Graph Embedding for Biological Dataset Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical decision support systems struggle to accurately classify cancer stages and detect new classes within datasets, particularly in late-stage cancer diagnosis and treatment, where conventional methods fail to provide efficient and effective treatment due to high-dimensional data complexity and human error in training sets.
Innovation Solution
The use of Graph Embedding to reduce dimensionality in biological datasets, allowing for supervised or unsupervised classification by assigning features to an embedded space based on posterior likelihood, partitioning features into disjoint regions, and generating new distributions for identifying new classes, thereby improving classification accuracy and detecting novel classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional classification methods are used on high-dimensional biological datasets, then the system can process the data, but the classification accuracy deteriorates due to high-dimensional data complexity
Solution Approach 1:
The patent applies Graph Embedding to transform high-dimensional biological data into a lower-dimensional embedded space while preserving the intrinsic manifold structure. This dimensionality reduction technique maps data points from high-dimensional space to a lower-dimensional space, reducing complexity while maintaining classification accuracy by preserving the essential relationships and patterns in the data.
2Adaptability or versatility
If conventional classification methods are used, then the system can operate, but the ability to detect new classes deteriorates due to reliance on pre-defined classes
Solution Approach 1:
The patent employs unsupervised learning algorithms that dynamically adapt to the data structure without relying on pre-defined class labels. The system automatically discovers and identifies new classes by analyzing the intrinsic patterns and relationships in the data, making the classification system flexible and adaptable to previously unknown categories while maintaining operational efficiency.
3Reliability
If human experts manually classify data, then the system can identify patterns, but the reliability deteriorates due to human error in training sets
Solution Approach 1:
The patent implements automated machine learning algorithms that self-train and self-validate without requiring manual expert intervention. The system automatically processes biological data, applies Graph Embedding for dimensionality reduction, and performs classification using supervised or unsupervised learning algorithms, thereby eliminating human errors in training set creation and improving classification consistency and reliability.
Data Source
AI summary
This invention relates to supervised or unsupervised classification of biological datasets. Specifically, the invention relates to the use of Graph Embedding as a method of reducing dimensionality thereby improving supervised classification of classes, both conventional and new ones.