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
  • [1] Unsupervised domain adaptation for object detection through mixed-domain and co-training learning
    Wei, Xing
    Qin, Xiongbo
    Zhao, Chong
    Qiao, Xuanyuan
    Lu, Yang
    MULTIMEDIA TOOLS AND APPLICATIONS, 2024, 83 (09) : 25213 - 25229
  • [2] Unsupervised domain adaptation for object detection through mixed-domain and co-training learning
    Xing Wei
    Xiongbo Qin
    Chong Zhao
    Xuanyuan Qiao
    Yang Lu
    Multimedia Tools and Applications, 2024, 83 (9) : 25213 - 25229
  • [3] Sym-Parameterized Dynamic Inference for Mixed-Domain Image Translation
    Chang, Simyung
    Park, SeongUk
    Yang, John
    Kwak, Nojun
    2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2019), 2019, : 4802 - 4810
  • [4] Training of Classification Models via Federated Learning and Homomorphic Encryption
    Angulo, Eduardo
    Marquez, Jose
    Villanueva-Polanco, Ricardo
    SENSORS, 2023, 23 (04)
  • [5] Echidna: Mixed-domain Computational Implementation via Decision Trees
    Merrill, Devon J.
    Garza, Jorge
    Swanson, Steven
    ACM SYMPOSIUM ON COMPUTATIONAL FABRICATION (SCF 2019), 2019,
  • [6] Mixed-domain modeling in Modelica
    Clauss, C
    Elmqvist, H
    Mattsson, SE
    Otter, M
    Schwarz, P
    SYSTEM SPECIFICATION AND DESIGN LANGUAGES: BEST OF FDL '02, 2003, : 29 - 40
  • [7] MIXED-DOMAIN ANALYSIS FOR CIRCUIT SIMULATION
    WALSH, K
    WOLFE, B
    VLSI SYSTEMS DESIGN, 1987, 8 (09): : 44 - &
  • [8] Efficient mixed-domain analysis of electrostatic MEMS
    Li, G
    Aluru, NR
    IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, 2003, 22 (09) : 1228 - 1242
  • [9] Layout verification for mixed-domain integrated MEMS
    Baidya, B
    Mukherjee, T
    IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, 2005, 24 (04) : 563 - 577
  • [10] Efficient mixed-domain analysis of electrostatic MEMS
    Li, G
    Aluru, NR
    IEEE/ACM INTERNATIONAL CONFERENCE ON CAD-02, DIGEST OF TECHNICAL PAPERS, 2002, : 474 - 477