Title:
APPLICATION OF A BAYESIAN/CAUSAL BASE COST-BASED OPTIMIZER TO A RELATIONAL BASE
Document Type and Number:
WIPO Patent Application WO/2024/042379
Kind Code:
A1
Abstract:
Version v1 of the Bayesian/causal base optimizer provides significant improvements. - The optimizer is able to recognize advanced data types (named entities) if the analyst has not defined them. - It is able to recognize roles if the analyst has not defined them. - Its algorithm implements the Bayesian network on super-types as well (in contrast to version v0). The algorithm divides the data model into triplets (S, V, L) and creates the histogram on each triplet. This is another important step forward in solving the problem of computing cardinality to which databases are subject at present. This is also a major theoretical and practical step in terms of completely integrating the concepts of Bayesian networks into the engine of a database. All this makes it possible to apply the Bayesian/causal base cost-based optimizer to a relational base, and makes it possible to solve the computation of cardinality more effectively in the world of existing relational bases. The idea is to create an additional scheme in the relational base and to install the module for generating histograms of the Bayesian template for a relational data model. Next, it is necessary to modify the algorithm of the relational database optimizer so as to allow it to search for "external cardinality" and the order in which the tables are accessed, the 2 elements having been computed by the Bayesian base. Therefore, the concept of D-separation and D-connections of the Bayesian base makes it possible to improve the BI/Big Data part.
More Like This:
Inventors:
JASTREBIC DRAGUTIN (FR)
Application Number:
PCT/IB2023/056190
Publication Date:
February 29, 2024
Filing Date:
June 15, 2023
Export Citation:
Assignee:
LYTICSWARE (FR)
International Classes:
G06F16/2453
Foreign References:
| US20020103793A1 | 2002-08-01 | |||
| FR3117229A1 | 2022-06-10 |
Other References:
ZONGHENG YANG ET AL: "NeuroCard: One Cardinality Estimator for All Tables", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 2 November 2020 (2020-11-02), XP081804472
Download PDF:
Previous Patent: CRYPTOCURRENCY MINER AND JOB DISTRIBUTION
Next Patent: DECORATIVE PANEL AND METHOD FOR MANUFACTURING DECORATIVE PANELS
Next Patent: DECORATIVE PANEL AND METHOD FOR MANUFACTURING DECORATIVE PANELS