Resistive Memory Crossbar Arrays for Matrix Vector Multiplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for matrix-vector multiplication are inefficient in terms of performance, area, and energy consumption, and lack the capability to effectively implement auto-associative neural network recall functionality.
Innovation Solution
The use of crossbar arrays composed of resistive memory devices, such as memristors, to store matrix coefficients and perform matrix-vector multiplication approximation, allowing for efficient operation below the threshold voltage to maintain memristance values and enable auto-associative neural network recall functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If digital integrated circuits are used to perform matrix-vector multiplication operations, then computation can be performed by Boolean logic circuits, but performance, area, and energy consumption are inefficient
Solution Approach 1:
The patent merges memory and computation functions into a single crossbar array architecture where resistive memory devices simultaneously store matrix coefficients and perform multiplication operations through analog current summation, eliminating the separation between memory access and computation in traditional digital systems
Solution Approach 2:
The patent replaces digital Boolean logic circuit computation with analog electrical current-based computation in the crossbar array, where matrix-vector multiplication is performed naturally through Ohm's law and Kirchhoff's current law, substituting complex digital logic operations with simple physical electrical phenomena
2Productivity
If digital integrated circuits are used to perform matrix-vector multiplication operations, then computation can be performed by Boolean logic circuits, but area efficiency is poor
Solution Approach 1:
The patent merges memory and computation functions into a single crossbar array architecture where resistive memory devices simultaneously store matrix coefficients and perform multiplication operations through analog current summation, eliminating the separation between memory access and computation in traditional digital systems
Solution Approach 2:
The patent transitions from one-dimensional sequential digital computation to two-dimensional parallel analog computation in the crossbar array, where multiple multiplications occur simultaneously across the array dimensions, dramatically improving computational throughput per unit area
3Adaptability or versatility
If threshold voltage is applied to enable state change in memristor, then memristance changes, but this causes difficulty in maintaining stable memristance values during computation
Solution Approach 1:
The patent applies different voltage conditions to different parts of the system: threshold voltage is applied only during programming operations to change memristance, while sub-threshold voltages are applied during computation operations to maintain stable memristance values, achieving both adaptability and reliability through spatially differentiated voltage control
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 significantly improves performance, area efficiency, and reduces energy consumption while enabling effective auto-associative neural network recall functionality by leveraging the unique properties of resistive memory devices.
Implementation Method 1
the instantaneous memristance varies with time and reflects the historical profile of the excitations through the device
Implementation Method 2
Based on circuit theory, an ideal memristor with memristance M builds the relationship between the magnetic flux φ and electric charge q that passes through the device, that is, dφ=M·dq
Implementation Method 3
the crossbar array inherently provides capabilities for this type of operation
Data Source
AI summary
An apparatus that performs the mathematical matrix-vector multiplication approximation operations using crossbar arrays of resistive memory devices (e.g. memristor, resistive random-access memory, spintronics, etc.). A crossbar array formed by resistive memory devices serves as a memory array that stores the coefficients of a matrix. Combined with input and output analog circuits, the crossbar array system realizes the method of performing matrix-vector multiplication approximation operations with significant performance, area and energy advantages over existing methods and designs. This invention also includes an extended method that realizes the auto-associative neural network recall function using the resistive memory crossbar architecture.