Semiconductor Device Stochastic Logic Circuit

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing semiconductor devices that calculate interaction models, such as the Ising model, face challenges in achieving high accuracy while maintaining low production costs and efficient energy usage, due to the complexity of implementing the thermal bath method's transition probability calculations.

Innovation Solution

A semiconductor device configuration that includes a memory, a reading unit, a majority circuit, and a write circuit, with a random number signal line injected before the majority circuit to perform stochastic state transitions based on the thermal bath method, reducing circuit area and cost while maintaining accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the thermal bath method is implemented using adder and multiplier to calculate transition probability, then calculation accuracy is improved, but circuit area increases and manufacturing cost increases

Engineering Contradiction:
Improvecalculation accuracyVSAvoidcircuit area
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts and eliminates the complex non-linear calculation components (adder and multiplier) from the thermal bath method implementation. By removing these expensive circuit elements and replacing them with simpler logic circuits, the invention achieves the same stochastic state transition functionality with reduced circuit area and lower manufacturing cost while maintaining calculation accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent replaces expensive, complex circuit components (adder and multiplier) with cheaper, simpler logic circuit components. This substitution uses less expensive hardware elements that can be manufactured more easily, reducing both circuit area and manufacturing cost while achieving the required computational functionality for the SA method.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

2Measurement precision

If the thermal bath method is implemented using adder and multiplier to calculate transition probability, then calculation accuracy is improved, but energy consumption increases

Engineering Contradiction:
Improvecalculation accuracyVSAvoidenergy consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent removes the energy-intensive adder and multiplier circuits from the implementation. By extracting these high-power components and replacing them with simpler logic circuits, the invention significantly reduces energy consumption while maintaining the accuracy required for proper stochastic state transition in the SA method.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent substitutes expensive, high-energy-consuming circuit components with cheaper, low-energy-consuming logic circuit components. This replacement reduces the energy footprint of the system while preserving the functional requirements for accurate transition probability calculation in the simulated annealing process.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

3Device complexity

If a stochastic algorithm omitting transition probability calculation is used, then circuit area is reduced and manufacturing cost is reduced, but solution accuracy is lowered

Engineering Contradiction:
Improvecircuit areaVSAvoidsolution accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent implements a partial version of the full thermal bath method by using logic circuits to achieve the essential stochastic state transition functionality without completing the full non-linear calculation sequence. This partial implementation maintains sufficient accuracy for the SA method while avoiding the complexity of complete transition probability calculation through adder and multiplier circuits.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent introduces logic circuits as an intermediary mechanism that bridges the gap between simplified algorithms and accurate transition probability calculation. These logic circuits serve as a mediator that provides the necessary stochastic behavior and accuracy without requiring the full complexity of traditional non-linear calculation methods.

Inventive Principle:
Principle #24Intermediary (Mediator)

4Measurement precision

If more transitions are performed to obtain highly accurate solution, then solution accuracy is improved, but calculation time increases

Engineering Contradiction:
Improvesolution accuracyVSAvoidcalculation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the mechanical iteration approach (performing many sequential transitions) with a more efficient logic circuit-based system. By substituting the mechanical repeated transition process with optimized logic circuit operations, the invention achieves high solution accuracy with fewer iterations, thereby reducing calculation time while maintaining precision.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3343468B1Semiconductor device and information processing device
Publication Date: 2019.12.25 HITACHI LTD
  • EP3343468B1 patent drawingFigure 1
  • EP3343468B1 patent drawingFigure 2
  • EP3343468B1 patent drawingFigure 3

AI summary

Provided are a semiconductor device and an information processing device that can be manufactured easily at low cost and can calculate an arbitrary interaction model such as an Ising model. A semiconductor device that performs a non-linear operation includes a memory, a reading unit that reads data from the memory, a majority circuit that inputs a result of a predetermined operation on the data read by the reading unit, and a write circuit that receives an output of the majority circuit, a value of a predetermined signal is stochastically inverted at a preceding stage of the majority circuit.