Low Cost LSTM Implementation based on Stochastic Computing for Channel State Information Prediction

This paper proposes a low-cost hardware implementation of Long Short-Term Memory (LSTM) Neural Network for channel state information (CSI) prediction. We first employ LSTM algorithm to the predication of channel state information. To reduce the hardware cost, the stochastic computing is employed to design the LSTM accelerator. The complex arithmetic operations are converted to simple logic gates. For instance, the multiplication is performed by an AND gate and tanh function is implemented by a finite state machine. According to the implementation report, the proposed stochastic LSTM reduce the hardware cost about 70%, which provides a promising technology for the future communication system design.

[1]  Shi Jin,et al.  Channel Estimation for Massive MIMO Using Gaussian-Mixture Bayesian Learning , 2015, IEEE Transactions on Wireless Communications.

[2]  Jeffrey H. Reed,et al.  A new approach to signal classification using spectral correlation and neural networks , 2005, First IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks, 2005. DySPAN 2005..

[3]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[4]  João Canas Ferreira,et al.  An FPGA implementation of a long short-term memory neural network , 2016, 2016 International Conference on ReConFigurable Computing and FPGAs (ReConFig).

[5]  Jianhao Hu,et al.  Hardware Efficient Massive MIMO Detector Based on the Monte Carlo Tree Search Method , 2017, IEEE Journal on Emerging and Selected Topics in Circuits and Systems.

[6]  Syuan-Yi Chen,et al.  FPGA-Based Computed Force Control System Using Elman Neural Network for Linear Ultrasonic Motor , 2009, IEEE Transactions on Industrial Electronics.

[7]  Jianhao Hu,et al.  A novel FIR filter based on stochastic logic , 2013, 2013 IEEE International Symposium on Circuits and Systems (ISCAS2013).

[8]  Tao Jiang,et al.  Deep learning for wireless physical layer: Opportunities and challenges , 2017, China Communications.

[9]  Howard C. Card,et al.  Stochastic Neural Computation I: Computational Elements , 2001, IEEE Trans. Computers.

[10]  Jianhao Hu,et al.  Stochastic Iterative MIMO Detection System: Algorithm and Hardware Design , 2015, IEEE Transactions on Circuits and Systems I: Regular Papers.

[11]  Jianhao Hu,et al.  Stochastic MIMO Detector Based on the Markov Chain Monte Carlo Algorithm , 2014, IEEE Transactions on Signal Processing.

[12]  Jürgen Schmidhuber,et al.  Learning to Forget: Continual Prediction with LSTM , 2000, Neural Computation.

[13]  Jianhao Hu,et al.  High Throughput Stochastic Log-MAP Turbo-Decoder Based on Low Bits Computation , 2013, IEEE Signal Processing Letters.

[14]  Jianhao Hu,et al.  Hardware and Energy-Efficient Stochastic LU Decomposition Scheme for MIMO Receivers , 2016, IEEE Transactions on Very Large Scale Integration (VLSI) Systems.

[15]  Jianhao Hu,et al.  A low complexity SCMA detector based on stochastic computation , 2017, 2017 IEEE 60th International Midwest Symposium on Circuits and Systems (MWSCAS).