Auto-Load Circuit for Neural Network Lookup Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current arithmetic devices for neural networks face challenges in efficiently performing auto-load operations, particularly in activating and managing auto-load commands and addresses based on power source voltage levels, which affects the storage and retrieval of look-up table data for activation functions.
Innovation Solution
The implementation of an arithmetic device with an auto-command/address generation circuit that generates signals for auto-load operations, a first data storage circuit for outputting look-up table data corresponding to activation functions, and a second data storage circuit for storing this data, allowing for sequential activation and address changes during auto-load operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If auto-load operations are activated based on power source voltage levels, then the efficiency of neural network operations is improved, but the device complexity increases due to additional control circuits
Solution Approach 1:
The auto-command/address generation circuit automatically generates control signals and addresses for loading look-up table data without external intervention. The circuit self-activates based on power source voltage levels, performing the load operation autonomously to improve efficiency while minimizing the need for additional control logic
Solution Approach 2:
The circuit monitors power source voltage levels and prepares to activate auto-load operations in advance. By detecting voltage conditions beforehand and sequentially generating commands only when needed, the system avoids continuous operation and reduces overall device complexity while maintaining high efficiency during critical periods
2Productivity
If look-up table data is stored and retrieved efficiently, then the performance of activation functions is improved, but the data storage circuit complexity increases
Solution Approach 1:
The data storage system is divided into two separate circuits: a first data storage circuit for holding look-up table data and a second data storage circuit for storing processed results. This segmentation allows each circuit to be optimized for its specific function, improving activation function performance while keeping individual circuit complexities manageable
Solution Approach 2:
The auto-command/address generation circuit acts as an intermediary that coordinates data transfer between the first and second data storage circuits. It generates the necessary commands and addresses sequentially, enabling efficient data movement without requiring complex direct interfacing between storage circuits
3Reliability
If sequential activation of auto-load commands is implemented, then the reliability of power-up operations is improved, but the operation time increases
Solution Approach 1:
The auto-load operation uses periodic sequential activation of commands and addresses during power-up. The auto-command/address generation circuit generates signals in a predetermined sequence at controlled intervals, ensuring reliable initialization while completing the entire load operation within a limited time frame through rhythmic, predictable timing
Data Source
AI summary
An arithmetic device includes an auto-command/address generation circuit, a first data storage circuit, and a second data storage circuit. The auto-command/address generation circuit generates an auto-load selection signal that activates an auto-load operation based on a level of a power source voltage. In addition, the auto-command/address generation circuit generates an auto-load command for the auto-load operation. The first data storage circuit outputs look-up table data, to which an activation function is applied, based on the auto-load command. The second data storage circuit stores the look-up table data, output from the first data storage circuit, based on the auto-load command.