Proceedings of the 17th Conference on Computer Science and Intelligence Systems

DOI: http://dx.doi.org/10.15439/2022F48

Citation: Proceedings of the 17th Conference on Computer Science and Intelligence Systems, M. Ganzha, L. Maciaszek, M. Paprzycki, D. Ślęzak (eds). ACSIS, Vol. 30, pages 205–209 (2022)

Full text

Abstract. The paper reports some results on neural architectures for learning numerical concepts from visual data. We use datasets of small images with single-pixel dots (one to six per image) to learn the abstraction of small integers, and other numerical concepts (e.g. even versus odd numbers). Both fully-connected and convolutional architectures are investigated. The obtained results indicate that two categories of numericalproperties apparently exist (in the context of discussed problems). In the first category, the properties can be learned without acquiring the counting skills, e.g. the notion of small, medium and large numbers. In the second category, explicit counting is embedded into the architecture so that the concepts are learned from numbers rather than directly from visual data. In general, we find that CNN architectures (if properly crafted) are more efficient in the discussed problems and (additionally) come with more plausible explainability.

References

1. A. J. Kersey and J. F. Cantlon, “Primitive concepts of number and the developing human brain,” Language Learning and Development, vol. 13, no. 2, pp. 191–214, 2017. http://dx.doi.org/10.1080/15475441.2016.1264878
2. M. H. Fischer and S. Shaki, “Number concepts: abstract and embodied,” Phil. Trans. Royal Society B, vol. 373, no. 1752, p. 20170125, 2018. http://dx.doi.org/10.1098/rstb.2017.0125
3. E. Walach and L. Wolf, “Learning to count with cnn boosting,” in Proceedings of the 14th European Conference on Computer Vision, part II, vol. LNCS 9906, 2016. http://dx.doi.org/10.1007/978-3-319-46475-6_41 pp. 660–676.
4. S. Sabathiel, J. L. McClelland, and T. Solstad, “Emerging representations for counting in a neural network agent interacting with a multimodal environment,” vol. ALIFE 2020: The 2020 Conference on Artificial Life, 2020. http://dx.doi.org/10.1162/isal_a_00333 pp. 736–743.
5. V. Troiani, J. E. Peelle, R. Clark, and M. Grossman, “Is it logical to count on quantifiers? dissociable neural networks underlying numerical and logical quantifiers,” Neuropsychologia, vol. 47, no. 1, pp. 104–111, 2009. http://dx.doi.org/https://doi.org/10.1016/j.neuropsychologia.2008.08.015
6. C. Creatore, S. Sabathiel, and T. Solstad, “Learning exact enumeration and approximate estimation in deep neural network models,” Cognition, vol. 215, p. 104815, 2021. http://dx.doi.org/10.1016/j.cognition.2021.104815
7. S. Guan and M. Loew, “Understanding the ability of deep neural networks to count connected components in images,” in 2020 IEEE Applied Imagery Pattern Recognition Workshop (AIPR), 2020. http://dx.doi.org/10.1109/AIPR50011.2020.9425331 pp. 1–7.
8. A. Śluzek, “Counting dots: On learning numerical concepts from visual data,” Proceedings of the 3rd Polish Conference on Artificial Intelligence, April 2022, Gdynia, Poland, pp. 16–19, 2022.
9. M. Fang, Z. Zhou, S. Chen, and J. L. McClelland, “Can a recurrent neural network learn to count things?” Cognitive Science, 2018.
10. A. Cope, E. Vasilaki, D. Minors, C. Sabo, J. Marshall, and A. Barron, “Abstract concept learning in a simple neural network inspired by the insect brain,” PLoS Computational Biology, vol. 14, no. 9, p. e1006435, 2018. http://dx.doi.org/10.1371/journal.pcbi.1006435
11. M. Tomonaga and T. Matsuzawa, “Enumeration of briefly presented items by the chimpanzee (pan troglodytes) and humans (homo sapiens),” Animal Learning & Behavior, vol. 30, p. 143âĂŞ157, 2002. http://dx.doi.org/10.3758/BF03192916
12. K. Wynn, “Children’s understanding of counting,” Cognition, vol. 36, no. 2, pp. 155–193, 1990. http://dx.doi.org/10.1016/0010-0277(90)90003-3