The Database Corner

In the age of big data, similarity search can provide efficient mechanisms to query large information repositories with objects having no natural order, e.g., images and other complex data, being able to handle different data types and formats as long as they are coupled with a meaningful, metric distance function.

Although the k-Nearest Neighbors (kNN) query is the most known subclass of similarity search, other query types, such as top-k, skyline, and result diversification, often appear in the context of data retrieval, data analysis, and data mining. Typically, similarity searching methods are based on the topological framework of Metric Spaces, allowing the design of efficient and flexible indexing structures and searching algorithms.

Inherent problems of searching large repositories containing complex data formats dominate the research in similarity search. The problems range from data embedding to distributed and parallel algorithms, with many challenges still unsolved (such as handling distance concentration due to the curse of dimensionality) or in constant evolution (such as adding seamless support to similarity searches to Data/Data Lake Warehouses).

If you are a student with basic programming and math skills and want to learn more about this topic, you can contact me!