Shared Neural Network Layers for Low-Memory Vehicle AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural networks in vehicles consume significant resources due to their large size and complexity, necessitating a reduction in memory and processing requirements.
Innovation Solution
Implementing a group of neural networks with shared layer portions that are shared between different networks, determined based on input constraints and optimized for memory and performance, allowing for dynamic provisioning based on vehicle conditions and resource availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural networks are made large and complex to provide highly accurate responses under varying circumstances, then response accuracy is improved, but resource consumption increases
Solution Approach 1:
The patent merges multiple neural networks into a group structure where shared layer portions are common to multiple networks. This combining approach allows the system to maintain multiple specialized networks for different driving scenarios while sharing common computational resources, thereby reducing overall resource consumption while preserving response accuracy across varying circumstances
Solution Approach 2:
The shared layer portions serve multiple functions across different neural networks within the group. A single shared layer can be utilized by multiple specialized networks, making the computational resources universal and multi-functional. This eliminates redundant storage and computation, reducing resource consumption while maintaining the capability to handle various driving scenarios through different specialized networks
2Adaptability or versatility
If neural networks are made large to handle varying driving scenarios, then adaptability is improved, but memory consumption increases
Solution Approach 1:
The patent segments neural networks into shared layer portions and specialized portions. The shared layers contain common computational resources that can be reused across multiple scenarios, while specialized portions handle scenario-specific processing. This segmentation allows the system to be adaptable to various driving scenarios without requiring each scenario to have a complete separate network, thereby reducing memory consumption
Solution Approach 2:
Multiple neural networks are merged into a group structure where shared layer portions are common to multiple networks. This merging allows the system to maintain adaptability across different driving scenarios by having multiple specialized networks available, while the shared portions reduce the total memory footprint by eliminating redundancy
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of producing a group of neural networks, the method includes determining a first layer potion that is shared by a first neural network sub-group of the group; determining second layer portions that are sharable by second neural network sub-groups, such that different second layer portions are shared by different second neural network sub-groups of the group; and determining third layer portions that are sharable by third neural network sub-groups, such that different third layer portions are shared by different third neural network sub-groups of the group; wherein each neural network of the group further comprises a unique combination of layer portions.