本論文將深度學習(DL)技術結合於毫米波多輸入多輸出系統中的混合預編碼與合併器設計，透過適當的訓練過程可以提升預測的準確性。我們的訓練資料包含毫米波通道以及射頻預編碼和射頻合併器矩陣。藉由大量的訓練資料與Adam演算法根據當前參數所得到的均方誤差(MSEs)調整參數，以此提升神經網路預測的準確性。訓練結束後將測試集通道餵入訓練好的神經網路(NN)並得到預測結果也就是射頻預編碼與射頻合併器的相位，在給定一個射頻預編碼的情況下我們可以利用最小平方法求解獲得基頻預編碼，相似的方法應用在射頻合併器得到基頻合併器矩陣。在不同數據流數量的情況下，根據深度神經網路的結果計算頻譜效率並從頻譜效率中可以說明我們設計的方法是有競爭力的。

In this thesis, we apply deep learning (DL) techniques to solve hybrid precoding and combining design problems in millimeter wave (mmWave) multiple-input multiple-output (MIMO) systems. To increase the accuracy of prediction is achieved through an appropriately training process. Our training data set includes mmWave channels, and solutions of RF analog precoders along with combiners. By utilizing a lot of training data and the Adam optimizer to adjust the parameters based on the mean square errors (MSEs), we can improve the accuracy of prediction. After training process, we feed testing data set into neural network (NN) and obtain the solutions of the phases of RF analog precoder and combiner. Given the RF analog precoder, we can acquire baseband precoder by using least square solutions; and the similar methodology is applied to the design of the RF analog combiner along with acquiring the baseband combiner. In the cases of different numbers of data streams, we calculate the spectral efficiency based on the outputs of DNN in the simulations results, and it can be observed that our method is competitive to the existing schemes.

[1] X. Zhang, A. F. Molisch and S.-Y. Kung, "Variable-phase-shift-based RF-baseband codesign for MIMO antenna selection," in IEEE Transactions on Signal Processing, vol. 53, no. 11, pp. 4091-4103, Nov. 2005.
[2] O. E. Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi and R. W. Heath, "Spatially Sparse Precoding in Millimeter Wave MIMO Systems," in IEEE Transactions on Wireless Communications, vol. 13, no. 3, pp. 1499-1513, Mar. 2014.
[3] W. Ni and X. Dong, "Hybrid Block Diagonalization for Massive Multiuser MIMO Systems," in IEEE Transactions on Communications, vol. 64, no. 1, pp. 201-211, Jan. 2016.
[4] N. Li, Z. Wei, H. Yang, X. Zhang and D. Yang, "Hybrid Precoding for mmWave Massive MIMO Systems With Partially Connected Structure," in IEEE Access, vol. 5, pp. 15142-15151, 2017.
[5] J. Zhang, Y. Huang, T. Yu, J. Wang and M. Xiao, "Hybrid Precoding for Multi-Subarray Millimeter-Wave Communication Systems," in IEEE Wireless Communications Letters, vol. 7, no. 3, pp. 440-443, Jun. 2018.
[6] M. Abadi et al., "TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems," arXiv: 1603.04467, 2016. [Online]. Availabel: https://arxiv.org/abs/1603.04467.
[7] A. Paszke et al., "Automatic differentiation in Pytorch," in Proc. Neural Information Processing Systems, 2017, pp. 1-4.
[8] C. Wen, W. Shih and S. Jin, "Deep Learning for Massive MIMO CSI Feedback," in IEEE Wireless Communications Letters, vol. 7, no. 5, pp. 748-751, Oct. 2018.
[9] K. Kim, J. Lee and J. Choi, "Deep Learning Based Pilot Allocation Scheme (DL-PAS) for 5G Massive MIMO System," in IEEE Communications Letters, vol. 22, no. 4, pp. 828-831, Apr. 2018.
[10] H. Ye, G. Y. Li and B. Juang, "Power of Deep Learning for Channel Estimation and Signal Detection in OFDM Systems," in IEEE Wireless Communications Letters, vol. 7, no. 1, pp. 114-117, Feb. 2018.
[11] M. Kim, W. Lee and D. Cho, "A Novel PAPR Reduction Scheme for OFDM System Based on Deep Learning," in IEEE Communications Letters, vol. 22, no. 3, pp. 510-513, Mar. 2018.
[12] A. Felix, S. Cammerer, S. Dörner, J. Hoydis and S. Ten Brink, "OFDM-Autoencoder for End-to-End Learning of Communications Systems," in Proc. IEEE International Workshop on Signal Processing Advances in Wireless Communications, Kalamata, 2018, pp. 1-5.
[13] W. Lee, "Resource Allocation for Multi-Channel Underlay Cognitive Radio Network Based on Deep Neural Network," in IEEE Communications Letters, vol. 22, no. 9, pp. 1942-1945, Sept. 2018.
[14] L. Fang and L. Wu, "Deep learning detection method for signal demodulation in short range multipath channel," in Proc. IEEE International Conference on Opto-Electronic Information Processing, Singapore, 2017, pp. 16-20.
[15] B. Kim, J. Kim, H. Chae, D. Yoon and J. W. Choi, "Deep neural network-based automatic modulation classification technique," in Proc. International Conference on Information and Communication Technology Convergence, Jeju, 2016, pp. 579-582.
[16] T. Gruber, S. Cammerer, J. Hoydis and S. t. Brink, "On deep learning-based channel decoding," in Proc. Conference on Information Sciences and Systems, Baltimore, MD, 2017, pp. 1-6.
[17] K. He, X. Zhang, S. Ren and J. Sun, "Delving Deep into Rectifiers: Surpassing Human-level Performance on ImageNet Classification", in Proc. IEEE International Conference on Computer Vision, Santiago, 2015 pp. 1026-1034.
[18] D. P. Kingma and J. Ba, "ADAM: A Method for Stochastic Optimization", arXiv: 1412.6980, 2014. [Online]. Available: https://arxiv.org/abs/1412.6980.
[19] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever and R. Salakhutdinov, "Dropout: A Simple Way to Prevent Neural Networks from Overfitting", in Journal of Machine Learning Research, vol. 15, no. 1, pp.1929-1958, Jan. 2014.
[20] A. L. Maas, A. Y. Hannun and A. Y. Ng, "Rectifier Nonlinearities Improve Neural Network Acoustic Models", in Proc. International Conference on Machine Learning, 2013.