation: Methods and Challenges [article]

Chuhan Wu, Fangzhao Wu, Yongfeng Huang, Xing Xie
2022 arXiv   pre-print
Next, we introduce the public datasets and evaluation methods for personalized news recommendation.  ...  We first review the techniques for tackling each core problem in a personalized news recommender system and the challenges they face.  ...  The local model updates are uploaded to a central server that coordinates a number of user clients for model training.  ... 
arXiv:2106.08934v3 fatcat:iagqsw73hrehxaxpvpydvtr26m

Algorithm Fairness in AI for Medicine and Healthcare [article]

Richard J. Chen, Tiffany Y. Chen, Jana Lipkova, Judy J. Wang, Drew F.K. Williamson, Ming Y. Lu, Sharifa Sahai, Faisal Mahmood
2022 arXiv   pre-print
Lastly, we also review emerging technology for mitigating bias via federated learning, disentanglement, and model explainability, and their role in AI-SaMD development.  ...  Recent evaluation of AI models stratified across race sub-populations have revealed inequalities in how patients are diagnosed, given treatments, and billed for healthcare costs.  ...  site in updating the global model and the varying frequencies at which different sites participate in training 209 .  ... 
arXiv:2110.00603v2 fatcat:pspb6bqqxjh45an5mhqohysswu

A Comprehensive Survey on Trustworthy Graph Neural Networks: Privacy, Robustness, Fairness, and Explainability [article]

Enyan Dai, Tianxiang Zhao, Huaisheng Zhu, Junjie Xu, Zhimeng Guo, Hui Liu, Jiliang Tang, Suhang Wang
2022 arXiv   pre-print
For example, existing works demonstrate that attackers can fool the GNNs to give the outcome they desire with unnoticeable perturbation on training graph.  ...  For each aspect, we give the taxonomy of the related methods and formulate the general frameworks for the multiple categories of trustworthy GNNs.  ...  Then, federated learning methods are leveraged to further update the global model.  ... 
arXiv:2204.08570v1 fatcat:7c3pkxitrbhgxj6fytn6f3r644

Disentangled Federated Learning for Tackling Attributes Skew via Invariant Aggregation and Diversity Transferring [article]

Zhengquan Luo, Yunlong Wang, Zilei Wang, Zhenan Sun, Tieniu Tan
2022 arXiv   pre-print
To cope with these, we proposed disentangled federated learning (DFL) to disentangle the domain-specific and cross-invariant attributes into two complementary branches, which are trained by the proposed  ...  Importantly, convergence analysis proves that the FL system can be stably converged even if incomplete client models participate in the global aggregation, which greatly expands the application scope of  ...  During each process, the parameters of some model parts are frozen for more targeted training.  ... 
arXiv:2206.06818v1 fatcat:4d75lcz45ved5aqmtjjsg34ofu

No Fear of Heterogeneity: Classifier Calibration for Federated Learning with Non-IID Data [article]

Mi Luo, Fei Chen, Dapeng Hu, Yifan Zhang, Jian Liang, Jiashi Feng
2021 arXiv   pre-print
after federated training.  ...  A central challenge in training classification models in the real-world federated system is learning with non-IID data.  ...  Acknowledgement We would like to thank the anonymous reviewers for their insightful comments and suggestions.  ... 
arXiv:2106.05001v2 fatcat:fmbnjqpmdbc6dj3mlc6usg3nhe

Personalized News Recommendation: Methods and Challenges

Chuhan Wu, Fangzhao Wu, Yongfeng Huang, Xing Xie
2022 ACM Transactions on Information Systems  
Next, we introduce the public datasets and evaluation methods for personalized news recommendation.  ...  Personalized news recommendation is important for users to find interested news information and alleviate information overload.  ...  The local model updates are uploaded to a central server that coordinates a number of user clients for model training.  ... 
doi:10.1145/3530257 fatcat:xzghh6cut5ahhgxz4mkzgy74ja

Generative Models for Effective ML on Private, Decentralized Datasets [article]

Sean Augenstein, H. Brendan McMahan, Daniel Ramage, Swaroop Ramaswamy, Peter Kairouz, Mingqing Chen, Rajiv Mathews, Blaise Aguera y Arcas
2020 arXiv   pre-print
To improve real-world applications of machine learning, experienced modelers develop intuition about their datasets, their models, and how the two interact.  ...  This paper demonstrates that generative models - trained using federated methods and with formal differential privacy guarantees - can be used effectively to debug many commonly occurring data issues even  ...  job is to develop and improve the machine learned models.  ... 
arXiv:1911.06679v2 fatcat:qdupc7zyh5gwpgu5yj2fim2kdu

Intrinisic Gradient Compression for Federated Learning [article]

Luke Melas-Kyriazi, Franklyn Wang
2021 arXiv   pre-print
Federated learning is a rapidly-growing area of research which enables a large number of clients to jointly train a machine learning model on privately-held data.  ...  One of the largest barriers to wider adoption of federated learning is the communication cost of sending model updates from and to the clients, which is accentuated by the fact that many of these devices  ...  in model training.  ... 
arXiv:2112.02656v1 fatcat:bmkxosl22rgnln5ikdbaayzofi

PrivFairFL: Privacy-Preserving Group Fairness in Federated Learning [article]

Sikha Pentyala, Nicola Neophytou, Anderson Nascimento, Martine De Cock, Golnoosh Farnadi
2022 arXiv   pre-print
In doing so, we propose a method for training group-fair ML models in cross-device FL under complete and formal privacy guarantees, without requiring the clients to disclose their sensitive attribute values  ...  Achieving group fairness in Federated Learning (FL) is challenging because mitigating bias inherently requires using the sensitive attribute values of all clients, while FL is aimed precisely at protecting  ...  Research Award for Privacy Enhancing Technologies, and the Google Cloud Research Credits Program.  ... 
arXiv:2205.11584v1 fatcat:ii4dzz6qtvdjtgxsmjbq7drgai

Worker overconfidence: Field evidence and implications for employee turnover and firm profits

Mitchell Hoffman, Stephen V. Burks
2020 Quantitative Economics  
To study the implications of overconfidence for worker welfare and firm profits, we estimate a structural learning model with biased beliefs that accounts for many key features of the data.  ...  Combining weekly productivity data with weekly productivity beliefs for a large sample of truckers over 2 years, we show that workers tend to systematically and persistently overpredict their productivity  ...  For work on overoptimism and stock options for nonexecutive workers see, for example, Oyer and Schaefer (2005) . 344 Hoffman and Burks Quantitative Economics 11 (2020)  ... 
doi:10.3982/qe834 fatcat:e65kk25asvaslbjprtyf2bpawu
« Previous Showing results 1 — 15 out of 222 results