License: Creative Commons Attribution 3.0 Unported license (CC BY 3.0)
When quoting this document, please refer to the following
DOI: 10.4230/LIPIcs.CSL.2021.10
URN: urn:nbn:de:0030-drops-134447
URL: https://drops.dagstuhl.de/opus/volltexte/2021/13444/
Go to the corresponding LIPIcs Volume Portal


van Bergerem, Steffen ; Schweikardt, Nicole

Learning Concepts Described By Weight Aggregation Logic

pdf-format:
LIPIcs-CSL-2021-10.pdf (0.7 MB)


Abstract

We consider weighted structures, which extend ordinary relational structures by assigning weights, i.e. elements from a particular group or ring, to tuples present in the structure. We introduce an extension of first-order logic that allows to aggregate weights of tuples, compare such aggregates, and use them to build more complex formulas. We provide locality properties of fragments of this logic including Feferman-Vaught decompositions and a Gaifman normal form for a fragment called FOW₁, as well as a localisation theorem for a larger fragment called FOWA₁. This fragment can express concepts from various machine learning scenarios. Using the locality properties, we show that concepts definable in FOWA₁ over a weighted background structure of at most polylogarithmic degree are agnostically PAC-learnable in polylogarithmic time after pseudo-linear time preprocessing.

BibTeX - Entry

@InProceedings{vanbergerem_et_al:LIPIcs:2021:13444,
  author =	{Steffen van Bergerem and Nicole Schweikardt},
  title =	{{Learning Concepts Described By Weight Aggregation Logic}},
  booktitle =	{29th EACSL Annual Conference on Computer Science Logic (CSL 2021)},
  pages =	{10:1--10:18},
  series =	{Leibniz International Proceedings in Informatics (LIPIcs)},
  ISBN =	{978-3-95977-175-7},
  ISSN =	{1868-8969},
  year =	{2021},
  volume =	{183},
  editor =	{Christel Baier and Jean Goubault-Larrecq},
  publisher =	{Schloss Dagstuhl--Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/opus/volltexte/2021/13444},
  URN =		{urn:nbn:de:0030-drops-134447},
  doi =		{10.4230/LIPIcs.CSL.2021.10},
  annote =	{Keywords: first-order definable concept learning, agnostic probably approximately correct learning, classification problems, locality, Feferman-Vaught decomposition, Gaifman normal form, first-order logic with counting, weight aggregation logic}
}

Keywords: first-order definable concept learning, agnostic probably approximately correct learning, classification problems, locality, Feferman-Vaught decomposition, Gaifman normal form, first-order logic with counting, weight aggregation logic
Collection: 29th EACSL Annual Conference on Computer Science Logic (CSL 2021)
Issue Date: 2021
Date of publication: 13.01.2021


DROPS-Home | Fulltext Search | Imprint | Privacy Published by LZI