Convolutional Neural Network Activation Indicator Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Processing data for feature detection operations, such as object detection or object recognition, in methods like convolutional neural networks is computationally intensive, leading to high processing demands.
Innovation Solution
A method and system that process data in convolutional neural networks by generating inputs for activation functions, storing indicators of these inputs, and determining whether to provide outputs based on evaluations of these indicators, allowing for reduced processing by comparing signatures of image data frames to determine if significant changes have occurred.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is processed in convolutional neural networks for feature detection, then classification accuracy is improved, but computational load and processing demands increase
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing indicator values for inputs to activation functions before actual inference. These indicators capture essential information about the input data's relationship to the activation function's threshold behavior, allowing the system to make intelligent decisions about whether full processing is needed without performing the complete computationally intensive operation every time.
Solution Approach 2:
The patent changes the parameter representation by transforming the raw input data into indicator values that represent the input's relationship to the activation function. This parameter transformation allows the system to work with compressed representations that maintain the essential information needed for accurate classification while significantly reducing the computational requirements for subsequent processing decisions.
2Reliability
If all data frames are processed through the activation function, then classification accuracy is maintained, but processing time and energy consumption increase
Solution Approach 1:
The system performs preliminary computation of indicator values during training or initialization, storing these pre-computed indicators for later use during inference. This preliminary action eliminates the need to re-compute complex activation function relationships for every data frame, allowing rapid decisions about whether full processing is necessary based on the pre-stored indicator information.
Solution Approach 2:
The patent applies partial action by selectively processing only those data frames that require it, based on the indicator evaluation. Instead of universally processing all data frames through the complete pipeline, the system processes only the subset where the indicators suggest full processing is necessary, thereby reducing overall processing time while maintaining accuracy for critical cases.
Data Source
AI summary
Aspects of the present disclosure relate to a computer-implemented method of processing data portion. The method comprises processing a first data portion in a convolutional neural network to generate a first input to an activation function in the convolutional neural network; providing a first output by applying the activation function to the first input; and storing an indicator, representative of the first input to the activation function, for the first data portion. The method further comprises determining whether to provide a second output by applying the activation function to a second input, generated from a second data portion, based at least in part on an evaluation of the indicator for the first data portion.