近年來，聯邦式學習(Federated Learning) 因為人工智慧(Artificial Intelligence) 日 益受到矚目，資料頓時成相當重要的資源，然而資料的儲存、傳輸、運算，對於大 多的應用服務提供公司(Application Service Provider) 來說，儼然成為了相當大的壓 力，由於，行動邊緣運算(Mobile Edge Computing) 的出現，以及日益普及，聯邦式 學習逐漸因其，去中心化的概念，以及保護隱私權特性，成為了當紅的熱門技術。 直觀地說，擁有更多的資料參與，有機會擁有更高的準確率，然而，越多的使 用者提供資料，要給付更高的費用(Incentive Payment)，並且，每個使用者裝置擁 有的運算及傳輸資源是異質的，會造成完成時間(Completion Time) 的掌握更加困 難。 因此，在本論文中，我們提出了一個最佳化問題，名為CoTENSOR， 同時考量 給付的費用以及完成時間。於此問題中，我們不僅討論了每個使用者裝置的不同費 用要求，異質的傳輸運算能力，以及為了確保產生足夠準確的結果，增加了資料量 總和的限制，並且設想做為中心伺服器的基地台擁有多個通道，能使多個使用者裝 置同時上傳。 接著，我們證明了CoTENSOR的難度，並且提出了一個近似演算法，名為 CoACT，此演算法同時考量了給付的費用以及完成時間，並且維持了準確率。 CoACT首先產生了所有可能的組合，依此產生出每個使用者裝置相對應的虛擬費用 (Pseudo Training Cost)，並利用動態規劃(Dynamic Programming) 的想法找尋最小的 虛擬費用，找出組成最小費用的使用者裝置集合，並且利用近似的技術降低其時間 複雜度。 實驗結果顯示，我們的演算法(CoACT) 不僅優於其他傳統方法(例如：隨機選取 裝置、選取最便宜的裝置、找尋增加時間最小的裝置)大約40%，也確保了準確率 的水準，並且在達到目標準確率上亦有相當好的表現，以及由於演算法的特性，可 以執行在平行的環境，我們也為此作了實驗。

Federated Learning (FL) has emerged in recent years to decentralize the model training with user devices. FL not only preserves data privacy but also ensures the model accuracy. To optimize the performance of FL, many critical issues, such as completion time minimization and device selection have been explored in the literature. However, none of previous work jointly considers the above two issues. In addition, few related studies consider multiple available channels. Therefore, in this thesis, we consider a more real and complex scenario, multiple channels are available to be allocated, which means it is available for multiple devices to upload at the same time. Moreover, channel allocation and uploading scheduling for the user devices in mobile networks also influence the completion time in each communication round of FL. To optimize all the issues above jointly, we first formulate a new optimization problem, named CoTENSOR, and prove its NP-hardness. Subsequently, a (3+e)-approximation algorithm, named CoACT, is proposed to find a near-optimal solution. Specifically, CoACT presents a new notion, pseudo training cost (detailed later), to evaluate the selection of each device so as to acquire an adequate set of devices to collaboratively train a global model. It also compacts the upload scheduling for those selected devices over multiple channels to jointly reduce total cost including the incentive payment (given to users in exchange for their computing and data) and completion time (of a global training round). We conduct an experimental evaluation with CIFAR10, which is a large-scale and public released image dataset and build and train a simple deep neural network model based on FL framework. Simulation results manifest CoACT can reduce the total cost by 40% compared with the traditional approaches and ensure test accuracy and train loss.

