Distributed Ledger Worker Nodes Collective Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for collective learning in distributed computing systems face inefficiencies due to straggling worker nodes, staleness, and resource mismanagement, particularly in asynchronous training processes, leading to reduced system performance and stability.

Innovation Solution

A distributed computer control system utilizing a distributed ledger arrangement and smart contracts to coordinate collective learning among worker nodes, ensuring accountability, reliability, and efficient resource utilization through dynamic synchronization mode switching and reward mechanisms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If synchronous training process is used, then training coordination is improved, but system time efficiency deteriorates due to straggler nodes

Engineering Contradiction:
Improvetraining coordinationVSAvoidsystem time efficiency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent implements dynamic synchronization mode switching that allows the system to transition between synchronous and asynchronous training processes based on real-time system conditions. This enables the system to adapt its coordination mechanism dynamically, using synchronous mode when coordination is critical and asynchronous mode when time efficiency is prioritized, thereby resolving the contradiction between training coordination and system time efficiency.

Inventive Principle:
Principle #15Dynamics

2Productivity

If asynchronous training process is used, then system time efficiency is improved, but training reliability deteriorates due to staleness

Engineering Contradiction:
Improvesystem time efficiencyVSAvoidtraining reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system dynamically adjusts the synchronization mode based on monitored performance metrics and system state. When training reliability is compromised due to staleness in asynchronous mode, the system can switch to synchronous mode to ensure proper coordination. This dynamic adaptation allows the system to maintain high productivity while preserving training reliability when necessary.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent incorporates feedback mechanisms that monitor training progress, node performance, and model staleness. This feedback information is used to determine when to switch between synchronous and asynchronous modes, ensuring that the system maintains optimal balance between time efficiency and training reliability based on actual system conditions.

Inventive Principle:
Principle #23Feedback

3Reliability

If centralized parameter servers are used, then model aggregation coordination is improved, but system complexity and single point of failure risk increase

Engineering Contradiction:
Improvemodel aggregation coordinationVSAvoidsystem architecture complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts the coordination function from a centralized parameter server architecture and distributes it across multiple worker nodes through peer-to-peer communication mechanisms. This eliminates the single point of failure inherent in centralized systems while maintaining model aggregation coordination through distributed consensus protocols, thereby reducing system architecture complexity and improving reliability.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system merges the roles of parameter server and worker nodes, allowing worker nodes to perform both training computations and model aggregation coordination. This consolidation eliminates the need for separate centralized parameter servers, reducing system complexity while maintaining coordinated model aggregation through the distributed worker node network.

Inventive Principle:
Principle #5Merging (Combining)

4Adaptability or versatility

If worker nodes operate remotely in different geographical locations, then system scalability is improved, but training coordination efficiency deteriorates

Engineering Contradiction:
Improvesystem scalabilityVSAvoidtraining coordination efficiency
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent introduces intermediary synchronization mechanisms and communication protocols that facilitate efficient coordination between geographically distributed worker nodes. These intermediaries manage data synchronization, model aggregation, and coordination tasks, enabling remote nodes to collaborate effectively despite geographical separation, thereby maintaining training coordination efficiency while preserving system scalability.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3851921B1Distributed computer control system and method of operation thereof via collective learning
Publication Date: 2026.03.11 UVUE LTD
  • EP3851921B1 patent drawingFigure 1
  • EP3851921B1 patent drawingFigure 2
  • EP3851921B1 patent drawingFigure 3

AI summary

There is provided a distributed computer control system and a method of operation thereof. The distributed computer control system includes a plurality of worker nodes. The collective learning of the worker nodes occurs within the distributed computer control system. The operation of distributed computer control system is coordinated by employing a distributed ledger arrangement. The distributed computer control system comprises a monitoring arrangement to provide learning parameters to the worker nodes by way of executing a smart contract. Each of the worker nodes is operable to train a computing model based on the learning parameters and a set of training data. The monitoring arrangement provides a reward to at least one of the worker nodes for training the computing model. The reward is determined based on a number and frequency of intermediary computing models provided by a given worker node to a remainder of the worker nodes, and a quality of the intermediary computing models.