Time Series Retrieval Using Residual Network Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Content-based Time Series Retrieval (CTSR) systems face challenges in efficiently computing similarity scores between time series data from diverse domains, particularly in real-time interactions.
Innovation Solution
The method involves obtaining known time series from a database, computing pairwise distance matrices with learned templates, stacking these matrices to form tensors, and processing the tensors using a residual network to generate feature vectors. This process allows for efficient similarity determination between known and unknown time series.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a high-capacity model is used to measure similarity between time series from diverse domains, then measurement precision is improved, but computation time increases
Solution Approach 1:
The patent pre-computes and stores feature vectors for known time series in a database before retrieval operations. This preliminary action allows the system to avoid re-computing complex similarity measurements during real-time queries, thus maintaining high measurement precision while reducing computation time during actual retrieval operations.
Solution Approach 2:
The patent creates simplified representations (feature vectors) of complex time series data. These feature vectors serve as copies that capture the essential characteristics of the original time series in a compressed format, enabling fast similarity comparisons without processing the full complexity of the original data during retrieval operations.
2Productivity
If real-time interaction is required for time series retrieval, then productivity is improved, but model computation efficiency deteriorates
Solution Approach 1:
The patent extracts and pre-computes feature vectors from time series data and stores them in a database. This extraction separates the complex computation from the real-time retrieval operation, allowing the system to maintain high productivity during queries while the computational complexity is handled during the offline pre-processing phase.
Solution Approach 2:
The patent segments the time series retrieval process into two distinct phases: an offline phase for pre-computing and storing feature vectors, and an online phase for fast retrieval based on pre-extracted features. This segmentation allows the system to achieve real-time interaction performance without the burden of complex real-time model computations.
3Measurement precision
If pairwise distance matrices are computed for all known time series and templates, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent pre-computes pairwise distance matrices between known time series and learned templates, then stores these pre-computed results in a database. This preliminary action allows the system to retrieve accurate distance measurements quickly during queries without re-computing the complex distance matrices, thus improving measurement precision while managing processing complexity through offline computation.
Data Source
AI summary
Methods, systems, and computer program products are provided for content-based time series retrieval. An example system includes at least one processor configured to: obtain, from at least one database, a plurality of known time series; for each known time series of the plurality of known time series: compute a pairwise distance matrix between that known time series and each learned template of a plurality of learned templates to generate a plurality of pairwise distance matrices; stack the plurality of pairwise distance matrices together to generate a tensor; and process, with the residual network, the tensor, wherein the residual network receives, as input, the tensor, and provides, as output, a feature vector for that known time series; and provide the feature vector for each known time series of the plurality of known time series.