Neural Network Group Switching for Multi-Environment Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neural networks face accuracy loss and catastrophic memory issues when operating in varying environments, leading to suboptimal performance as they are trained with increasingly large datasets to cover multiple settings.
Innovation Solution
The system adapts by collecting training data for specific settings, associating characteristics with neural network coefficients and structures, and storing this information in a database management system, allowing for the retrieval and instantiation of the appropriate neural network configuration based on current settings, enabling seamless transitions between different environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If training data sets are increased to cover multiple different settings, then the neural network can operate in more environments, but accuracy is lost and catastrophic memory loss occurs
Solution Approach 1:
The patent segments the training process by creating separate training data sets for different settings (e.g., sunny conditions, rainy conditions, night conditions). Each neural network is trained independently on setting-specific data, allowing the system to maintain high accuracy for each individual setting while achieving versatility through multiple specialized networks.
Solution Approach 2:
The patent applies local quality by training each neural network with setting-specific characteristics. Each network develops specialized coefficients and structures optimized for its particular setting, rather than using a single generic network. This allows each network to maintain high accuracy for its specific environment while the system as a whole handles multiple settings.
2Adaptability or versatility
If training data sets are increased to cover multiple different settings, then the neural network can operate in more environments, but catastrophic memory loss occurs
Solution Approach 1:
The patent segments the learning task across multiple specialized neural networks, each trained on specific setting data. This prevents any single network from attempting to memorize all settings simultaneously, avoiding catastrophic forgetting while maintaining the ability to operate across multiple environments through network selection based on current settings.
Solution Approach 2:
The patent changes the parameter of training data composition by using setting-specific data sets for each network. This allows each network to optimize its memory allocation for relevant settings, preventing memory loss by not exposing individual networks to conflicting information from diverse settings.
3Device complexity
If a single neural network is used for multiple settings, then device complexity is reduced, but accuracy loss occurs
Solution Approach 1:
The patent segments the neural network system into multiple specialized networks, each optimized for specific settings. This segmentation improves accuracy for each setting while the overall system complexity is managed through automated network selection and switching based on environmental conditions.
Solution Approach 2:
The patent introduces dynamic selection of neural networks based on current settings. The system dynamically determines which specialized network to use based on environmental conditions, making the overall system adaptable while maintaining high accuracy through setting-specific network selection.
Data Source
AI summary
Methods and systems for controlling an autonomous machine. The autonomous machine has sensors generating input data, while the controller includes two or more neural networks that inference using the input data and generate output data. The neural networks can be trained using an identical set of training data set. The output data from each of the neural networks are monitored to ensure that the integrity of the operation is maintained by, for example, the output data from one neural network is compared with the output data from another neural network to verify the consistency. If the comparison yields that the integrity of the system is not maintained at an acceptable level, the controller can stop using the output in controlling the autonomous machine.