Neural Network Circuit Command Fetch Unit for Embedded IoT
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Convolutional neural networks (CNNs) face challenges in being implemented in embedded devices like IoT devices due to the difficulty of embedding large-scale dedicated circuits and the limited hardware resources, such as CPU or memory, which hinder sufficient operational performance.
Innovation Solution
A neural network circuit comprising a convolution operation circuit, a quantization operation circuit, and a command fetch unit that performs convolution operations, quantization operations, and reads commands from external memory to optimize performance and resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If large-scale dedicated circuits are used for convolutional neural networks, then operational performance is improved, but device complexity increases and embeddability in embedded devices deteriorates
Solution Approach 1:
The neural network processing is divided into separate functional modules: a convolution operation circuit for performing convolution operations, a quantization operation circuit for performing quantization operations, and a command fetch unit for reading commands. This segmentation allows each module to be optimized independently and embedded in resource-constrained devices while maintaining overall performance.
Solution Approach 2:
The convolution operation circuit and quantization operation circuit are designed as universal hardware components that can process different neural network operations through command-based control. The command fetch unit reads operation commands from external memory, allowing the same hardware circuit to perform multiple functions based on different command sequences, reducing the need for dedicated circuits for each operation type.
2Device complexity
If software alone is used for convolutional neural networks in embedded devices, then device complexity is reduced, but operational performance deteriorates
Solution Approach 1:
A command fetch unit serves as an intermediary between external memory and the operation circuits. It reads operation commands from external memory and supplies them to the convolution or quantization operation circuits, enabling the hardware to execute complex neural network operations without requiring all logic to be embedded in the device itself. This intermediary approach allows performance closer to dedicated hardware while maintaining embeddability.
3Measurement precision
If operations are performed without quantization, then computational precision is improved, but power consumption and processing time increase
Solution Approach 1:
The quantization operation circuit changes the precision parameter of data by performing quantization operations on convolution output data. This parameter change reduces the bit-width of data representations, thereby reducing power consumption and processing time while maintaining sufficient precision for embedded device applications. The circuit can adjust quantization parameters based on command inputs to balance precision and energy consumption.
Data Source
AI summary
A neural network circuit comprising a convolution operation circuit that performs a convolution operation on input data; a quantization operation circuit that performs a quantization operation on convolution operation output data from the convolution operation circuit; and a command fetch unit that reads, from an external memory, commands for operating the convolution operation circuit or the quantization operation circuit.