Series Memory Cells for Neural Network Weighted Inputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory systems fail to effectively implement weighted inputs in neural networks, which are crucial for modeling the relative influence of inputs on neuron behavior, limiting their operational efficiency.
Innovation Solution
The integration of memory cells configured to generate weighted inputs by applying specific voltages to pairs of memory cells in series, where the threshold voltage of one cell acts as a weight, enabling proportional current flow and emulating synaptic behavior in neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If memory cells are configured to generate weighted inputs using series-connected cells with threshold voltages as weights, then neural network processing accuracy is improved, but device complexity increases
Solution Approach 1:
The patent merges the functions of weight storage and input processing into a single memory cell structure. By connecting memory cells in series and using threshold voltages as weights, the system combines multiple neural network operations (weighting, summation, activation) into unified memory cell operations, thereby improving processing accuracy while managing device complexity through functional integration.
Solution Approach 2:
The memory cells are designed to perform multiple functions: storing data, providing weighted inputs to artificial neurons, and enabling neural network computations. The series-connected memory cell configuration allows the same hardware structure to serve as both weight storage and signal processing element, enhancing versatility without requiring separate dedicated components for each function.
2Productivity
If standard memory systems are used without weighted input generation, then device complexity is reduced, but neural network operational efficiency deteriorates
Solution Approach 1:
The patent implements preliminary action by pre-configuring memory cells with specific threshold voltages that represent weights before neural network operations begin. This pre-establishment of weighted connections allows the system to perform neural network computations more efficiently during operation, as the weighting function is already embedded in the memory cell characteristics rather than requiring real-time computation.
3Measurement precision
If voltage is applied to series-connected memory cells to generate proportional current, then synaptic weight representation is improved, but energy consumption increases
Solution Approach 1:
The patent replaces traditional electronic multiplication operations with a physics-based approach using Ohm's law and series circuit characteristics. By applying voltage to series-connected memory cells and measuring the resulting current, the system performs weighted input generation through natural electrical phenomena rather than complex computational operations, improving energy efficiency while maintaining precise synaptic weight representation.
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 operational efficiency of neural networks by accurately representing synaptic weights, improving the processing and decision-making capabilities of memory systems.
Implementation Method 1
applying a second voltage to a second memory cell coupled in series with the first memory cell to activate the second memory cell so that current flows through the first and second memory cells... The first voltage and a threshold voltage of the second memory cell can be such that the current is proportional to a product of the first voltage and the threshold voltage of the second memory cell
Data Source
AI summary
A method can include applying a first voltage to a first memory cell to activate the first memory cell, applying a second voltage to a second memory cell coupled in series with the first memory cell to activate the second memory cell so that current flows through the first and second memory cells, and generating an output responsive to the current. The first voltage and a threshold voltage of the second memory cell can be such that the current is proportional to a product of the first voltage and the threshold voltage of the second memory cell.