Feature Dictionary Encoding for Privacy-Aware ANN Bandwidth Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed training of artificial neural networks (ANNs) faces challenges related to data privacy and bandwidth limitations, as multiple processors or distributed computing can lead to concerns about data authenticity and network performance.
Innovation Solution
A system utilizing a feature dictionary for centralized training of ANNs, where multiple computing devices host different versions of ANNs and feature dictionaries. Inputs are encoded using a data compression algorithm, reducing data transmission and enhancing privacy through obfuscation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If distributed computing is used to train ANNs, then training effectiveness is improved, but data privacy concerns and network performance issues arise
Solution Approach 1:
The patent extracts only the essential features from input data using a feature dictionary, rather than transmitting or storing complete raw data. This allows distributed training to proceed with compressed feature representations, improving both privacy protection and network efficiency while maintaining training effectiveness
Solution Approach 2:
The feature dictionary acts as an intermediary between raw input data and the ANN training process. It transforms complex input data into compressed feature codes that can be safely shared across distributed systems, resolving the contradiction between data utility and privacy protection
2Productivity
If distributed computing is used to train ANNs, then training effectiveness is improved, but network bandwidth limitations worsen
Solution Approach 1:
The patent extracts only the essential features from input data using a feature dictionary, rather than transmitting or storing complete raw data. This allows distributed training to proceed with compressed feature representations, improving both privacy protection and network efficiency while maintaining training effectiveness
Solution Approach 2:
The patent changes the representation parameters of input data by encoding it into compressed feature codes through the feature dictionary. This parameter transformation dramatically reduces the quantity of data that needs to be transmitted over the network while preserving the essential information needed for effective ANN training
3Quantity of substance
If feature encoding is used to reduce data transmission, then bandwidth is improved, but data compression complexity increases
Solution Approach 1:
The feature dictionary is pre-computed and stored before the actual data transmission and training process. This preliminary action organizes the compression mappings in advance, so that during distributed training, only simple lookups and code transmissions are needed, reducing the complexity of real-time compression operations
Data Source
AI summary
A system having multiple devices that can host different versions of an artificial neural network (ANN) as well as different versions of a feature dictionary. In the system, encoded inputs for the ANN can be decoded by the feature dictionary, which allows for encoded input to be sent to a master version of the ANN over a network instead of an original version of the input which usually includes more data than the encoded input. Thus, by using the feature dictionary for training of a master ANN there can be reduction of data transmission.