Multiview descriptor of neural activations, through the fusion of functional graphs and persistent homology
DOI:
https://doi.org/10.70577/cieninter.v4i4.162Keywords:
Deep learning, Interpretability, Multi-view learning, Persistent homology, Topological data analysis.Abstract
This study proposes a multiview descriptor of the internal activations of neural networks: a fixed-length set of numbers that summarizes a network by combining two different descriptions of the same object, one drawn from graph-theoretic measures of the connectivity among neurons and the other from persistent homology, which captures the shape of the activation cloud. Its validation used forty neural networks of two types, multilayer perceptrons and convolutional networks, arranged in five groups and trained on three image datasets. The independent variable is the group of each network; the dependent variables are the numbers the descriptor computes for it (ten graph measures, twelve persistence statistics and their fusion), none of which uses accuracy. The fusion yielded highly significant results for operation and prediction, agreement, and performance, as assessed by cross-validation (0.949, 0.957 and 0.972 on the three datasets), non-parametric tests (29 of 30 tests significant) and ROC analysis (0.949 with the fusion, 0.854 and 0.869 with each view and 0.754 with a baseline), respectively. Because the two network types differ widely in accuracy, two further checks, one restricted to the multilayer perceptrons and one that equalized accuracy within each type, showed that the descriptor's advantage does not rest on that difference.
Downloads
References
Alain, G., y Bengio, Y. (2016). Understanding intermediate layers using linear classifier probes. arXiv preprint arXiv:1610.01644. https://doi.org/10.48550/arXiv.1610.01644
Arlot, S., y Celisse, A. (2010). A survey of cross-validation procedures for model selection. Statistics Surveys, 4, 40–79. https://doi.org/10.1214/09-SS054 DOI: https://doi.org/10.1214/09-SS054
Ballester, R., Arnal Clemente, X., Casacuberta, C., Madadi, M., Corneanu, C. A., y Escalera, S. (2024). Predicting the generalization gap in neural networks using topological data analysis. Neurocomputing, 596, 127787. https://doi.org/10.1016/j.neucom.2024.127787 DOI: https://doi.org/10.1016/j.neucom.2024.127787
Ballester, R., Casacuberta, C., y Escalera, S. (2026). Topological data analysis for neural networks. Springer. https://doi.org/10.1007/978-3-032-08283-1 DOI: https://doi.org/10.1007/978-3-032-08283-1
Baltrusaitis, T., Ahuja, C., y Morency, L. P. (2019). Multimodal machine learning: A survey and taxonomy. IEEE Transactions on Pattern Analysis and Machine Intelligence, 41(2), 423–443. https://doi.org/10.1109/TPAMI.2018.2798607 DOI: https://doi.org/10.1109/TPAMI.2018.2798607
Bauer, U. (2021). Ripser: efficient computation of Vietoris–Rips persistence barcodes. Journal of Applied and Computational Topology, 5, 391–423. https://doi.org/10.1007/s41468-021-00071-5 DOI: https://doi.org/10.1007/s41468-021-00071-5
Benjamini, Y., y Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B, 57(1), 289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x DOI: https://doi.org/10.1111/j.2517-6161.1995.tb02031.x
Bullmore, E., y Sporns, O. (2009). Complex brain networks: graph theoretical analysis of structural and functional systems. Nature Reviews Neuroscience, 10(3), 186–198. https://doi.org/10.1038/nrn2575 DOI: https://doi.org/10.1038/nrn2575
Chazal, F., y Michel, B. (2021). An introduction to topological data analysis: fundamental and practical aspects for data scientists. Frontiers in Artificial Intelligence, 4, 667963. https://doi.org/10.3389/frai.2021.667963 DOI: https://doi.org/10.3389/frai.2021.667963