1 Introduction 1
2 Related Work 5
2.1 Mobile Edge Computing 5
2.2 Decentralized Learning Framework Evolution and
Analysis 6
2.3 Decentralized Training Efficiency Optimization 7
2.4 Summary 7
3 The Completion-Time-Aware Federated Learning via Resource-Constrained Devices Problem 8
3.1 Problem Formulation 8
3.2 Summary 14
4 Collaborative Heterogeneous Devices Selection and Scheduling Algorithm for Federated Learning 15
4.1 Collaborative Heterogeneous Devices Selection and
Scheduling Algorithm for Federated Learning 15
4.2 Candidate Devices (CD) Grouping 18
4.3 Candidate Devices (CD) Rounding 19
4.4 Candidate Devices (CD) Selection 20
4.5 Candidate Devices (CD) Scheduling 22
4.6 Final Solutions (FS) Selection 24
4.7 Time Complexity Analysis 24
4.8 Approximation Ratio Analysis 25
4.9 Summary 27
5 Performance Evaluation 28
5.1 Simulation Settings 28
5.2 Total Cost Trade-off and Obscure Total Cost 29
5.3 Heterogeneous Computing Resources 30
5.4 Data Quantity Requirement and Data Quantity per Device 32
5.5 The Effect of Number of Channels 32
5.6 Effect of Data Quantity Requirement on Accuracy and Loss 33
5.7 Running Time Speedup with Parallel Processing 35
5.8 Summary 36
6 Conclusions 37

[1] F. Zhou, L. Liu, K. Zhang, G. Trajcevski, J. Wu, and T. Zhong, “Deeplink: A deep learning approach for user identity linkage,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2018.
[2] T.-Y. Yang, C. G. Brinton, and C. Joe-Wong, “Predicting learner interactions in social learning networks,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2018.
[3] Y. Bao, Y. Peng, and C. Wu, “Deep learning-based job placement in distributed machine learning clusters,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2019.
[4] T. Ching, D. S. Himmelstein, B. K. Beaulieu-Jones, A. A. Kalinin, B. T. Do, G. P. Way, E. Ferrero, P.-M. Agapow, M. Zietz, M. M. Hoffman, et al., “Opportunities and obstacles for deep learning in biology and medicine,” Journal of The Royal Society
Interface, vol. 15, no. 141, p. 20 170 387, 2018.
[5] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of deep bidirectional transformers for language understanding,” in Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), vol. 1, Jun. 2019, pp. 4171–4186.
[6] A. M. Kaplan, “If you love something, let it go mobile: Mobile marketing and mobile social media 4x4,” Business horizons, vol. 55, pp. 129–139, 2012.
[7] A. Hard, K. Rao, R. Mathews, S. Ramaswamy, F. Beaufays, S. Augenstein, H. Eichner, C. Kiddon, and D. Ramage, “Federated learning for mobile keyboard prediction,” arXiv preprint arXiv:1811.03604, 2018.
[8] T. Yang, G. Andrew, H. Eichner, H. Sun, W. Li, N. Kong, D. Ramage, and F. Beaufays, “Applied federated learning: Improving google keyboard query suggestions,” arXiv preprint arXiv:1812.02903, 2018.
[9] S. Ramaswamy, R. Mathews, K. Rao, and F. Beaufays, “Federated learning for emoji prediction in a mobile keyboard,” arXiv preprint arXiv:1906.04329, 2019.
[10] (2019). Cisco visual networking index: Global mobile data traffic forecast update. White paper.
[11] H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communicationefficient learning of deep networks from decentralized data,” in Proceedings of International Conference on Artificial Intelligence and Statistics (AISTATS), 2016.
[12] T. Nishio and R. Yonetani, “Client selection for federated learning with heterogeneous resources in mobile edge,” in Proceedings of IEEE International Conference on Communications (ICC), 2019.
[13] Y. Zhang and M. Van der Schaar, “Reputation-based incentive protocols in crowdsourcing applications,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2012.
[14] S. Shi, X. Chu, and B. Li, “MG-WFBP: Efficient data communication for distributed synchronous SGD algorithms,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2019.
[15] Y. Mao, C. You, J. Zhang, K. Huang, and K. B. Letaief, “A survey on mobile edge computing: The communication perspective,” IEEE Communications Surveys & Tutorials, vol. 19, no. 4, pp. 2322–2358, 2017.
[16] S.-H. Hsu, C.-H. Lin, C.-Y. Wang, and W.-T. Chen, “Breaking bandwidth limitation for mission-critical iot using semisequential multiple relays,” IEEE Internet of Things Journal, vol. 5, no. 5, pp. 3316–3329, 2017.
[17] J. Koneˇcný, H. B. McMahan, F. X. Yu, P. Richtarik, A. T. Suresh, and D. Bacon, “Federated learning: Strategies for improving communication efficiency,” in Proceedings of Conference on Neural Information Processing Systems (NIPS) Workshop on Private Multi-Party Machine Learning, 2016.
[18] K. Yang, T. Jiang, Y. Shi, and Z. Ding, “Federated learning based on over-the-air computation,” in Proceedings of IEEE International Conference on Communications (ICC), 2019, pp. 1–6.
