Low-Discrepancy Sequence Generator for Accurate Stochastic Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stochastic computing systems face challenges in achieving high accuracy and energy efficiency due to the high hardware cost and complexity of using pseudo-random sequences like Sobol and Halton sequences for generating stochastic bit-streams, which affect the latency and energy consumption of computations.
Innovation Solution
A novel Powers-of-2 Low-Discrepancy Sequence Generator (P2LSG) is introduced, utilizing a lightweight and energy-efficient design based on Powers-of-2 Van der Corput sequences, implemented with a log2(N)-bit counter and hardwiring scheme, capable of generating multiple sequences in parallel to enhance accuracy and reduce hardware costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pseudo-random sequences like Sobol and Halton sequences are used for generating stochastic bit-streams, then accuracy of stochastic computing is improved, but hardware cost and complexity increase
Solution Approach 1:
The patent segments the sequence generation into multiple independent Linear Feedback Shift Registers (LFSRs), each generating a separate bit-stream. By using multiple simple LFSRs instead of a single complex pseudo-random sequence generator, the system achieves the statistical properties of sophisticated sequences while using simpler, more hardware-efficient components. The segmentation allows parallel generation of uncorrelated sequences.
Solution Approach 2:
The patent creates multiple copies of simple LFSR circuits to generate multiple stochastic bit-streams simultaneously. Instead of implementing one complex Sobol or Halton sequence generator, the system uses multiple copies of a basic LFSR unit, each configured with different feedback polynomials and initial states, to produce the required number of uncorrelated sequences for accurate stochastic computing operations.
2Measurement precision
If pseudo-random sequences like Sobol and Halton sequences are used for generating stochastic bit-streams, then accuracy of stochastic computing is improved, but energy consumption increases
Solution Approach 1:
The patent segments the sequence generation into multiple independent Linear Feedback Shift Registers (LFSRs), each generating a separate bit-stream. By using multiple simple LFSRs instead of a single complex pseudo-random sequence generator, the system achieves the statistical properties of sophisticated sequences while using simpler, more hardware-efficient components. The segmentation allows parallel generation of uncorrelated sequences.
Solution Approach 2:
The patent creates multiple copies of simple LFSR circuits to generate multiple stochastic bit-streams simultaneously. Instead of implementing one complex Sobol or Halton sequence generator, the system uses multiple copies of a basic LFSR unit, each configured with different feedback polynomials and initial states, to produce the required number of uncorrelated sequences for accurate stochastic computing operations.
3Measurement precision
If pseudo-random sequences like Sobol and Halton sequences are used for generating stochastic bit-streams, then accuracy of stochastic computing is improved, but latency increases
Solution Approach 1:
The patent segments the sequence generation into multiple independent Linear Feedback Shift Registers (LFSRs), each generating a separate bit-stream. By using multiple simple LFSRs instead of a single complex pseudo-random sequence generator, the system achieves the statistical properties of sophisticated sequences while using simpler, more hardware-efficient components. The segmentation allows parallel generation of uncorrelated sequences.
Solution Approach 2:
The patent pre-configures multiple LFSRs with different feedback polynomials and initial states during the design phase, so that they are ready to generate uncorrelated sequences immediately upon activation. This preliminary setup eliminates the need for complex runtime calculations or state transitions required by Sobol and Halton generators, reducing initialization latency and enabling immediate parallel sequence generation.
Data Source
AI summary
A random number generator for accurate and energy-efficient stochastic computing, providing a low-cost and energy-efficient Low-discrepancy Sequence Generator derived from Powers-of-2 Van der Corput (VDC) sequences.