Training Mixed-Domain Translation Models via Federated Learning

被引:0
|
作者
Passban, Peyman [1 ]
Roosta, Tanya [1 ]
Gupta, Rahul [1 ]
Chadha, Ankit [1 ]
Chung, Clement [1 ]
机构
[1] Amazon, Seattle, WA 98121 USA
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Training mixed-domain translation models is a complex task that demands tailored architectures and costly data preparation techniques. In this work, we leverage federated learning (FL) in order to tackle the problem. Our investigation demonstrates that with slight modifications in the training process, neural machine translation (NMT) engines can be easily adapted when an FL-based aggregation is applied to fuse different domains. Experimental results also show that engines built via FL are able to perform on par with state-of-the-art baselines that rely on centralized training techniques. We evaluate our hypothesis in the presence of five datasets with different sizes, from different domains, to translate from German into English and discuss how FL and NMT can mutually benefit from each other. In addition to providing benchmarking results on the union of FL and NMT, we also propose a novel technique to dynamically control the communication bandwidth by selecting impactful parameters during FL updates. This is a significant achievement considering the large size of NMT engines that need to be exchanged between FL parties.
引用
收藏
页码:2576 / 2586
页数:11
相关论文
共 50 条
  • [31] A Model Predictive Control Based Path Tracker in Mixed-Domain
    Hu, Jia
    Feng, Yongwei
    Li, Xin
    Wang, Haoran
    2021 32ND IEEE INTELLIGENT VEHICLES SYMPOSIUM (IV), 2021, : 1255 - 1260
  • [32] A novel design methodology for the mixed-domain optimization of a MEMS accelerometer
    Pak, Murat
    Fernandez, Francisco, V
    Dundar, Gunhan
    INTEGRATION-THE VLSI JOURNAL, 2018, 62 : 314 - 321
  • [33] Pisces: Efficient Federated Learning via Guided Asynchronous Training
    Jiang, Zhifeng
    Wang, Wei
    Li, Baochun
    Li, Bo
    PROCEEDINGS OF THE 13TH SYMPOSIUM ON CLOUD COMPUTING, SOCC 2022, 2022, : 370 - 385
  • [34] Mixed-Signal and Mixed-Domain Instrumentation for Emerging Technology Device Characterization
    Ribeiro, Diogo
    Boaventura, Alirio
    Cruz, Pedro
    Carvalho, Nuno Borges
    2014 XXXITH URSI GENERAL ASSEMBLY AND SCIENTIFIC SYMPOSIUM (URSI GASS), 2014,
  • [35] Mitigating Demographic Bias of Federated Learning Models via Robust-Fair Domain Smoothing: A Domain-Shifting Approach
    Zeng, Huimin
    Yue, Zhenrui
    Jiang, Qian
    Zhang, Yang
    Shang, Lanyu
    Zong, Ruohan
    Wang, Dong
    2024 IEEE 44TH INTERNATIONAL CONFERENCE ON DISTRIBUTED COMPUTING SYSTEMS, ICDCS 2024, 2024, : 785 - 796
  • [36] LMA: lightweight mixed-domain attention for efficient network design
    Yang Yu
    Yi Zhang
    Zhe Song
    Cheng-Kai Tang
    Applied Intelligence, 2023, 53 : 13432 - 13451
  • [37] LMA: lightweight mixed-domain attention for efficient network design
    Yu, Yang
    Zhang, Yi
    Song, Zhe
    Tang, Cheng-Kai
    APPLIED INTELLIGENCE, 2023, 53 (11) : 13432 - 13451
  • [38] All-Digital Transmitter With a Mixed-Domain Combination Filter
    Cordeiro, R. F.
    Oliveira, Arnaldo S. R.
    Vieira, J. M. N.
    IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS, 2016, 63 (01) : 4 - 8
  • [39] Simulation of PEM fuel cells using the mixed-domain model
    Meng, Hua
    Wuhan Ligong Daxue Xuebao/Journal of Wuhan University of Technology, 2006, 28 (SUPPL. 2): : 587 - 591
  • [40] A novel evolutionary engineering design approach for mixed-domain systems
    Fan, Z
    Seo, KS
    Hu, JJ
    Goodman, ED
    Rosenberg, RC
    ENGINEERING OPTIMIZATION, 2004, 36 (02) : 127 - 147