Chen, Z., Fu, L., Yao, J., Guo, W., Plant, C., y Wang, S. (2023). Learnable graph convolutional network and feature fusion for multi-view learning. Information Fusion, 95, 109–119. https://doi.org/10.1016/j.inffus.2023.02.013 DOI: https://doi.org/10.1016/j.inffus.2023.02.013
Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2.ª ed.). Lawrence Erlbaum Associates. https://doi.org/10.4324/9780203771587 DOI: https://doi.org/10.4324/9780203771587
Corneanu, C. A., Escalera, S., y Martinez, A. M. (2020). Computing the testing error without a testing set. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2674–2682. https://doi.org/10.1109/CVPR42600.2020.00275 DOI: https://doi.org/10.1109/CVPR42600.2020.00275
Demsar, J. (2006). Statistical comparisons of classifiers over multiple data sets. Journal of Machine Learning Research, 7, 1–30. https://jmlr.org/papers/v7/demsar06a.html
Edelsbrunner, H., Letscher, D., y Zomorodian, A. (2002). Topological persistence and simplification. Discrete & Computational Geometry, 28(4), 511–533. https://doi.org/10.1007/s00454-002-2885-2 DOI: https://doi.org/10.1007/s00454-002-2885-2
García, S., Fernández, A., Luengo, J., y Herrera, F. (2010). Advanced nonparametric tests for multiple comparisons in the design of experiments in computational intelligence and data mining: Experimental analysis of power. Information Sciences, 180(10), 2044–2064. https://doi.org/10.1016/j.ins.2009.12.010 DOI: https://doi.org/10.1016/j.ins.2009.12.010
Hao, P., Gao, W., y Hu, L. (2025). Embedded feature fusion for multi-view multi-label feature selection. Pattern Recognition, 157, 110888. https://doi.org/10.1016/j.patcog.2024.110888 DOI: https://doi.org/10.1016/j.patcog.2024.110888
Hoerl, A. E., y Kennard, R. W. (1970). Ridge Regression: Biased Estimation for Nonorthogonal Problems. Technometrics, 12(1), 55–67. https://doi.org/10.1080/00401706.1970.10488634 DOI: https://doi.org/10.1080/00401706.1970.10488634
Horta, V. A. C., Tiddi, I., Little, S., y Mileo, A. (2021). Extracting knowledge from deep neural networks through graph analysis. Future Generation Computer Systems, 120, 109–118. https://doi.org/10.1016/j.future.2021.02.009 DOI: https://doi.org/10.1016/j.future.2021.02.009
Kendall, M. G. (1938). A new measure of rank correlation. Biometrika, 30(1/2), 81–93. https://doi.org/10.2307/2332226 DOI: https://doi.org/10.1093/biomet/30.1-2.81
Klabunde, M., Schumacher, T., Strohmaier, M., y Lemmerich, F. (2025). Similarity of neural network models: A survey of functional and representational measures. ACM Computing Surveys, 57(9), 1–52. https://doi.org/10.1145/3728458 DOI: https://doi.org/10.1145/3728458
Krizhevsky, A. (2009). Learning Multiple Layers of Features from Tiny Images. https://www.cs.toronto.edu/~kriz/learning-features-2009-TR.pdf
Liang, P. P., Zadeh, A., y Morency, L. P. (2024). Foundations & trends in multimodal machine learning: Principles, challenges, and open questions. ACM Computing Surveys, 56(10), 1–42. https://doi.org/10.1145/3656580 DOI: https://doi.org/10.1145/3656580
Loshchilov, I., y Hutter, F. (2019). Decoupled weight decay regularization. International Conference on Learning Representations (ICLR). https://openreview.net/forum?id=Bkg6RiCqY7
Mann, H. B., y Whitney, D. R. (1947). On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics, 18(1), 50–60. https://doi.org/10.1214/aoms/1177730491 DOI: https://doi.org/10.1214/aoms/1177730491
Mellor, J., Turner, J., Storkey, A., y Crowley, E. J. (2021). Neural architecture search without training. Proceedings of the 38th International Conference on Machine Learning (ICML), 7588–7598. https://proceedings.mlr.press/v139/mellor21a.html
