Feature Dictionary Encoding for Bandwidth-Efficient ANN Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed training of artificial neural networks (ANNs) faces challenges related to data privacy concerns and network performance limitations, particularly when using multiple processors or distributed computing.
Innovation Solution
A system is implemented that uses a feature dictionary for centralized training of ANNs, where inputs are encoded and obfuscated at local devices, reducing data transmission and enhancing privacy by only transmitting encoded data over the network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If distributed computing with multiple processors is used for ANN training, then training effectiveness and network complexity are improved, but data privacy concerns and network performance limitations worsen
Solution Approach 1:
The patent extracts only the essential features from raw data and transmits these extracted features to the central server for training. This separation of feature extraction (done locally) from model training (done centrally) allows distributed computing benefits while protecting raw data privacy.
Solution Approach 2:
The patent introduces an intermediary encoding layer that transforms raw data into encoded representations before transmission. This intermediary step protects data privacy by ensuring that even if transmitted data is intercepted, the original raw data cannot be easily recovered.
2Productivity
If distributed computing with multiple processors is used for ANN training, then training effectiveness is improved, but network bandwidth requirements increase
Solution Approach 1:
The patent extracts only the essential features from raw data and transmits these extracted features to the central server for training. This separation of feature extraction (done locally) from model training (done centrally) allows distributed computing benefits while protecting raw data privacy.
Solution Approach 2:
The patent segments the training process into two distinct phases: local feature extraction and centralized model training. This segmentation allows each phase to be optimized independently, reducing overall network bandwidth requirements.
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.