Weighted-Adder-Based Polynomial Computation Using Correlated Unipolar Stochastic Bitstreams

Abstract

Stochastic computing (SC) is an unconventional computing paradigm, which processes numerical values through stochastic bitstreams. This representation can be interpreted as probabilities and makes stochastic circuits implemented with simple logic gates. This brief proposes a stochastic weighted-adder based approach for computing polynomials by using correlation logic in unipolar format. This approach can significantly reduce the use of RNGs to lower hardware cost and can handily compute polynomials with arbitrary coefficients. Experimental results show that the proposed designs outperform previous counterparts in computing accuracy and hardware cost. For example, the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$5^{th}$ </tex-math></inline-formula> -order Maclaurin polynomial of the sigmoid( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${x}$ </tex-math></inline-formula> ) that contains positive and negative coefficients can have at least an 87% improvement in mean squared error and a 21% reduction in area.