Comparator Arithmetic Circuit for Accurate Neural Product-Sum
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural network hardware devices require a large number of units to perform product-sum operations, which can be complex and inaccurate, especially in simulating neurons with high accuracy and simplicity.
Innovation Solution
An arithmetic device configured with a first resistor, a second resistor, a comparator, M cross switches, and M coefficient circuits, capable of executing product-sum operations with binary input signals and coefficients, utilizing constant current sources to switch between different current values based on input and coefficient values, and performing sign function processing to generate binary output signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of units are used to perform product-sum operations in neural network hardware, then the accuracy of neural network operations can be improved, but the device complexity increases
Solution Approach 1:
The patent combines multiple product-sum operation units into a single integrated circuit that processes multiple input signals simultaneously. The circuit integrates resistors, constant current sources, and switching elements to perform parallel computations, reducing the number of discrete units needed while maintaining accuracy.
Solution Approach 2:
The invention creates a universal arithmetic device that can handle multiple input signals and coefficients through a single configurable circuit. By using switching elements controlled by coefficient signals, the same hardware structure can perform different product-sum operations, eliminating the need for dedicated units for each operation.
2Device complexity
If a simple configuration is used for product-sum operation units, then the device complexity is reduced, but the accuracy of simulating neurons deteriorates
Solution Approach 1:
The patent uses parameter changes in the form of coefficient signals to control switching elements, allowing the same simple circuit configuration to adapt to different computational requirements. By changing the switching states based on coefficient values, the circuit maintains computational accuracy without increasing structural complexity.
Solution Approach 2:
The invention introduces constant current sources as intermediary elements that linearize the relationship between input signals and output currents. These current sources act as mediators that convert binary input signals into proportional current values, improving the linearity and accuracy of the simulation while keeping the overall configuration simple.
3Ease of operation
If binary input signals and coefficients are used for product-sum operations, then the ease of operation is improved, but the manufacturing precision requirements increase
Solution Approach 1:
The patent segments the computation into discrete binary input signals and coefficient signals that control switching elements. This segmentation allows for easier operation with binary data while managing manufacturing precision by confining critical analog operations to well-defined circuit elements with controlled tolerances.
Solution Approach 2:
The invention replaces complex mechanical or multi-state switching mechanisms with electronic switching elements controlled by binary signals. This substitution simplifies operation while managing manufacturing precision through standard semiconductor fabrication processes that can reliably produce the required resistor and switching element characteristics.
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 device achieves high accuracy in simulating neurons with a simple configuration by executing product-sum operations and improving linearity, thereby enhancing the accuracy of neural network operations.
Implementation Method 1
Each of the M coefficient circuits includes a first constant current source and a second constant current source, and is capable of changing a current difference between a current flowing through the first constant current source and a current flowing through the second constant current source
Implementation Method 2
The comparator is configured to output the output signal representing a value corresponding to a result of comparison between a first voltage generated at the first comparison terminal and a second voltage generated at the second comparison terminal
Data Source
AI summary
According to an embodiment, an arithmetic device includes a comparator, M cross switches, and M coefficient circuits. The comparator compares a first voltage generated at a first comparison terminal and a second voltage generated at a second comparison terminal. The M cross switches are provided corresponding to the M input signals. The M coefficient circuits are provided corresponding to the M coefficients, and each includes a first constant current source and a second constant current source. Each of the M cross switches performs switching between a straight state and a reverse state. In each of the M coefficient circuits, the first constant current source is connected between a positive output terminal of the corresponding coefficient circuit and a reference potential, and the second constant current source is connected between a negative output terminal of the corresponding coefficient circuit and the reference potential.