Stochastic computing (SC) hinges on the generation and use of stochastic bitstreams-streams of randomly generated 1s and 0s, with the probabilities of p and 1-p, respectively. We consider approaches for stochastic bitstream generation, considering randomness, circuit area/performance/cost, and the impact of the various approaches on SC accuracy. We first review the widely used Linear-Feedback Shift Register (LFSR)-based approach and variants. Alternative low-discrepancy sequences are then discussed, followed by techniques that leverage post-CMOS technologies and metastability of devices as sources of randomness. We conclude with a discussion on correlations between bitstreams, and how (1) correlations can be reduced/eliminated, and (2) correlations may actually be leveraged to positive effect in certain circumstances.
CITATION STYLE
Hsiao, H., Anderson, J., & Hara-Azumi, Y. (2019). Generating stochastic bitstreams. In Stochastic Computing: Techniques and Applications (pp. 137–152). Springer International Publishing. https://doi.org/10.1007/978-3-030-03730-7_7
Mendeley helps you to discover research relevant for your work.