Pseudo-Random Bit-Stream Down-Sampling for Stochastic Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deterministic approaches to stochastic computing suffer from poor progressive precision and high energy consumption due to their reliance on unary streams, which require longer operation times and higher energy usage to achieve accurate results, even when slight inaccuracies are acceptable.
Innovation Solution
The development of high-quality down-sampling techniques for deterministic stochastic computing that generate pseudo-random but accurate bit streams, modifying the structure of stream generators to produce completely accurate results in a shorter time and with lower energy consumption, by introducing pseudo-randomization and truncating output streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If deterministic approaches use unary streams to ensure completely accurate results, then manufacturing precision is improved, but productivity deteriorates due to longer operation times
Solution Approach 1:
The patent changes the fundamental parameter of bit stream structure from unary format (single 1 followed by 0s) to pseudo-random format (multiple 1s distributed throughout the stream). This parameter change allows the system to maintain deterministic accuracy while reducing the number of clock cycles needed, thereby improving productivity without sacrificing computation precision
Solution Approach 2:
Instead of using traditional unary streams that guarantee accuracy but slow down processing, the patent inverts the approach by using pseudo-random streams that were previously associated with stochastic (inaccurate) computing. By applying deterministic logic to pseudo-random streams, the patent achieves both speed and accuracy, effectively inverting the conventional wisdom that accuracy requires unary format
2Manufacturing precision
If deterministic approaches use unary streams to achieve accurate results, then manufacturing precision is improved, but use of energy worsens due to higher energy consumption
Solution Approach 1:
By changing the bit stream parameter from unary to pseudo-random format, the patent reduces the number of clock cycles required for computation. Since energy consumption in digital circuits is directly proportional to the number of clock cycles, this parameter change simultaneously improves energy efficiency while maintaining computation accuracy through deterministic logic operations
3Productivity
If conventional random stream-based stochastic computing is used to reduce processing time, then productivity is improved, but manufacturing precision deteriorates due to random fluctuation and correlation problems
Solution Approach 1:
The patent introduces deterministic logic operations as an intermediary between pseudo-random bit streams and the final computation result. This intermediary layer of deterministic logic eliminates the random fluctuation and correlation problems inherent in conventional stochastic computing, thereby achieving both fast processing (inherited from pseudo-random streams) and high accuracy (guaranteed by deterministic logic)
Solution Approach 2:
The patent extracts and removes the harmful random fluctuation and correlation effects from the computation process by applying deterministic logic. By taking out these problematic stochastic elements while retaining the speed benefits of pseudo-random streams, the patent achieves both high productivity and high precision simultaneously
Data Source
AI summary
This disclosure describes techniques for processing data bits using pseudo-random deterministic bit-streams. In some examples, a device includes a pseudo-random bit-stream generator configured to generate bit combinations encoding first and second numerical values based on a proportion of the data bits in the sequence that are high relative to the total data bits in the sequence. The device also includes a stochastic computational unit configured to perform a computational operation on the bit combinations and produce an output bit-stream having a set of data bits indicating a result of the computational operation, wherein the data bits of the output bit-stream represent the result based on a probability that any data bit in the set of data bits of the output bit-stream is high.