研究生: |
黃子玹 Zih-Syuan Huang |
---|---|
論文名稱: |
直接端對端基於局部學習輔助之異構模型拆分聯邦轉移學習 Direct Edge-to-Edge Local-learning-assisted Heterogeneous Model-based Split Federated Transfer Learning |
指導教授: |
陸敬互
Ching-Hu Lu |
口試委員: |
蘇順豐
Shun-Feng Su 鍾聖倫 Sheng-Luen Chung 馬尚彬 Shang-Pin Ma 廖峻鋒 Chun-Feng Liao |
學位類別: |
碩士 Master |
系所名稱: |
電資學院 - 電機工程系 Department of Electrical Engineering |
論文出版年: | 2023 |
畢業學年度: | 111 |
語文別: | 中文 |
論文頁數: | 118 |
中文關鍵詞: | 智慧物聯網系統 、深度學習 、分布式系統 、邊緣模型 、邊緣運算 、局部學習 、基於模型的轉移學習 、多域混和相似性 、直接端對端 、多對多轉移學習 、跨筒倉 、拆分聯邦轉移學習 、異構模型融合 |
外文關鍵詞: | AIoT, Deep Learning, Distributed System, Edge Model, Edge Computing, Local Learning, Model-based Transfer Learning, Multi-domain Hybrid Similarity, Direct Edge-to-Edge, Many-to-Many Transfer Learning, Cross-silo, Split Federated Transfer Learning, Heterogeneous Model Fusion |
相關次數: | 點閱:381 下載:4 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
隨著科技產品製作技術逐漸成熟,使物聯網系統得以快速發展,其產生的龐大數據量,可使用深度神經網路處理,不過標記資料與訓練模型皆需要花費人力、時間,與運算資源,而透過轉移學習可以解決該問題。但傳統的系統皆需要中心化伺服器進行轉移學習的運算,雖已有研究提出分布式系統之直接端對端轉移學習,但其在同場域 (single-silo) 中忽視基於模型的轉移學習,導致仍有學習效率偏低且傳輸出成本偏高的缺點。因此,本研究透過「局部學習輔助之基於模型的轉移學習」轉移模型知識以提升目標任務的準確度,並利用「多域混和相似性學習模組」計算特徵圖相似度分數,以選擇到合適的轉移來源與目標。本研究進一步將上述兩模組結合於直接端對端多對多的轉移學習情境,其應用在一對多的轉移學習情況時,在平均減少48.57%的傳輸成本下,任務目標模型的辨識度平均提升42.32%。另外,在多對一的情況時,在減少48.57%的傳輸成本下,任務目標模型的辨識度提升63.41%。針對跨場域 (cross-silo) 的部分,既有研究直接端對端轉移學習亦忽略跨筒倉信息交換,導致無法減少跨店或跨筒倉學習的資源投入。透過本研究所提出之「輕量化拆分聯邦之基於模型的轉移學習」可以成功將異構資料的模型佈署於新店家當中,並且最多可以獲得84%準確度的初始模型。而兩具有異構資料及模型架構的筒倉亦可透過我們所提出的「跨層拆分聯邦模型聚合學習」進行模型融合,以學習到其他領域當中的多元知識,於實驗當中最高可以提升10.85%的模型準確度,藉此提升自身裝置的準確度與泛化能力。
With the gradual maturation of technology in the production of consumer goods, the Internet of Things (IoT) systems have experienced rapid development, resulting in a massive amount of data that can be processed using deep neural networks. However, annotating the data and training the models require significant manpower, time, and computational resources. Transfer learning can address this problem. Traditional systems rely on centralized servers for transfer learning. Although there have been studies proposing distributed systems for direct edge-to-edge transfer learning, they neglect model-based transfer learning within the same domain, leading to lower learning efficiency and higher transmission costs. Therefore, this study proposes "Local-learning-assisted Model-based Transfer Learning" to transfer model-level knowledge and improve the accuracy of the target task. It also utilizes a "Multi-Domain Hybrid Similarity Learning" to calculate feature map similarity scores, enabling the selection of appropriate transfer sources and targets. Furthermore, this study combines the two modules in a direct edge-to-edge many-to-many transfer learning scenario. When applied to a one-to-many transfer learning situation, it achieves an average improvement of 42.32% in the recognition accuracy of the task target model, while reducing transmission costs by an average of 48.57%. In the case of many-to-one transfer learning, the recognition accuracy of the task target model improves by 63.41% with a 48.57% reduction in transmission costs. Regarding cross-silo scenarios, existing research on direct edge-to-edge transfer learning also overlooks the exchange of information across different silos, resulting in an inability to reduce resource investment in learning across different stores or silos. Through the proposed "Lightweight Split Federated Model-based Transfer Learning," this study successfully deploys models for heterogeneous data in new silos, achieving an initial model accuracy of up to 84%. Additionally, two silos with heterogeneous data and model architectures can be merged using the "Cross-layer Split Federated Model Aggregation Learning" proposed in this study to learn diverse knowledge from other domains. In experiments, this approach achieves a maximum improvement of 10.85% in model accuracy, enhancing the accuracy and generalization capability of the models.
