Adaptive Neural Networks for Varying Settings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neural networks face accuracy loss and catastrophic memory issues when operating in varying environments, requiring extensive training data sets that can lead to decreased performance.
Innovation Solution
A system that collects training data for specific settings, associates characteristics with neural network coefficients and structures, and stores them in a database management system, allowing for adaptive reconfiguration of neural networks based on changing settings by retrieving and instantiating the appropriate coefficients and structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If training data sets are increased to include samples from many different settings, then the neural network can operate in more different settings, but the neural network loses accuracy and encounters catastrophic memory loss
Solution Approach 1:
The patent segments the training data by organizing it into multiple data structures, where each data structure corresponds to a specific setting or environment. Instead of using a single large training set that causes catastrophic forgetting, the system divides the training data into manageable segments that can be selectively applied based on the current operating conditions.
Solution Approach 2:
The patent implements dynamic selection of training data structures based on the current setting. The system adaptively chooses which trained neural network model to deploy by matching the current environmental parameters with the settings for which each model was trained, allowing the system to dynamically adapt to changing conditions without suffering from catastrophic memory loss.
2Adaptability or versatility
If training data sets are increased to include samples from many different settings, then the neural network can operate in more different settings, but the training process becomes more complex and resource-intensive
Solution Approach 1:
The patent segments the training data by organizing it into multiple data structures, where each data structure corresponds to a specific setting or environment. Instead of using a single large training set that causes catastrophic forgetting, the system divides the training data into manageable segments that can be selectively applied based on the current operating conditions.
Solution Approach 2:
The patent applies local quality by associating specific characteristics with each training data structure. Each data structure is optimized for a particular setting with specific quality attributes, allowing the system to select the most appropriate model for the current conditions rather than using a generic one-size-fits-all approach.
3Adaptability or versatility
If a single neural network is trained to operate in multiple settings, then the system can handle various environments, but the neural network encounters catastrophic memory loss
Solution Approach 1:
The patent segments the training data by organizing it into multiple data structures, where each data structure corresponds to a specific setting or environment. Instead of using a single large training set that causes catastrophic forgetting, the system divides the training data into manageable segments that can be selectively applied based on the current operating conditions.
Solution Approach 2:
The patent creates multiple copies of the neural network, each trained on a specific segment of the training data corresponding to a particular setting. Rather than modifying a single network to handle all settings (which causes catastrophic forgetting), the system maintains multiple specialized networks and selects the appropriate copy based on the current operating conditions.
Data Source
AI summary
Methods and systems that allow neural network systems to maintain or increase operational accuracy while being able to operate in various settings. A set of training data is collected over each of at least two different settings. Each setting has a set of characteristics. Examples of setting characteristic types can be time, geographical location, and/or weather condition. Each set of training data is used to train a neural network resulting in a set of coefficients. For each setting, the setting characteristics are associated with the corresponding neural network having the resulting coefficients and neural network structure. A neural network, having the coefficients and neural network structure resulted after training using the training data collected over a setting, would yield optimal results when operated in/under the setting. A database management system can store information relating to, for example, the setting characteristics, neural network coefficients, and/or neural network structures.