{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,12,31]],"date-time":"2025-12-31T16:15:23Z","timestamp":1767197723306,"version":"3.48.0"},"reference-count":35,"publisher":"MDPI AG","issue":"1","license":[{"start":{"date-parts":[[2025,12,31]],"date-time":"2025-12-31T00:00:00Z","timestamp":1767139200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100018537","name":"National Science and Technology Major Project","doi-asserted-by":"publisher","award":["2021ZD0201302"],"award-info":[{"award-number":["2021ZD0201302"]}],"id":[{"id":"10.13039\/501100018537","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Zhongguancun Laboratory"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Entropy"],"abstract":"<jats:p>Robustness to distributional shifts remains a critical limitation for deploying deep neural networks (DNNs) in real-world applications. While DNNs excel in standard benchmarks, their performance often deteriorates under unseen or perturbed conditions. Understanding how internal information representations relate to such robustness remains underexplored. In this work, we propose an interpretable framework for robustness assessment based on partial information decomposition (PID), which quantifies how neurons redundantly, uniquely, or synergistically encode task-relevant information. Analysis of PID measures computed from clean inputs reveals that models characterized by higher redundancy rates and lower synergy rates tend to maintain more stable performance under various natural corruptions. Additionally, a higher rate of unique information is positively associated with improved classification accuracy on the data from which the measure is computed. These findings provide new insights for understanding and comparing model behavior through internal information analysis, and highlight the feasibility of lightweight robustness assessment without requiring extensive access to corrupted data.<\/jats:p>","DOI":"10.3390\/e28010050","type":"journal-article","created":{"date-parts":[[2025,12,31]],"date-time":"2025-12-31T16:08:00Z","timestamp":1767197280000},"page":"50","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Interpreting Performance of Deep Neural Networks with Partial Information Decomposition"],"prefix":"10.3390","volume":"28","author":[{"ORCID":"https:\/\/orcid.org\/0009-0004-3200-1005","authenticated-orcid":false,"given":"Tianyue","family":"Liu","sequence":"first","affiliation":[{"name":"School of Artificial Intelligence, Beihang University, Beijing 100191, China"},{"name":"Zhongguancun Laboratory, Beijing 100194, China"},{"name":"LMIB and SKLCCSE, Beihang University, Beijing 100191, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Binghui","family":"Guo","sequence":"additional","affiliation":[{"name":"School of Artificial Intelligence, Beihang University, Beijing 100191, China"},{"name":"Zhongguancun Laboratory, Beijing 100194, China"},{"name":"LMIB and SKLCCSE, Beihang University, Beijing 100191, China"},{"name":"Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing, Beijing 100083, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4340-5414","authenticated-orcid":false,"given":"Ziqiao","family":"Yin","sequence":"additional","affiliation":[{"name":"School of Artificial Intelligence, Beihang University, Beijing 100191, China"},{"name":"Zhongguancun Laboratory, Beijing 100194, China"},{"name":"LMIB and SKLCCSE, Beihang University, Beijing 100191, China"},{"name":"Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing, Beijing 100083, China"},{"name":"Hangzhou Internation Innovation Institute of Beihang University, Hangzhou 311115, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhilong","family":"Mi","sequence":"additional","affiliation":[{"name":"School of Artificial Intelligence, Beihang University, Beijing 100191, China"},{"name":"LMIB and SKLCCSE, Beihang University, Beijing 100191, China"},{"name":"Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing, Beijing 100083, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Donghui","family":"Jin","sequence":"additional","affiliation":[{"name":"School of Artificial Intelligence, Beihang University, Beijing 100191, China"},{"name":"LMIB and SKLCCSE, Beihang University, Beijing 100191, China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1968","published-online":{"date-parts":[[2025,12,31]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"2352","DOI":"10.1162\/neco_a_00990","article-title":"Deep convolutional neural networks for image classification: A comprehensive review","volume":"29","author":"Rawat","year":"2017","journal-title":"Neural Comput."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3158369","article-title":"Deep learning based recommender system: A survey and new perspectives","volume":"52","author":"Zhang","year":"2019","journal-title":"ACM Comput. Surv. (CSUR)"},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"55","DOI":"10.1109\/MCI.2018.2840738","article-title":"Recent trends in deep learning based natural language processing","volume":"13","author":"Young","year":"2018","journal-title":"IEEE Comput. Intell. Mag."},{"key":"ref_4","doi-asserted-by":"crossref","unstructured":"Torralba, A., and Efros, A.A. (2011). Unbiased look at dataset bias. CVPR 2011, IEEE.","DOI":"10.1109\/CVPR.2011.5995347"},{"key":"ref_5","unstructured":"Hendrycks, D., and Gimpel, K. (2017, January 24\u201326). A baseline for detecting misclassified and out-of-distribution examples in neural networks. Proceedings of the International Conference on Learning Representations, Toulon, France."