Multiply-Accumulate Circuit with Segmented Weight Units
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analog-type circuits for multiply-accumulate operations in neural networks face challenges in simplifying circuit configuration and reducing power consumption while maintaining accurate results.
Innovation Solution
The proposed arithmetic apparatus includes multiple input lines and multiply-accumulate devices with positive and negative weight multiplication units, where the sum totals of absolute weight values are equal, allowing for the generation of a voltage signal representing the sum of product values, which reduces distortion and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If an analog-type circuit is used for multiply-accumulate operations, then power consumption is reduced, but circuit configuration complexity increases
Solution Approach 1:
The multiply-accumulate device is segmented into distinct functional units: multiple multiplication units (each handling specific weight ranges) and a shared accumulation unit. This segmentation allows each unit to be optimized independently while maintaining overall system simplicity, resolving the contradiction between analog efficiency and circuit complexity.
Solution Approach 2:
The accumulation unit is designed as a universal component that can accumulate charges from multiple different multiplication units simultaneously. This multi-functional design reduces the total number of components needed, as one accumulation unit serves multiple purposes across different weight ranges, thereby reducing circuit configuration complexity while maintaining analog power efficiency.
2Measurement precision
If multiple multiplication units with different weight ranges are used, then calculation accuracy is improved, but device complexity increases
Solution Approach 1:
Different multiplication units are assigned to handle specific weight ranges (e.g., first multiplication units for weights 0-1, second multiplication units for weights 1-2). This local specialization allows each unit to be optimized for its specific range, improving calculation accuracy for different weight magnitudes while keeping each individual unit relatively simple in structure.
Solution Approach 2:
Multiple multiplication units that handle different weight ranges are merged into a single multiply-accumulate device that shares a common accumulation unit. This merging approach improves calculation accuracy across different weight ranges while avoiding the complexity of completely separate accumulation units for each multiplication unit, thus balancing accuracy improvement with device complexity control.
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 configuration simplifies the circuit configuration and reduces power consumption while maintaining accurate multiply-accumulate operations by using voltage signals to represent the sum of product values, thereby improving the efficiency of neural network processing.
Implementation Method 1
an accumulation unit that accumulates charges corresponding to the product values
Implementation Method 2
generates a voltage signal representing a sum of the product values by a voltage on the basis of the voltage of the accumulation unit
Data Source
AI summary
An arithmetic apparatus includes input lines and multiply-accumulate devices. An electrical signal for an input value is input into each of the input lines. Multiplication units include a positive weight multiplication unit that generates a positive weight charge for a product value obtained by multiplying the input value by a positive weight value and/or a negative weight multiplication unit that generates a negative weight charge for a product value obtained by multiplying the input value by a negative weight value. An output unit of the multiply-accumulate device includes an accumulation unit that accumulates the positive weight charge and the negative weight charge, generates a voltage signal representing a sum of the product values by a voltage on the basis of the voltage of the accumulation unit, and outputs the multiply-accumulate signal on the basis of the voltage signal.