The thesis will focus on the Lanczos method (and other methods such as the power method) used in the context of informational retrieval. The thesis will involve a review of the background and mathematics behind informational retrieval and the different methods (both direct and iterative) used, as well as an experimental study comparing the numerical behavior of Lanczos and another method, in both single and double precision, in terms of precision and recall for 2 or more datasets.
Seznam odborné literatury
*Eldén, Lars. Matrix methods in data mining and pattern recognition. Vol. 15. Siam, 2019.
*Strang, Gilbert. Linear algebra and learning from data. Wellesley-Cambridge Press, 2019.
*Eldén, Lars. "Numerical linear algebra in data mining." Acta Numerica 15 (2006): 327-384.
Předběžná náplň práce
Data Mining and informatics is a huge emerging topic, attracting interest from both industry and academia. Methods in data mining are, at their core, based in numerical linear algebra. Iterative methods in numerical linear algebra have a wide variety of applications, including analyzing text, web search engines, and optimizing neural networks. This thesis involves exploring the use of iterative methods within data science applications.
Předběžná náplň práce v anglickém jazyce
Data Mining and informatics is a huge emerging topic, attracting interest from both industry and academia. Methods in data mining are, at their core, based in numerical linear algebra. Iterative methods in numerical linear algebra have a wide variety of applications, including analyzing text, web search engines, and optimizing neural networks. This thesis involves exploring the use of iterative methods within data science applications.
- zadáno a potvrzeno stud. odd.