[1] M. Abdelhaq, "Internet of Things Fundamentals, Architectures, Challenges and Solutions: A Survey," IJCSNS, vol. 22, no. 1, p. 189, 2022.
[2] S. Shadroo, A. M. Rahmani, and A. Rezaee, "Survey on the application of deep learning in the Internet of Things," Telecommunication Systems, vol. 79, no. 4, pp. 601-627, 2022.
[3] M. Zamini and E. Kim, "A Survey on Computational Intelligence-based Transfer Learning," arXiv preprint arXiv:2206.10593, 2022.
[4] H. Hua, Y. Li, T. Wang, N. Dong, W. Li, and J. Cao, "Edge Computing with Artificial Intelligence: A Machine Learning Perspective," ACM Computing Surveys, vol. 55, no. 9, pp. 1-35, 2023.
[5] F. Massimi, P. Ferrara, and F. Benedetto, "Deep Learning Methods for Space Situational Awareness in Mega-Constellations Satellite-Based Internet of Things Networks," Sensors, vol. 23, no. 1, p. 124, 2023.
[6] C.-H. Lu and X.-Z. Lin, "Toward Direct Edge-to-Edge Transfer Learning for IoT-Enabled Edge Cameras," IEEE Internet of Things Journal, vol. 8, no. 6, pp. 4931-4943, 2020.
[7] C.-H. Lu and Y.-M. Zhou, "Direct Edge-to-Edge Many-to-Many Latent Feature Transfer Learning," IEEE Internet of Things Journal, 2021.
[8] 蔡永楨, "直接端對端具注意力之多重表示潛藏特徵轉移學習," 碩士, 電機工程系, 國立臺灣科技大學, 2021. [Online]. Available: https://etheses.lib.ntust.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=id=%22G0M10907318%22.&searchmode=basic
[9] Q. Yang, Y. Liu, T. Chen, and Y. Tong, "Federated machine learning: Concept and applications," ACM Transactions on Intelligent Systems and Technology (TIST), vol. 10, no. 2, pp. 1-19, 2019.
[10] M. Gholizade, H. Soltanizadeh, and M. Rahmanimanesh, "A Survey of Transfer Learning and Categories," Modeling and Simulation in Electrical and Electronics Engineering, vol. 1, no. 3, pp. 17-25, 2021.
[11] S.-L. Chung and C.-H. Lu, "Client/Server-based Concurrent Generic Equipment Emulator," in International IEEE/IAS Conference on Industrial Automation and Control: Emerging Technologies, 22-27 May 1995 1995, pp. 593-597.
[12] S. Saha and T. Ahmad, "Federated Transfer Learning: concept and applications," arXiv preprint arXiv:2010.15561, 2020.
[13] H. G. Abreha, M. Hayajneh, and M. A. Serhani, "Federated learning in edge computing: a systematic survey," Sensors, vol. 22, no. 2, p. 450, 2022.
[14] X. Zhong and H. Ban, "Pre-trained network-based transfer learning: A small-sample machine learning approach to nuclear power plant classification problem," Annals of Nuclear Energy, vol. 175, p. 109201, 2022.
[15] F. Zhuang et al., "A comprehensive survey on transfer learning," Proceedings of the IEEE, vol. 109, no. 1, pp. 43-76, 2020.
[16] T. Liu, S. Xie, J. Yu, L. Niu, and W. Sun, "Classification of thyroid nodules in ultrasound images using deep model based transfer learning and hybrid features," in 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2017: IEEE, pp. 919-923.
[17] C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, "Rethinking the inception architecture for computer vision," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 2818-2826.
[18] C. Szegedy, S. Ioffe, V. Vanhoucke, and A. A. Alemi, "Inception-v4, inception-resnet and the impact of residual connections on learning," in Thirty-first AAAI conference on artificial intelligence, 2017.
[19] Q. Peng, Z. Kong, L. Zhu, and T. Zhang, "A Multi-label Scene Categorization Model Based on Deep Convolutional Neural Network," Communications, Signal Processing, and Systems: Proceedings of the 2018 CSPS Volume III: Systems, vol. 517, p. 128, 2019.
[20] X. Zheng, L. Lin, S. Liang, B. Rao, and R. Zhan, "A Transfer Learning Method for Deep Networks with Small Sample Sizes," Journal of Physics: Conference Series, vol. 1631, no. 1, p. 012072, 2020/09/01 2020, doi: 10.1088/1742-6596/1631/1/012072.
[21] J. Liang, D. Hu, Y. Wang, R. He, and J. Feng, "Source data-absent unsupervised domain adaptation through hypothesis transfer and labeling transfer," IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021.
[22] Y. Ganin et al., "Domain-adversarial training of neural networks," The journal of machine learning research, vol. 17, no. 1, pp. 2096-2030, 2016.
[23] Z. Li, Y. Zhang, Y. Wei, Y. Wu, and Q. Yang, "End-to-End Adversarial Memory Network for Cross-domain Sentiment Classification," in IJCAI, 2017, pp. 2237-2243.
[24] Z. Ying et al., "TAI-SARNET: Deep transferred atrous-inception CNN for small samples SAR ATR," Sensors, vol. 20, no. 6, p. 1724, 2020.
