|
Machine learning, neural networks
Lattice theory for machine learning
D. V. Vinogradov Federal Research Center "Computer Science and Control" of Russian Academy of Sciences, Moscow, Russia
Abstract:
We represent theoretical foundations for similarity-based machine learning system construction. The key technique is Formal Concept Analysis, a modern branch of Lattice Theory. We introduce bitset encoding algorithms for objects described by both discrete and continuous attributes. Then we discuss Markov chain Monte Carlo method. After presentations of main steps of machine learning we provide a result on sufficient number of hypotheses to generate. We conclude with discussion of results of experimental approbation of our approach with respect to several datasets from UCI Machine Learning repository.
Keywords:
lattice, FCA, JSM-method, bitset, machine learning.
Citation:
D. V. Vinogradov, “Lattice theory for machine learning”, Artificial Intelligence and Decision Making, 2021, no. 1, 25–32; Scientific and Technical Information Processing, 49:5 (2022), 379–384
Linking options:
https://www.mathnet.ru/eng/iipr89 https://www.mathnet.ru/eng/iipr/y2021/i1/p25
|
Statistics & downloads: |
Abstract page: | 23 | Full-text PDF : | 32 | References: | 1 |
|