MoL-2022-15: Filipetto, Valentino (2022) Constructing queries from data examples. [Report]
![[thumbnail of MoL-2022-15.text.pdf]](https://eprints.illc.uva.nl/style/images/fileicons/text.png) |
Text
MoL-2022-15.text.pdf Download (575kB) |
Abstract
The Query-by-Example problem (QBE) is about the existence of a fitting query for a given database instance and positive/negative labelled data examples. The question is: given an input database instance and sets of user examples, does there exist an explanatory query whose answers to it include the positive examples but not the negatives? Intuitively, what a fitting query does it to generalize the data according to certain constraints given by the user examples. From a complexity theoretical-standpoint, it was proven in 2010 that the Query-by-Example problem is coNExpTime-complete if we consider conjunctive queries as query language.
While in the relevant literature it is sometimes assumed that the fitting queries for the Query-by-Example problem cannot contain constants, in this thesis we consider precisely what happens if constants can appear in them and we check is there are changes in terms of computational complexity. Throughout the thesis we show how natural it is to allow constants in the queries and we present inputs for QBE where, if constants are now allowed to appear in the queries, no fitting query exists. Our analysis shows that, in most cases, constants do not really influence the computational complexity of QBE.
| Item Type: | Report |
|---|---|
| Report Nr: | MoL-2022-15 |
| Series Name: | Master of Logic Thesis (MoL) Series |
| Year: | 2022 |
| Subjects: | Computation Logic |
| Depositing User: | Dr Marco Vervoort |
| Date Deposited: | 25 Aug 2022 16:03 |
| Last Modified: | 25 Aug 2022 16:03 |
| URI: | https://eprints.illc.uva.nl/id/eprint/2214 |
Actions (login required)
 |
View Item |
