Memory String Threshold Encoding for Accurate AI Data Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory devices face challenges in efficiently implementing thermometer encoding for data searching in artificial intelligence applications, particularly in terms of design efficiency and accuracy.
Innovation Solution
A memory device with a memory string comprising multiple switch elements, each with distinct threshold voltage levels, is configured to generate a string current signal based on the comparison of input and stored data, allowing for efficient data searching through the use of word line signals to control the switch elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple switch elements with distinct threshold voltage levels are used for thermometer encoding, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The memory device is segmented into multiple memory cells, each containing switch elements with distinct threshold voltage levels. This segmentation allows the system to achieve high measurement precision through thermometer encoding while managing device complexity by dividing the overall structure into manageable, repeating units.
Solution Approach 2:
Each switch element within the memory cells is assigned a specific threshold voltage level, creating local quality variations that enable precise data searching. This local differentiation of threshold voltages allows the system to encode and compare multiple data values simultaneously, improving measurement precision without requiring a monolithic complex structure.
2Productivity
If multiple memory cells with different threshold voltage levels are implemented, then productivity is improved, but manufacturing precision requirements increase
Solution Approach 1:
The invention utilizes parameter changes by varying the threshold voltage levels of switch elements across different memory cells. This approach enables parallel data searching operations, improving productivity, while the threshold voltage is controlled as a key parameter during manufacturing to ensure consistent and accurate device performance.
3Reliability
If switch elements are coupled in series for comparison operation, then reliability is improved, but use of energy increases
Solution Approach 1:
The series-coupled switch elements are activated periodically through word line signals that apply voltage at specific time intervals. This periodic activation enables reliable data comparison operations while reducing overall energy consumption, as the switch elements are not continuously conducting current but only during the comparison cycles.
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
The solution enables accurate and efficient data searching by generating distinct current levels based on the comparison of input and stored data, enhancing the performance of memory devices in artificial intelligence applications.
Implementation Method 1
each with distinct threshold voltage levels, is configured to generate a string current signal based on the comparison of input and stored data
Data Source
AI summary
A memory device includes a memory string. The memory string is configured to stored store data, and compare input data with the store data to generate a string current signal, wherein the memory string comprises memory cells, the memory cells comprises switch elements coupled in series, when a quantized value of the input data is equal to a quantized value of the stored data, a first switch element of the switch elements is turned on and the string current signal has a first current level, and when the quantized value of the input data is different from the quantized value of the stored data, the first switch element of the switch elements is turned off and the string current signal has a second current level lower than the first current level.