Reversible stochastic computing

Stochastic computing is reemerging as one of the alternative technique of conventional binary computing for error tolerant, low-power, and low resource applications. In the recent past, several works have been reported for the implementation of stochastic circuits. All these works have been carried out using irreversible gates or Boolean functions as building blocks, which will not be supported in the near future for the power efficient realizations due to Landauer principle. Therefore, it is timely to consider the reversible stochastic computing design. Till date, as per the author's best knowledge, no work has been reported on the reversible stochastic computing. Besides, no work has been reported on the study of feasibility of existing reversible gates for stochastic computing. In this paper, reversible gates have been examined for their stochastic operations in unipolar, bipolar, and inverse bipolar formats. In addition, some novel universal reversible gates are proposed, which are well suited for reversible stochastic computing using these formats. Finally, the implementation of reversible two-input linear neuron using the proposed reversible gates has been demonstrated.