Unsupervised Vector Embeddings for Unseen Data Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional AI-powered databases lack the ability to execute cognitive intelligence queries for unseen data, limiting their applicability and usefulness, especially in supporting multiple SQL types for functions like prediction.
Innovation Solution
The method involves generating vector embeddings for training data using unsupervised machine learning models, allowing for the identification of closest matches in unseen data to perform tasks such as prediction, classification, and data imputation, through techniques like Modified Best Matching and Smoothing Inverse Frequency operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional supervised training methods are used for AI database models, then the model can be trained on labeled data, but it cannot perform cognitive intelligence queries on unseen data
Solution Approach 1:
The patent inverts the conventional supervised learning approach by using unsupervised learning to generate vector embeddings without labeled data. Instead of requiring labeled training data to learn representations, the system learns from raw unstructured data alone, then uses these embeddings to perform cognitive intelligence queries on unseen data. This inversion enables the model to adapt to new data without requiring retraining with labels.
Solution Approach 2:
The system performs self-service by automatically generating vector embeddings from unstructured training data without human annotation. The unsupervised learning model autonomously learns semantic representations and uses them to enable cognitive queries on unseen data, eliminating the need for external labeled data or manual intervention for model training.
2Adaptability or versatility
If semantic vectors are generated from unstructured representation of structured relational data, then cognitive intelligence queries can be performed, but unseen data can only undergo exploratory analysis
Solution Approach 1:
The patent changes the fundamental parameter of how semantic vectors are generated by transitioning from supervised learning (requiring labeled data) to unsupervised learning (using only raw data). This parameter change enables the system to generate reliable vector embeddings for unseen data, transforming it from exploratory analysis only to supporting full cognitive intelligence queries with improved inference accuracy.
3Adaptability or versatility
If AI database methodology is limited to explored data only, then the system remains simple, but it lacks wider applicability and usefulness
Solution Approach 1:
The patent implements universality by designing a system where unsupervised vector embeddings serve multiple functions: they enable prediction queries, classification, similarity searches, inductive reasoning, semantic clustering, and data imputation. This single unsupervised embedding framework provides multi-functional support for various SQL types and cognitive intelligence queries, greatly expanding applicability without proportionally increasing system complexity.
Data Source
AI summary
A computer-implemented method of performing queries using Artificial Intelligence (AI) database embeddings includes the operations of generating a plurality of vector embeddings describing a training data from a database for training a machine learning model. A test vector embedding is generated from the plurality of vector embeddings based on training data for unseen data from one or more rows of the database. One or more vectors from the plurality of vector embeddings describing the training data that are a closest match to the test vector embedding are identified. A task is determined based upon the unseen data. The determined task is performed using the trained machine learning model.