Mixed-Precision Memcomputing System for Computational Precision

Solution Overview

Problem

Conventional memcomputing techniques face challenges in achieving sufficient computational precision due to device variability and stochasticity, which limits their effectiveness for many computational tasks.

Innovation Solution

A mixed-precision memcomputing system is implemented, combining low-precision memcomputing hardware with highly precise digital combinational circuitry to iteratively increase computational precision, using resistive memory elements and digital circuitry for analog computations, allowing for improved precision in calculations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If conventional memcomputing techniques are used, then energy efficiency and speed are improved, but computational precision deteriorates due to device variability and stochasticity

Engineering Contradiction:

Improveenergy efficiencyVSAvoidcomputational precision

Core Design Contradiction:

Use of energy by moving objectVSMeasurement precision

Solution Approach 1:

The computational system is segmented into two distinct components: a low-precision memcomputing unit for energy-efficient parallel computations and a high-precision digital processing unit for accuracy-critical operations. This segmentation allows each component to operate in its optimal precision range, resolving the contradiction between energy efficiency and computational precision.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different precision levels are applied locally to different parts of the computational system. The memcomputing unit operates at low precision where energy efficiency is paramount, while the digital processing unit operates at high precision where accuracy is required. This local quality differentiation allows the system to achieve both energy efficiency and computational precision in appropriate contexts.

Inventive Principle:

Principle #3Local quality

2Productivity

If conventional memcomputing techniques are used, then computation speed is improved, but computational precision deteriorates due to device variability

Engineering Contradiction:

Improvecomputation speedVSAvoidcomputational precision

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The computational system is segmented into two distinct components: a low-precision memcomputing unit for energy-efficient parallel computations and a high-precision digital processing unit for accuracy-critical operations. This segmentation allows each component to operate in its optimal precision range, resolving the contradiction between energy efficiency and computational precision.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically switches between low-precision memcomputing and high-precision digital processing based on the computational requirements of different tasks. This dynamic adaptation allows the system to optimize for speed when using memcomputing and for precision when using digital processing, resolving the contradiction between computation speed and computational precision.

Inventive Principle:

Principle #15Dynamics

3Measurement precision

If high precision digital computing is used, then computational precision is improved, but energy consumption increases

Engineering Contradiction:

Improvecomputational precisionVSAvoidenergy consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

Different precision levels are applied locally to different parts of the computational system. The memcomputing unit operates at low precision where energy efficiency is paramount, while the digital processing unit operates at high precision where accuracy is required. This local quality differentiation allows the system to achieve both energy efficiency and computational precision in appropriate contexts.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system changes the precision parameter dynamically based on computational requirements. For energy-intensive tasks, the system switches to low-precision memcomputing mode. For accuracy-critical tasks, it switches to high-precision digital processing mode. This parameter change resolves the contradiction between computational precision and energy consumption.

Inventive Principle:

Principle #35Parameter changes

4Measurement precision

If high precision digital computing is used, then computational precision is improved, but computation speed deteriorates

Engineering Contradiction:

Improvecomputational precisionVSAvoidcomputation speed

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The computational system is segmented into two distinct components: a low-precision memcomputing unit for energy-efficient parallel computations and a high-precision digital processing unit for accuracy-critical operations. This segmentation allows each component to operate in its optimal precision range, resolving the contradiction between energy efficiency and computational precision.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically switches between low-precision memcomputing and high-precision digital processing based on the computational requirements of different tasks. This dynamic adaptation allows the system to optimize for speed when using memcomputing and for precision when using digital processing, resolving the contradiction between computation speed and computational precision.

Inventive Principle:

Principle #15Dynamics

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach enhances the energy efficiency and speed of computations by leveraging different precision levels for floating-point arithmetic, effectively addressing the limitations of conventional memcomputing systems.

Implementation Method 1

The computational memory may comprise an array of resistive memory elements having resistance or conductance values stored therein, the respective resistance or conductance values being programmable

Methodology Applied

Scientific EffectElectrical Resistance: Electrical Resistance

Implementation Method 2

Each of at least a subset of the resistive memory elements may be selected from the group consisting of phase change memory, metal oxide resistive random-access memory, conductive bridge RAM, and magnetic RAM

Methodology Applied

Scientific EffectPhase Change: Phase Change

Data Source

PatentUS10114613B2Mixed-precision memcomputing system

Publication Date: 2018.10.30 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A computing system includes computational memory and digital combinational circuitry operatively coupled with the computational memory. The computational memory is configured to perform computations at a prescribed precision. The digital combinational circuitry is configured to increase the precision of the computations performed by the computational memory. The computational memory and the digital combinational circuitry may be configured to iteratively perform a computation to a predefined precision. The computational memory may include circuitry configured to perform analog computation using values stored in the computational memory, and the digital combinational circuitry may include a central processing unit, a graphics processing unit and/or application specific circuitry. The computational memory may include an array of resistive memory elements having resistance or conductance values stored therein, the respective resistance or conductance values being programmable.