Resistive Device Arrays for Scalable Matrix Operations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for computing maximization algorithms, such as those used in information processing systems, face inefficiencies due to the quadratic time complexity of matrix operations in von Neumann architecture, making real-time processing with a large number of independent sources computationally expensive.
Innovation Solution
Implementing matrix operations using a network of resistive device arrays, which perform forward and backward passes to update matrix weights in parallel, reducing computational steps and time complexity to O(n) by separating diagonal and off-diagonal terms across multiple arrays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If von Neumann architecture is used for matrix operations, then computational accuracy is maintained, but computation time increases quadratically with the number of independent sources N
Solution Approach 1:
The patent replaces the mechanical sequential processing of von Neumann architecture with an analog electrical system using resistive devices. Matrix operations are performed by applying voltage inputs to resistive device arrays where current flows through the devices according to Ohm's law, naturally computing matrix-vector multiplications in parallel without sequential mechanical steps.
Solution Approach 2:
The patent transitions from the sequential time dimension of digital computation to the spatial parallelism of analog computation. By mapping matrix operations onto a two-dimensional array of resistive devices, the system performs computations that would take O(N²) sequential steps in O(1) analog time, effectively adding a spatial dimension to the computational process.
2Reliability
If conventional computing architecture is used for large-scale matrix operations, then system stability is maintained, but computational cost becomes prohibitive for real-time applications
Solution Approach 1:
The patent replaces conventional digital computing architecture with an analog electrical system using resistive devices. Matrix operations are performed by applying voltage inputs to resistive device arrays where current flows through the devices according to Ohm's law, naturally computing matrix-vector multiplications in parallel without sequential mechanical steps.
Solution Approach 2:
The resistive device array performs computations autonomously through the physical laws of electricity. When voltage inputs are applied, the current distribution through the resistive network automatically computes the matrix operations without requiring external control or sequential processing steps, enabling self-service real-time computation.
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 enables fast and scalable computation of matrix operations, particularly for high-order matrices, by performing analog updates in parallel, thus reducing the computational burden and enabling real-time processing with large numbers of independent sources.
Implementation Method 1
a network of resistive device arrays may include a first array and a second array... passing a forward pass by pulsing an input voltage value from an input of the first array and an input of the second array, and reading output voltage values at an output of the first array and an output of the second array
Data Source
AI summary
In some aspects, a method may include initializing a first array and a second array with a random voltage value, passing a forward pass by pulsing an input voltage value from an input of the first array and an input of the second array, and reading output voltage values at an output of the first array and an output of the second array. The method may further include passing a backward pass into the inputs of both of the first and second arrays, and reading voltage values at the inputs of the first and second arrays. The method may further include updating, with the first array, a first matrix update on the first array, updating, with the second array, a first matrix update on the second, and updating, with the second array, a second matrix update on the second array.