Blockchain-Verified Collaborative Deep Learning Model Updates

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Deep learning models face challenges in achieving strong generalization and high accuracy due to the limitations of single data sources and privacy concerns, which hinder collaborative training while being susceptible to malicious or low-quality data sources.

Innovation Solution

A collaborative deep learning method utilizing a global model shared among user terminals, where changes to the model parameters are stored on a blockchain, verified through consensus contracts, and updated based on evaluation matrices to ensure secure and accurate parameter convergence.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If collaborative deep learning is implemented to improve model generalization and accuracy through multiple data sources, then the model performance is improved, but the system becomes vulnerable to malicious users and low-quality data sources

Engineering Contradiction:

Improvemodel generalization and accuracyVSAvoidmalicious users and low-quality data sources

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent introduces blockchain technology as an intermediary layer between multiple user terminals and the global model. The blockchain stores parameter changes from different users, providing a trusted mediation mechanism that verifies and validates contributions before updating the global model, thus protecting against malicious inputs while maintaining collaborative benefits

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements a feedback mechanism where the blockchain records and verifies parameter changes from each user terminal. Through consensus mechanisms and evaluation matrices stored on the blockchain, the system provides feedback on the quality and validity of each contribution, enabling the global model to learn from reliable sources while rejecting malicious or low-quality inputs

Inventive Principle:

Principle #23Feedback

2Manufacturing precision

If multiple user terminals contribute parameter changes to improve the global model, then the model accuracy is enhanced, but the complexity of verifying and evaluating these changes increases

Engineering Contradiction:

Improvemodel accuracyVSAvoidverification and evaluation system

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The blockchain system performs self-verification through its inherent consensus mechanisms and cryptographic validation. Each parameter change is automatically verified and evaluated by the distributed network nodes, eliminating the need for complex centralized verification systems while maintaining high accuracy through decentralized trust

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The blockchain serves multiple functions simultaneously: it stores parameter changes, verifies their authenticity through consensus, evaluates their quality via evaluation matrices, and manages the global model updates. This multi-functionality reduces overall system complexity by consolidating verification and evaluation tasks into a single universal platform

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Productivity

If all parameter changes from all user terminals are used to update the global model, then the training speed is fast, but the model convergence becomes unstable due to malicious or low-quality data

Engineering Contradiction:

Improvetraining speedVSAvoidmodel convergence stability

Core Design Contradiction:

ProductivityVSStability of the object's composition

Solution Approach 1:

The system applies different quality standards and verification levels to different parameter changes based on their source and content. The blockchain evaluates each contribution individually and applies selective updating, giving appropriate weight to high-quality changes while filtering out malicious or low-quality inputs, thus maintaining both training efficiency and convergence stability

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system performs preliminary verification and evaluation of parameter changes before they are applied to the global model. The blockchain validates contributions in advance through consensus mechanisms and evaluation matrices, ensuring that only verified quality changes are incorporated, which prevents instability while maintaining fast training through pre-filtered high-quality data

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11954592B2Collaborative deep learning methods and collaborative deep learning apparatuses

Publication Date: 2024.04.09 TSINGHUA UNIVERSITY

AI summary

The disclosure provides a collaborative deep learning method and a collaborative deep learning apparatus. The method includes: sending an instruction for downloading a global model to a plurality of user terminals; receiving a set of changes from each user terminal; storing the set of changes; recording a hash value of the set of changes into a blockchain; obtaining a storage transaction number from the blockchain for the hash value of the set of changes; sending the set of changes and the storage transaction number to the plurality of user terminals; receiving the set of target user terminals from the blockchain; updating the current parameters of the global model based on sets of changes corresponding to the set of target user terminals; and returning the sending the instruction, to update the global model until the global model meets a preset condition.