Some Estimates of Computational Complexity When Predicting the Properties of New Objects Using Characteristic Functions
| Autor: | M. I. Zabezhailo |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Empirical data
Computational complexity theory Computer science media_common.quotation_subject 05 social sciences Binary number Sample (statistics) 050905 science studies computer.software_genre Similarity analysis Algebraic operation Quality (business) Data mining 0509 other social sciences 050904 information & library sciences General Economics Econometrics and Finance computer Reliability (statistics) media_common |
| Zdroj: | Automatic Documentation and Mathematical Linguistics. 54:298-305 |
| ISSN: | 1934-8371 0005-1055 |
| DOI: | 10.3103/s0005105520060072 |
| Popis: | This paper discusses approaches to evaluating the quality of intelligent data analysis results in diagnostic tasks. The reliability (indisputability) of empirical dependencies established during training (interpolation–extrapolation) on precedents is evaluated using a special mathematical tool, that is, characteristic functions. Characteristic functions are generated on the available sample of empirical data based on similarity analysis of precedent descriptions, formalized as a binary algebraic operation. Some estimates of the computational complexity of applying the proposed mathematical technique of characteristic functions to predicting (diagnosing) the properties of newly studied precedents are presented. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink