A Beta neuron in CMOS subthreshold mode

Beta Basis Function Neural Networks (BBFNN) are powerful systems for learning and universal approximation characteristics. In this paper, we present a hardware implementation of the Beta neuron using the CMOS subthreshold-mode. We describe a low power low voltage analog Beta neuron circuit. Three main modules are used to realize the Beta function: a logarithmic current to voltage converter, a multiplier and an exponential voltage to current converter. Simulation results prove the validity of our neural hardware implementation. The control parameters of the Beta function are independent and are made easily by current sources. This analog implementation can be used easily to realize analog BBFNN.

[1]  Radu Dogaru,et al.  A modified RBF neural network for efficient current-mode VLSI implementation , 1996, Proceedings of Fifth International Conference on Microelectronics for Neural Networks.

[2]  Eric A. Vittoz,et al.  Micropower Techniques , 1994 .

[3]  L X Wang,et al.  Fuzzy basis functions, universal approximation, and orthogonal least-squares learning , 1992, IEEE Trans. Neural Networks.

[4]  Michael D. Godfrey.  CMOS device modeling for subthreshold circuits , 1992 .

[5]  J. Ramirez-Angulo,et al.  Architectures for analog VLSI implementation of neural networks for solving linear equations with inequality constraints , 1992, [Proceedings] 1992 IEEE International Symposium on Circuits and Systems.

[6]  David Zhang,et al.  VLSI compressor design with applications to digital neural networks , 1997, IEEE Trans. Very Large Scale Integr. Syst..