[25] Y. Gao, "News Video Classification Model Based on ResNet-2 and Transfer Learning," Security and Communication Networks, vol. 2021, 2021.
[26] Q. Xia, W. Ye, Z. Tao, J. Wu, and Q. Li, "A survey of federated learning for edge computing: Research problems and solutions," High-Confidence Computing, vol. 1, no. 1, p. 100008, 2021.
[27] K. Kopparapu, E. Lin, J. G. Breslin, and B. Sudharsan, "TinyFedTL: Federated Transfer Learning on Ubiquitous Tiny IoT Devices," in 2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops), 2022: IEEE, pp. 79-81.
[28] H. Yi, T. Bie, and T. Yan, "Framework Construction of an Adversarial Federated Transfer Learning Classifier," arXiv preprint arXiv:2211.04734, 2022.
[29] Y. Sun, N. Chong, and O. Hideya, "Multi-Source Domain Adaptation Based on Federated Knowledge Alignment," arXiv preprint arXiv:2203.11635, 2022.
[30] T. Sha, X. Yu, Z. Shi, Y. Xue, S. Wang, and S. Hu, "Feature Map Transfer: Vertical Federated Learning for CNN Models," in International Conference on Data Mining and Big Data, 2021: Springer, pp. 37-44.
[31] T. Berghout, T. Bentrcia, M. A. Ferrag, and M. Benbouzid, "A Heterogeneous Federated Transfer Learning Approach with Extreme Aggregation and Speed," Mathematics, vol. 10, no. 19, p. 3528, 2022.
[32] U. Majeed, S. S. Hassan, and C. S. Hong, "Cross-Silo Model-Based secure federated transfer learning for Flow-Based traffic classification," in 2021 International Conference on Information Networking (ICOIN), 2021: IEEE, pp. 588-593.
[33] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[34] A. Nøkland and L. H. Eidnes, "Training neural networks with local error signals," in International conference on machine learning, 2019: PMLR, pp. 4839-4850.
[35] M. Laskin et al., "Parallel training of deep networks with local updates," arXiv preprint arXiv:2012.03837, 2020.
[36] J. Zhu, H. Zeng, S. Liao, Z. Lei, C. Cai, and L. Zheng, "Deep hybrid similarity learning for person re-identification," Ieee T Circ Syst Vid, vol. 28, no. 11, pp. 3183-3193, 2017.
[37] X. Tian, W. Dinghai, and W. Huaiguang, "Research on Application of Transfer Learning in Equipment Fault Diagnosis," in Journal of Physics: Conference Series, 2021, vol. 1986, no. 1: IOP Publishing, p. 012099.
[38] O. Gupta and R. Raskar, "Distributed learning of deep neural network over multiple agents," Journal of Network and Computer Applications, vol. 116, pp. 1-8, 2018.
[39] P. Vepakomma, O. Gupta, T. Swedish, and R. Raskar, "Split learning for health: Distributed deep learning without sharing raw patient data," arXiv preprint arXiv:1812.00564, 2018.
[40] S.-C. Huang, A. Pareek, S. Seyyedi, I. Banerjee, and M. P. Lungren, "Fusion of medical imaging and electronic health records using deep learning: a systematic review and implementation guidelines," NPJ digital medicine, vol. 3, no. 1, p. 136, 2020.
[41] S. P. Singh and M. Jaggi, "Model fusion via optimal transport," Advances in Neural Information Processing Systems, vol. 33, pp. 22045-22055, 2020.
[42] D.-J. Han, H. I. Bhatti, J. Lee, and J. Moon, "Accelerating Federated Split Learning via Local-Loss-Based Training," 2021.
[43] C. Thapa, P. C. M. Arachchige, S. Camtepe, and L. Sun, "Splitfed: When federated learning meets split learning," in Proceedings of the AAAI Conference on Artificial Intelligence, 2022, vol. 36, no. 8, pp. 8485-8493.
[44] X. Lu, Y. Liao, C. Liu, P. Lio, and P. Hui, "Heterogeneous Model Fusion Federated Learning Mechanism Based on Model Mapping," IEEE Internet of Things Journal, vol. 9, no. 8, pp. 6058-6068, 2021.
[45] I. Hussain, Q. He, and Z. Chen, "Automatic fruit recognition based on DCNN for commercial source trace system," Int. J. Comput. Sci. Appl, vol. 8, no. 2/3, pp. 01-14, 2018.
[46] Y.-F. Song, Z. Zhang, C. Shan, and L. Wang, "Constructing stronger and faster baselines for skeleton-based action recognition," IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 2, pp. 1474-1488, 2022.
[47] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, "Mobilenetv2: Inverted residuals and linear bottlenecks," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 4510-4520.
[48] M. Iman, H. R. Arabnia, and K. Rasheed, "A review of deep transfer learning and recent advancements," Technologies, vol. 11, no. 2, p. 40, 2023.
[49] B. Neyshabur, H. Sedghi, and C. Zhang, "What is being transferred in transfer learning?," Advances in neural information processing systems, vol. 33, pp. 512-523, 2020.