Hardware Implementation of Tanh Function Based on Second-Order Approximation and Error Compensation

ZHANG Bowen; CHEN Gang; CHEN Xu; LU Huaxiang

doi:10.13911/j.cnki.1004-3365.210031

[1] RAMANAIAH K V, SRIDHAR S. Hardware implementation of artificial neural networks ［J］. I-Manager’s J Embedded Syst, 2014, 3(4): 31-34.

[2] SHARMA S, ATHAIYA A. Activation functions in neural networks ［J］. Int J Engineer Appl Sci Tech, 2020, 4(12): 310-316.

[3] LEBOEUF K, NAMIN A H, MUSCEDERE R, et al. High speed VLSI implementation of the hyperbolic tangent sigmoid function ［C］ // 3rd Int Conf Convergence Hybrid Inform Tech. Busan, South Korea. 2008: 1070-1073.

[5] LIN C W, WANG J S. A digital circuit design of hyperbolic tangent sigmoid function for neural networks ［C］ // IEEE ISCAS. Seattle, WA, USA. 2008: 856-859.

[7] HAJDUK Z. Hardware implementation of hyperbolic tangent and sigmoid activation functions ［J］. Bulletin Polish Academy Sci-Tech Sci, 2018, 66(5): 563-577.

[9] SARANYA S, ELANGO B. Implementation of PWL and LUT based approximation for hyperbolic tangent activation function in VLSI ［C］ // Int Conf Commun Signal Process. Melmaruvathur, India. 2014: 1778-1782.

[10] NAMIN A H, LEBOEUF K, MUSCEDERE R, et al. Efficient hardware implementation of the hyperbolic tangent sigmoid function ［C］ // IEEE Int Symp Circ Syst. Taiwan, China. 2009: 2117-2120.

[11] LI Z, WANG L, GUO S, et al. Laius: an 8-bit fixed-point CNN hardware inference engine ［C］ // IEEE ISPA/IUCC. Guangzhou, China. 2017: 143-150.

[12] ZAMANLOOY B, MIRHASSANI M. Efficient VLSI implementation of neural networks with hyperbolic tangent activation function ［J］. IEEE Trans VLSI Syst, 2014, 22(1): 39-48.