Knowledge Graph Layer for Human-Guided Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated machine learning systems fail to incorporate human intuition, expertise, and domain knowledge, leading to suboptimal performance and a lack of understanding in machine learning processes, and there is a need to efficiently harness human knowledge to improve machine learning outcomes.
Innovation Solution
A system and method that captures and enhances data, models it, reviews the results, visualizes the output, and provides recommendations using a recommendations engine that learns from user inputs to improve machine learning outcomes over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional machine learning approaches are used, then implementation is straightforward with well-defined algorithms, but the system cannot effectively leverage unstructured data or human expertise
Solution Approach 1:
The system segments knowledge into discrete units called knowledge graphs with entities, attributes, and relationships. This segmentation allows unstructured data to be organized into manageable components that can be processed by machine learning algorithms, resolving the contradiction between handling unstructured data and maintaining system manageability.
Solution Approach 2:
The patent introduces knowledge graphs as an intermediary layer between unstructured data and machine learning algorithms. This intermediary structure enables the system to leverage human expertise and unstructured data without requiring complete reengineering of the ML pipeline, thus improving adaptability while controlling complexity.
2Measurement precision
If more data is collected to improve model accuracy, then learning performance increases, but data quality and relevance become harder to ensure
Solution Approach 1:
The system applies local quality by associating confidence scores with individual knowledge graph entities and relationships. This allows the system to handle large volumes of data while maintaining quality control at the local level, where each data point can be individually assessed and weighted according to its reliability.
Solution Approach 2:
The patent changes the parameter of data representation by transforming raw data into structured knowledge graphs with explicit confidence metrics. This parameter transformation enables the system to process larger datasets while maintaining accuracy through confidence-weighted learning.
3Reliability
If human expertise is incorporated into the system, then decision-making quality improves, but the system becomes less scalable
Solution Approach 1:
The system copies human expertise into structured knowledge graphs that can be systematically organized and distributed. Instead of relying on individual human experts, the system creates replicable knowledge structures that can be scaled across multiple applications and users while maintaining decision-making quality.
Solution Approach 2:
The knowledge graph structure serves multiple functions: it captures human expertise, structures unstructured data, and provides a unified interface for machine learning algorithms. This multi-functionality allows the system to maintain reliability while improving scalability through a single versatile framework.
4Loss of information
If the system processes more unstructured data, then knowledge coverage increases, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary action by pre-processing unstructured data into knowledge graphs before machine learning processing. This advance structuring of data into entities, attributes, and relationships reduces the computational burden during actual ML operations, thereby increasing knowledge coverage while controlling processing time.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method of harnessing knowledge and expertise to improve machine learning is disclosed. The system and method include capturing the data to input, preparing the captured data, enhancing the prepared data, modeling and learning the process associated with the enhanced data, reviewing the result of the learning and modeling to produce an output, visualizing the reviewed output, and input and recommendations from recommendations engine that make recommendations of techniques and configurations to use.