Memory Cell Current Summation for Euclidean Distance Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory technologies lack efficient methods for calculating the Euclidean distance between data points, which is crucial for similarity measurements in fields like geometry and deep learning.
Innovation Solution
A memory device comprising memory strings with switch elements that generate cell current signals proportional to the square of the difference between store and input bits, allowing for the calculation of Euclidean distances through the summation of these signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional memory technologies are used for Euclidean distance calculation, then the calculation can be performed, but the efficiency and precision are insufficient
Solution Approach 1:
The patent replaces conventional electronic computation systems with a physical memory system that directly calculates Euclidean distance through analog current signals. Memory cells store data as threshold voltage levels, and during read operations, current values flowing through selected memory strings are proportional to the squared difference between stored and input values. These currents are summed to produce a bit line signal proportional to the squared Euclidean distance, eliminating the need for traditional mechanical or electronic calculation circuits and significantly improving both efficiency and precision.
2Measurement precision
If multiple memory cells are used to improve calculation accuracy, then the precision improves, but the device complexity increases
Solution Approach 1:
The patent makes standard memory cells multi-functional by enabling them to perform both data storage and Euclidean distance calculation simultaneously. The same memory cells that store data in threshold voltage levels are used to generate current signals proportional to the squared difference between stored and input values. This eliminates the need for separate calculation hardware, reducing device complexity while maintaining high precision through the parallel operation of multiple memory cells.
3Measurement precision
If threshold voltage levels are increased to improve signal differentiation, then the measurement precision improves, but the energy consumption increases
Solution Approach 1:
The patent employs sense amplifiers that automatically detect and amplify the small current differences generated by memory cells during read operations. The sense amplifiers convert the analog current signals, which are proportional to the squared difference between threshold voltage levels and input voltages, into distinguishable digital signals. This self-service mechanism allows for high signal differentiation precision without requiring excessive threshold voltage levels, thereby controlling energy consumption while maintaining measurement precision.
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
Enables efficient calculation of Euclidean distances by generating bit line signals proportional to the square of the distance between data points, enhancing similarity measurements in memory operations.
Implementation Method 1
a current value of the cell current signal is proportional to a square of a difference between a threshold voltage level of the switch element and a voltage level of a control terminal signal
Data Source
AI summary
A memory device includes first and second memory strings. The first memory string includes a first switch element and generates a first cell current signal. The second memory string comprising a second switch element and configured to generate a second cell current signal. The first and second switch elements operate as a first memory cell storing a first store bit, control terminals of the first and second switch elements receive first and second word line signals, respectively, the first and second word line signals carries a first input bit, and a summation of a current value of the first cell current signal and a current value of the second cell current signal is proportional to a square of a difference between a logic value of the first store bit and a logic value of the first input bit.