[19] ——, “Federated learning via over-the-air computation,” IEEE Transactions onWireless Communications, vol. 19, no. 3, pp. 2022–2035, 2020.
[20] M. Mohri, G. Sivek, and A. T. Suresh, “Agnostic federated learning,” in Proceedings of International Conference on Machine Learning (ICML), 2019.
[21] J. Wang and G. Joshi, “Cooperative SGD: A unified framework for the design and analysis of communication-efficient SGD algorithms,” arXiv preprint arXiv:1808.07576, 2018.
[22] Y. Gao, Y. Chen, and K. R. Liu, “On cost-effective incentive mechanisms in microtask crowdsourcing,” IEEE Transactions on Computational Intelligence and AI in Games, vol. 7, no. 1, pp. 3–15, 2014.
[23] S. A. Kravchenko and F. Werner, “Parallel machine scheduling problems with a single server,” Mathematical and Computer Modelling, vol. 26, no. 12, pp. 1–11, 1997.
[24] A. Krizhevsky, “Learning multiple layers of features from tiny images,” Citeseer, Tech. Rep., 2009.
[25] P. Rost, A. Banchs, I. Berberana, M. Breitbach, M. Doll, H. Droste, C. Mannweiler, M. A. Puente, K. Samdanis, and B. Sayadi, “Mobile network architecture evolution toward 5G,” IEEE Communications Magazine, vol. 54, no. 5, pp. 84–91, 2016.
[26] N. Abbas, Y. Zhang, A. Taherkordi, and T. Skeie, “Mobile edge computing: A survey,” IEEE Internet of Things Journal, vol. 5, no. 1, pp. 450–465, 2017.
[27] P. Schulz, A. Wolf, G. P. Fettweis, A. M. Waswa, D. M. Soleymani, A. Mitschele-Thiel, T. Dudda, M. Dod, M. Rehme, J. Voigt, et al., “Network architectures for demanding 5g performance requirements: Tailored toward specific needs of efficiency and flexibility,” IEEE Vehicular Technology Magazine, vol. 14, no. 2, pp. 33–43, 2019.
[28] X. Ran, H. Chen, X. Zhu, Z. Liu, and J. Chen, “DeepDecision: A mobile deep learning framework for edge video analytics,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2018.
[29] J. Xu, L. Chen, and P. Zhou, “Joint service caching and task offloading for mobile edge computing in dense networks,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2018.
[30] Y. He, J. Ren, G. Yu, and Y. Cai, “Joint computation offloading and resource allocation in D2D enabled mec networks,” in Proceedings of IEEE International Conference on Communications (ICC), 2019.
[31] Y. Xiao and M. Krunz, “Qoe and power efficiency tradeoff for fog computing networks with fog node cooperation,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2017.
[32] L. Tong and W. Gao, “Application-aware traffic scheduling for workload offloading in mobile clouds,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2016.
[33] L. Huang, S. Bi, and Y. J. Zhang, “Deep reinforcement learning for online computation offloading in wireless powered mobile-edge computing networks,” IEEE Trans. Mob. Comput., 2019.
[34] S. Wang et al., “Dynamic service placement for mobile micro-clouds with predicted future costs,” IEEE Trans. Parallel Distrib. Syst., vol. 28, pp. 1002–1016, 2016.
[35] L. Wang, L. Jiao, J. Li, and M. Mühlhäuser, “Online resource allocation for arbitrary user mobility in distributed edge clouds,” in Proceedings of IEEE International Conference on Distributed Computing Systems (ICDCS), 2017.
[36] L. Gu, D. Zeng,W. Li, S. Guo, A. Zomaya, and H. Jin, “Deep reinforcement learning based VNF management in geo-distributed edge computing,” in Proceedings of IEEE International Conference on Distributed Computing Systems (ICDCS), 2019.
[37] J. Park, S. Samarakoon, M. Bennis, and M. Debbah, “Wireless network intelligence at the edge,” Proceedings of the IEEE, vol. 107, no. 11, pp. 2204–2239, 2019.
[38] C. Zhang, H. Du, Q. Ye, C. Liu, and H. Yuan, “Dmra: A decentralized resource allocation scheme for multi-sp mobile edge computing,” in Proceedings of IEEE International Conference on Distributed Computing Systems (ICDCS), 2019.
