Configurable Stochastic Circuits for Parallel Problem Solving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional approaches to solving stochastic problems electronically are inefficient due to their reliance on deterministic computers and high-precision floating point arithmetic, which require excessive time and processing resources, and often unnecessary precision given the inherent randomness and uncertainty in stochastic processes.
Innovation Solution
The use of configurable circuits that can be configured to generate samples from probability distributions, allowing for the division of stochastic problems into smaller fragments that can be solved in parallel using less complex circuitry and lower precision arithmetic, such as stochastic tiles with memory to store information about random variables and control functionality to reconfigure for different fragments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deterministic computers with high-precision floating point arithmetic are used to solve stochastic problems, then measurement precision is improved, but productivity deteriorates due to excessive time and processing resources required
Solution Approach 1:
The patent replaces deterministic mechanical computation systems with stochastic computational systems. Instead of using traditional deterministic computers that perform sequential floating-point arithmetic operations, the invention employs stochastic circuits that naturally perform parallel probabilistic computations, substituting the mechanical computational approach with a fundamentally different computational paradigm that is inherently suited for stochastic problems
Solution Approach 2:
The patent divides complex stochastic problems into smaller, independent stochastic fragments that can be processed in parallel. Each fragment represents a subset of the overall probability distribution that can be solved independently, allowing multiple fragments to be computed simultaneously across multiple stochastic circuits, thereby increasing overall processing throughput while maintaining accuracy
2Measurement precision
If deterministic computers with high-precision floating point arithmetic are used to solve stochastic problems, then measurement precision is improved, but device complexity worsens due to excessive processing resources required
Solution Approach 1:
The patent replaces complex deterministic mechanical computation systems with simpler stochastic circuits. The stochastic circuits utilize natural probabilistic behavior of electronic components rather than complex sequential logic and floating-point arithmetic units, significantly reducing device complexity while maintaining computational capability for stochastic problems
Solution Approach 2:
The patent employs multiple simple, low-cost stochastic circuit elements that can be easily fabricated and replaced. Instead of using a few complex high-precision deterministic processors, the invention uses many simpler stochastic units whose individual complexity is low, though they work in parallel to achieve the required computational power and precision
3Productivity
If configurable circuits divide stochastic problems into fragments for parallel processing, then productivity is improved, but device complexity worsens due to reconfiguration requirements
Solution Approach 1:
The patent employs dynamically reconfigurable stochastic circuits that can adapt their configuration based on the specific stochastic fragment being processed. The circuits can be reconfigured to handle different problem types and fragment structures, allowing a single hardware platform to efficiently process diverse stochastic problems by dynamically adjusting its internal architecture rather than requiring separate dedicated circuits for each problem type
Data Source
AI summary
Techniques described herein may be used to solve a stochastic problem by dividing the stochastic problem into multiple fragments. In some cases, each fragment may be related to a random variable that forms a part of the problem, such that each fragment may produce samples from a probability distribution for that variable. Each fragment of the stochastic problem may then be assigned to a configurable circuit to solve the stochastic fragment. Configurable circuits may be implemented using any suitable combination of hardware and/or software, including using stochastic circuitry. In some embodiments, stochastic circuitry may include a stochastic tile and/or a stochastic memory.