Moor, M., Horn, M., Rieck, B., y Borgwardt, K. (2020). Topological autoencoders. Proceedings of the 37th International Conference on Machine Learning (ICML), 7045–7054. https://proceedings.mlr.press/v119/moor20a.html
Naitzat, G., Zhitnikov, A., y Lim, L. H. (2020). Topology of deep neural networks. Journal of Machine Learning Research, 21(184), 1–40. https://jmlr.org/papers/v21/20-345.html
Netzer, Y., Wang, T., Coates, A., Bissacco, A., Wu, B., y Ng, A. Y. (2011). Reading digits in natural images with unsupervised feature learning. NIPS Workshop on Deep Learning and Unsupervised Feature Learning. http://ufldl.stanford.edu/housenumbers/nips2011_housenumbers.pdf
Purvine, E., Brown, D., Jefferson, B., Joslyn, C., Praggastis, B., Rathore, A., Shapiro, M., Wang, B., y Zhou, Y. (2023). Experimental observations of the topology of convolutional neural network activations. Proceedings of the AAAI Conference on Artificial Intelligence, 37(8), 9470–9479. https://doi.org/10.1609/aaai.v37i8.26134 DOI: https://doi.org/10.1609/aaai.v37i8.26134
Raghu, M., Gilmer, J., Yosinski, J., y Sohl-Dickstein, J. (2017). SVCCA: Singular vector canonical correlation analysis for deep learning dynamics and interpretability. Advances in Neural Information Processing Systems (NeurIPS), 6076–6085. https://proceedings.neurips.cc/paper/2017/hash/dc6a7e655d7e5840e66733e9ee67cc69-Abstract.html
Recht, B., Roelofs, R., Schmidt, L., y Shankar, V. (2019). Do ImageNet classifiers generalize to ImageNet?. Proceedings of the 36th International Conference on Machine Learning (ICML), 5389–5400. https://proceedings.mlr.press/v97/recht19a.html
Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., y Batra, D. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. Proceedings of the IEEE International Conference on Computer Vision (ICCV), 618–626. https://doi.org/10.1109/ICCV.2017.74 DOI: https://doi.org/10.1109/ICCV.2017.74
Simonyan, K., Vedaldi, A., y Zisserman, A. (2013). Deep inside convolutional networks: Visualising image classification models and saliency maps. arXiv preprint arXiv:1312.6034. https://doi.org/10.48550/arXiv.1312.6034
Songdechakraiwut, T., y Wu, Y. (2025). Functional Connectomes of Neural Networks. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI-25), 20558–20566. https://doi.org/10.1609/aaai.v39i19.34265 DOI: https://doi.org/10.1609/aaai.v39i19.34265
Sun, S. (2013). A survey of multi-view machine learning. Neural Computing and Applications, 23(7–8), 2031–2038. https://doi.org/10.1007/s00521-013-1362-6 DOI: https://doi.org/10.1007/s00521-013-1362-6
Zhang, B., Dong, Z., Zhang, J., y Lin, H. (2023). Functional network: A novel framework for interpretability of deep neural networks. Neurocomputing, 519, 94–103. https://doi.org/10.1016/j.neucom.2022.11.035 DOI: https://doi.org/10.1016/j.neucom.2022.11.035
Zhang, B., He, Z., y Lin, H. (2024). A comprehensive review of deep neural network interpretation using topological data analysis. Neurocomputing, 609, 128513. https://doi.org/10.1016/j.neucom.2024.128513 DOI: https://doi.org/10.1016/j.neucom.2024.128513
Zhang, B., y Lin, H. (2024). Functional loops: Monitoring functional organization of deep neural networks using algebraic topology. Neural Networks, 174, 106239. https://doi.org/10.1016/j.neunet.2024.106239 DOI: https://doi.org/10.1016/j.neunet.2024.106239
Zhao, J., Xie, X., Xu, X., y Sun, S. (2017). Multi-view learning overview: Recent progress and new challenges. Information Fusion, 38, 43–54. https://doi.org/10.1016/j.inffus.2017.02.007 DOI: https://doi.org/10.1016/j.inffus.2017.02.007
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 César Daniel Reinoso Reinoso, Amalia Isabel Escudero Villa

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.








