Learn to use Transformed Temporal Similarity Search (Transformed TSS), a model designed to compress and search over time series data. TSS can take a window containing several thousand data points and compress it – reducing the dimensionality of the data by over 99% while retaining the original shape of the data. The smaller window size significantly reduces the memory and disk resources needed to insert and search over large time series datasets.
In this demo, we will use KDB.AI to perform vector searches over generated example market data to identify patterns of interest.
In addition to reducing memory and disk footprint, TSS supports ANN indexing. This means that Transformed TSS compressed embeddings can be attached to prebuilt indexes in the vector database. ANN indexing methods such as HNSW improve the speed and efficiency of vector searches.
Download the Jupyter Notebook at the GitHub repository.