[39] Y. Zhao, M. Li, L. Lai, N. Suda, D. Civin, and V. Chandra, “Federated learning with non-iid data,” arXiv preprint arXiv:1806.00582, 2018.
[40] C. Xie, K. Huang, P.-Y. Chen, and B. Li, “DBA: Distributed backdoor attacks against federated learning,” in Proceedings of International Conference on Learning Representations (ICLR), 2020.
[41] Z. Sun, P. Kairouz, A. T. Suresh, and H. B. McMahan, “Can you really backdoor federated learning?” arXiv preprint arXiv:1911.07963, 2019.
[42] K. Bonawitz, V. Ivanov, B. Kreuter, A. Marcedone, H. B. McMahan, S. Patel, D. Ramage, A. Segal, and K. Seth, “Practical secure aggregation for federated learning on user-held data,” in Proceedings of Conference on Neural Information Processing Systems (NIPS) Workshop on Private Multi-Party Machine Learning, 2016.
[43] K. Bonawitz, H. Eichner, W. Grieskamp, D. Huba, A. Ingerman, V. Ivanov, C. M. Kiddon, J. Koneˇcný, S. Mazzocchi, B. McMahan, T. V. Overveldt, D. Petrou, D. Ramage, and J. Roselander, “Towards federated learning at scale: System design,” in Proceedings of Conference on Systems and Machine Learning (SysML), 2019.
[44] V. Smith, C.-K. Chiang, M. Sanjabi, and A. S. Talwalkar, “Federated multi-task learning,” in Proceedings of Conference on Neural Information Processing Systems (NIPS), 2017.
[45] Q. Meng, W. Chen, Y. Wang, Z.-M. Ma, and T.-Y. Liu, “Convergence analysis of distributed stochastic gradient descent with shuffling,” Neurocomputing, vol. 337, pp. 46–57, 2019.
[46] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” in Proceedings of Conference on Neural Information Processing Systems (NIPS), 2014.
[47] R. Yonetani, T. Takahashi, A. Hashimoto, and Y. Ushiku, “Decentralized learning of generative adversarial networks from multi-client non-iid data,” arXiv preprint arXiv:1905.09684, 2019.
[48] R. Anil, G. Pereyra, A. T. Passos, R. Ormandi, G. Dahl, and G. Hinton, “Large scale distributed neural network training through online distillation,” in Proceedings of International Conference on Learning Representations (ICLR), 2018.
[49] E. Jeong, S. Oh, H. Kim, J. Park, M. Bennis, and S.-L. Kim, “Communicationefficient on-device machine learning: Federated distillation and augmentation under non-iid private data,” arXiv preprint arXiv:1811.11479, 2018.
[50] S.Wang, T. Tuor, T. Salonidis, K. K. Leung, C. Makaya, T. He, and K. Chan, “When edge meets learning: Adaptive control for resource constrained distributed machine learning,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2018.
[51] ——, “Adaptive federated learning in resource constrained edge computing systems,” IEEE Journal on Selected Areas in Communications, vol. 37, no. 6, pp. 1205–1221, 2019.
[52] N. H. Tran, W. Bao, A. Zomaya, N. M. NH, and C. S. Hong, “Federated learning over wireless networks: Optimization model design and analysis,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), 2019.
[53] L. P. Kaelbling, M. L. Littman, and A. W. Moore, “Reinforcement learning: A survey,” Journal of artificial intelligence research, vol. 4, pp. 237–285, 1996.
[54] X.Wang, Y. Han, C.Wang, Q. Zhao, X. Chen, and M. Chen, “In-edge AI: Intelligentizing mobile edge computing, caching and communication by federated learning,” IEEE Network, vol. 33, no. 5, pp. 156–165, 2019.
[55] D. P.Williamson and D. B. Shmoys, The Design of Approximation Algorithms. Cambridge University Press, 2011.
[56] M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, M. Isard, et al., “Tensorflow: A system for large-scale machine learning,” in Proceedings of USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2016.
[57] F. Chollet et al., Keras, https://keras.io, 2015.
[58] (2020). Google cloud storage pricing, [Online]. Available: https://cloud.google.com/storage/pricing#operations-pricing.
[59] (2020). Coinmarketcap, [Online]. Available: https://coinmarketcap.com/currencies/datum/.