Survey of Stochastic Computation on Factor Graphs

Stochastic computation is a new alternative approach for iterative computation on factor graphs. In this approach, the information is represented by the statistics of the bit stream which results in simple high-speed hardware implementation of graph-based algorithms. Despite the first purpose of its invention (i.e., low-precision digital circuits), the stochastic representation has recently been shown to be able to provide near-optimal decoding performance for practical Low-Density ParityCheck (LDPC) codes, with respect to Sum-Product Algorithm (SPA). This paper provides a survey of stochastic methods for graph-based iterative decoding, the state-of-the-art and, their possible new applications.

• Type: Conference
• Year: 2007
• Where: ISMVL
• Authors: Saeed Sharifi Tehrani, Shie Mannor, Warren J. Gross
• Keywords: Artificial Intelligence, ISMVL 2007, Iterative Computation, Near-optimal Decoding Performance, Stochastic Computation