ANN Model Layer Removal for Real-Time Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing artificial neural network (ANN) models require high-end infrastructure for real-time and accurate predictions, which is not always available, and existing optimization techniques are limited in scope and utility, particularly for applications like traffic density prediction models that handle continuous input streams.
Innovation Solution
A method and system that assign random activation threshold values to activation functions in ANN models, identify removable layers based on activated neurons, evaluate the relative loss of removing these layers, and derive a modified ANN model by removing unnecessary layers to optimize performance without impacting accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex ANN models with multiple layers are deployed for real-time predictions, then prediction accuracy is improved, but processing time increases and high-end infrastructure is required
Solution Approach 1:
The patent extracts and removes redundant or less important layers from the ANN model while retaining the essential layers that contribute to prediction accuracy. This layer removal process reduces model complexity and processing time while maintaining the core predictive functionality through selective extraction of critical model components.
Solution Approach 2:
The patent changes the architectural parameters of the ANN model by adjusting the number of layers and neurons based on their activation patterns and contribution to predictions. By dynamically modifying model parameters such as layer count and neuron distribution, the system achieves real-time processing capability while preserving accuracy through data-driven parameter optimization.
2Measurement precision
If complex ANN models with multiple layers are deployed for real-time predictions, then prediction accuracy is improved, but infrastructure requirements increase
Solution Approach 1:
The patent extracts and removes redundant or less important layers from the ANN model while retaining the essential layers that contribute to prediction accuracy. This layer removal process reduces model complexity and processing time while maintaining the core predictive functionality through selective extraction of critical model components.
Solution Approach 2:
The patent changes the architectural parameters of the ANN model by adjusting the number of layers and neurons based on their activation patterns and contribution to predictions. By dynamically modifying model parameters such as layer count and neuron distribution, the system achieves real-time processing capability while preserving accuracy through data-driven parameter optimization.
3Productivity
If existing optimization techniques are applied to ANN models, then some performance improvement is achieved, but the scope and utility are limited particularly for real-time applications
Solution Approach 1:
The patent introduces dynamic adaptability to ANN optimization by continuously monitoring neuron activation patterns and adjusting model architecture in response to real-time data characteristics. This dynamic approach enables the model to adapt its structure based on actual usage patterns, making it versatile for various real-time applications including traffic density prediction with continuous input streams.
Solution Approach 2:
The patent implements feedback mechanisms where the performance and activation patterns of the ANN model are continuously evaluated, and this information is used to iteratively refine the model architecture. The feedback loop enables automatic adjustment of model parameters based on real-time performance metrics, enhancing both productivity and adaptability across different application scenarios.
Data Source
AI summary
The disclosure relates to method and system for improving performance of an artificial neural network (ANN) model. The method includes receiving the ANN model and input dataset. The ANN model includes neurons arranged in multiple layers and employing corresponding activation functions. The method further includes assigning a random activation threshold value to each of the corresponding activation functions, determining activated neurons in each layer for a majority of input data in the input dataset based on the random activation threshold value for each of the corresponding activation functions, identifying removable layers based on a number of activated neurons and a pre-defined threshold value, evaluating a relative loss of the ANN model upon removing each removable layer from the ANN model and for a random input data in the input dataset, and deriving a modified ANN model by removing one or more of the removable layers based on the evaluation.