Ternary Input Processing Circuit Using Current Leakage Switching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for processing input variables in computing systems are limited in efficiently handling ternary values and performing binary or ternary operations, particularly in the context of artificial neural networks, where hardware-based acceleration is needed for effective processing.
Innovation Solution
A method and apparatus utilizing a current leakage device to selectively connect the load paths of transistors to circuit nodes based on input variables, combined with analog-to-digital converters to ascertain and process currents, enabling the processing of ternary variables and facilitating ternary multiplication and accumulation operations using field effect transistors, such as FeFETs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional methods are used for processing input variables, then binary operations can be performed, but ternary value processing efficiency is insufficient
Solution Approach 1:
The processing apparatus is segmented into multiple specialized components: a first processing unit for binary input variables and a second processing unit for ternary input variables. This segmentation allows each unit to be optimized for its specific function, improving ternary processing efficiency without requiring complete redesign of the entire system.
Solution Approach 2:
The processing apparatus is designed with multi-functionality to handle both binary and ternary operations. The same hardware infrastructure can process different types of input variables (binary and ternary) by switching between different processing modes or units, thereby improving productivity without proportionally increasing device complexity.
2Speed
If hardware-based acceleration is implemented for ternary operations, then processing speed improves, but device complexity increases
Solution Approach 1:
A dedicated second processing unit acts as an intermediary component specifically for ternary operations. This intermediary unit bridges the gap between standard binary processing hardware and the need for efficient ternary operations, providing hardware acceleration without requiring complete architectural overhaul.
Solution Approach 2:
The processing apparatus employs dynamic switching between different processing units based on the input variable type. When ternary operations are detected, the system dynamically activates the specialized second processing unit, providing speed improvement only when needed, thereby minimizing the impact on overall device complexity.
3Adaptability or versatility
If specialized processing units for ternary values are added, then processing capability improves, but manufacturing complexity increases
Solution Approach 1:
The apparatus is manufactured as modular segments: a first processing unit for binary operations and a second processing unit for ternary operations. This segmentation allows each unit to be manufactured and tested independently, then integrated into the final system, reducing overall manufacturing complexity while improving versatility.
4Productivity
If current leakage devices are used to connect load paths to circuit nodes, then ternary variable processing is enabled, but circuit complexity increases
Solution Approach 1:
The current leakage device is extracted as a separate, dedicated component rather than being integrated into the main transistor structure. This extraction allows the leakage device to be independently controlled and optimized for ternary variable processing, enabling the functionality without significantly increasing the complexity of the main circuit configuration.
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 allows for efficient hardware-based processing of ternary values and operations, enhancing the performance of artificial neural networks by effectively encoding and processing ternary inputs and weights, thereby accelerating inference tasks.
Implementation Method 1
application of a first input variable to a control electrode of a first transistor
Implementation Method 2
application of a second input variable to a current leakage device, wherein the current leakage device is designed to connect a load path of the first transistor selectively to a first circuit node or a second circuit node based on the second input variable
Implementation Method 3
ascertainment of the variable that characterizes the current flowing through the first circuit node, by means of a first, for example current-based, analog-to-digital converter device
Implementation Method 4
use of a field effect transistor, FET, for example MOSFET; as the first transistor and/or a or the second transistor, wherein, for example, use is made of a ferroelectric field effect transistor, FeFET
Data Source
AI summary
A method for processing input variables, comprising: application of a first input variable to a control electrode of a first transistor; application of a second input variable to a current leakage device, wherein the current leakage device is designed to connect a load path of the first transistor selectively to a first circuit node or a second circuit node based on the second input variable.