Time Series Retrieval Using Context Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in rapidly and accurately retrieving stored irregular time series data based on similarity to a query time series, particularly in clinical and other data-intensive fields where manual annotation and hard-coded rules are often required.
Innovation Solution
A time series retrieval system that uses a trained encoder/decoder, such as a neural network, to convert time series data into context vectors, allowing for the identification and retrieval of similar time series based on similarity to a query, utilizing a user interface for input and output and employing nearest neighbor methods for efficient matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation and hard-coded rules are used to retrieve patients matching specific criteria, then retrieval accuracy can be achieved, but retrieval speed and efficiency deteriorate
Solution Approach 1:
The patent replaces manual annotation and hard-coded rule-based systems with a neural network-based embedding model. The system converts time series data into contextual embeddings that capture semantic meaning, enabling automated similarity search without manual intervention. This substitution of mechanical/manual processes with intelligent algorithms simultaneously improves both accuracy and speed.
Solution Approach 2:
The patent transforms time series data from raw numerical sequences into contextual embeddings through a neural network encoder. This parameter transformation converts complex temporal patterns into compressed vector representations that preserve semantic information, enabling efficient similarity comparison while maintaining high retrieval accuracy.
2Measurement precision
If complex rules and manual annotation are implemented to handle irregular time series data, then retrieval accuracy improves, but system complexity increases
Solution Approach 1:
The patent replaces complex manual annotation processes and hard-coded rule systems with a unified neural network-based embedding model. This single system handles the complexity of irregular time series data internally through learned representations, presenting a simple interface to users while managing all computational complexity within the model architecture.
Solution Approach 2:
The patent introduces contextual embeddings as an intermediary representation between raw time series data and retrieval operations. These embeddings serve as a bridge that captures complex temporal patterns and irregularities, allowing the system to handle sophisticated data characteristics without requiring complex external processing rules.
3Device complexity
If traditional methods are used to search irregular time series data, then system simplicity is maintained, but retrieval accuracy and similarity matching capability deteriorate
Solution Approach 1:
The patent transforms time series data from raw numerical sequences into contextual embeddings through a neural network encoder. This parameter transformation converts complex temporal patterns into compressed vector representations that preserve semantic information, enabling efficient similarity comparison while maintaining high retrieval accuracy.
4Measurement precision
If more detailed processing and normalization of time series data are performed, then retrieval accuracy improves, but processing time increases
Solution Approach 1:
The patent performs data normalization, encoding, and contextual embedding generation as preliminary actions during the indexing phase. By pre-processing time series data into contextual embeddings and storing them in advance, the system eliminates the need for repeated processing during query execution, significantly reducing retrieval time while maintaining high accuracy.
Solution Approach 2:
The patent creates compressed vector representations (contextual embeddings) as copies of the original time series data. These embedding copies capture the essential semantic information of the full time series, allowing fast similarity search on the compact representations without requiring access to or processing of the complete original data during retrieval operations.
Data Source
AI summary
A method (100) for identifying time series data using a time series retrieval system (800), comprising: receiving (120) a plurality of time series, each time series comprising a plurality of datapoints, wherein a least some of the plurality of times series comprise datapoints obtained at irregular time intervals within the time period; storing (130) the received plurality of time series in a database; generate (140) a context vector for each of the plurality of time series; receiving (150) a request for identification of one or more of the plurality of time series based on similarity to a time series query; identifying (160), based on similarity to the query time series context vector, one or more of the stored generated context vectors; retrieving (170) each time series associated with the identified one or more stored generated context vectors; and providing (180) the retrieved time series.