Neural Network Communication Using Distance-Based Reference Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural network-based transmitters and receivers from different vendors have poor adaptability during communication due to the unwillingness of vendors to disclose trained neural networks, leading to challenges in adaptive communication.
Innovation Solution
Implement a method where a first apparatus processes data using a transmitter neural network and sends information about a calculated distance to a second apparatus, which uses this distance and a reference data to restore the original data, enabling adaptability through joint training and shared data usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vendors do not disclose trained neural networks, then vendor intellectual property is protected, but communication adaptability between different vendors deteriorates
Solution Approach 1:
The patent introduces a shared neural network as an intermediary component that is trained using data from multiple vendor-specific neural networks. This shared network serves as a mediator that enables different vendors' transmitters and receivers to communicate effectively without requiring them to disclose their proprietary neural networks. The shared neural network learns common patterns and representations that bridge the gap between different vendor implementations.
Solution Approach 2:
The patent segments the neural network functionality into two distinct parts: vendor-specific neural networks that remain proprietary and are not disclosed, and a shared neural network that is openly trained and used for cross-vendor compatibility. This segmentation allows each vendor to protect their intellectual property while still enabling interoperability through the shared component.
2Reliability
If joint training of transmitter and receiver neural networks is implemented, then communication performance is improved, but system complexity increases
Solution Approach 1:
The patent merges the training process of transmitter and receiver neural networks into a joint training framework. During joint training, both the transmitter neural network and receiver neural network are trained simultaneously using end-to-end optimization, allowing the system to achieve better communication performance by coordinating the optimization of both components together rather than separately.
3Adaptability or versatility
If distance calculation and data restoration methods are used, then communication adaptability is improved, but data transmission overhead increases
Solution Approach 1:
The patent extracts the essential communication capability from the proprietary neural networks by calculating a distance metric that captures the relationship between transmitted and received data. Instead of transmitting the entire neural network or large amounts of calibration data, the system extracts and transmits only the necessary distance information, which is then used to restore and adapt the data at the receiver side.
Data Source
AI summary
This application discloses a data processing method, which may be applied to a neural network-based communication scenario. The method includes: A first apparatus that serves as a sender may send a distance between two pieces of data to a second apparatus that serves as a receiver, where the two pieces of data may be data obtained after source data is processed by using a transmitter neural network, and data that has a reference function. In this way, after the data that has the reference function is introduced, a receiver neural network of the second apparatus can correctly receive data processed by using the transmitter neural network of the first apparatus, so that adaptability of communication between neural networks that are not jointly trained is improved.