},{"key":"ref_6","doi-asserted-by":"crossref","unstructured":"Dong, Y., Kang, C., Zhang, J., Zhu, Z., Wang, Y., Yang, X., Su, H., Wei, X., and Zhu, J. (2023, January 17\u201324). Benchmarking robustness of 3d object detection to common corruptions. Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, BC, Canada.","DOI":"10.1109\/CVPR52729.2023.00105"},{"key":"ref_7","doi-asserted-by":"crossref","unstructured":"Rusak, E., Schott, L., Zimmermann, R.S., Bitterwolf, J., Bringmann, O., Bethge, M., and Brendel, W. (2020). A simple way to make neural networks robust against diverse image corruptions. Computer Vision\u2013ECCV 2020, Proceedings of the 16th European Conference, Glasgow, UK, 23\u201328 August 2020, Proceedings, Part III 16, Springer.","DOI":"10.1007\/978-3-030-58580-8_4"},{"key":"ref_8","unstructured":"Geirhos, R., Temme, C.R., Rauber, J., Sch\u00fctt, H.H., Bethge, M., and Wichmann, A.F. (2018). Generalisation in humans and deep neural networks. arXiv."},{"key":"ref_9","first-page":"11539","article-title":"Improving robustness against common corruptions by covariate shift adaptation","volume":"33","author":"Schneider","year":"2020","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"ref_10","unstructured":"Chun, S., Oh, S.J., Yun, S., Han, D., Choe, J., and Yoo, Y. (2020). An empirical evaluation on robustness and uncertainty of regularization methods. arXiv."},{"key":"ref_11","doi-asserted-by":"crossref","unstructured":"Yu, S., Giraldo, L.G.S., and Pr\u00edncipe, J.C. (2021, January 19\u201327). Information-theoretic methods in deep neural networks: Recent advances and emerging opportunities. Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence (IJCAI-21), Virtual.","DOI":"10.24963\/ijcai.2021\/633"},{"key":"ref_12","doi-asserted-by":"crossref","unstructured":"Tishby, N., and Zaslavsky, N. (2015, January 11\u201315). Deep learning and the information bottleneck principle. Proceedings of the 2015 IEEE Information Theory Workshop (itw), Jeju Island, Republic of Korea.","DOI":"10.1109\/ITW.2015.7133169"},{"key":"ref_13","unstructured":"Shwartz-Ziv, R., and Tishby, N. (2017). Opening the black box of deep neural networks via information. arXiv."},{"key":"ref_14","doi-asserted-by":"crossref","unstructured":"Gabri\u00e9, M., Manoel, A., Luneau, C., Macris, N., Krzakala, F., and Zdeborov\u00e1, L. (2018). Entropy and mutual information in models of deep neural networks. arXiv.","DOI":"10.1088\/1742-5468\/ab3430"},{"key":"ref_15","unstructured":"Williams, P.L., and Beer, R.D. (2010). Nonnegative decomposition of multivariate information. arXiv."},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"352","DOI":"10.1016\/j.tics.2023.11.005","article-title":"Information decomposition and the informational architecture of the brain","volume":"28","author":"Luppi","year":"2024","journal-title":"Trends Cogn. Sci."},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"e2300888120","DOI":"10.1073\/pnas.2300888120","article-title":"Partial entropy decomposition reveals higher-order information structures in human brain activity","volume":"120","author":"Varley","year":"2023","journal-title":"Proc. Natl. Acad. Sci. USA"},{"key":"ref_18","first-page":"1","article-title":"A rigorous information-theoretic definition of redundancy and relevancy in feature selection based on (partial) information decomposition","volume":"24","author":"Wollstadt","year":"2023","journal-title":"J. Mach. Learn. Res."},{"key":"ref_19","doi-asserted-by":"crossref","first-page":"25","DOI":"10.1016\/j.bandc.2015.09.004","article-title":"Partial information decomposition as a unified approach to the specification of neural goal functions","volume":"112","author":"Wibral","year":"2017","journal-title":"Brain Cogn."},{"key":"ref_20","first-page":"2045","article-title":"Diffusion pid: Interpreting diffusion via partial information decomposition","volume":"37","author":"Dewan","year":"2024","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Proca, A.M., Rosas, F.E., Luppi, A.I., Bor, D., Crosby, M., and Mediano, P.A. (2024). Synergistic information supports modality integration and flexible learning in neural networks solving multiple tasks. Plos Comput. Biol., 20.","DOI":"10.1371\/journal.pcbi.1012178"},{"key":"ref_22","doi-asserted-by":"crossref","unstructured":"Tax, T.M., Mediano, P.A., and Shanahan, M. (2017). The partial information decomposition of generative neural network models. Entropy, 19.","DOI":"10.3390\/e19090474"},{"key":"ref_23","unstructured":"Rolnick, D., and Tegmark, M. (May, January 30). The power of deeper networks for expressing natural functions. Proceedings of the International Conference on Learning Representations, Vancouver, BC, Canada."},{"key":"ref_24","doi-asserted-by":"crossref","first-page":"379","DOI":"10.1002\/j.1538-7305.1948.tb01338.x","article-title":"A mathematical theory of communication","volume":"27","author":"Shannon","year":"1948","journal-title":"Bell Syst. Tech. J."},{"key":"ref_25","doi-asserted-by":"crossref","first-page":"052802","DOI":"10.1103\/PhysRevE.91.052802","article-title":"Exploration of synergistic and redundant information sharing in static and dynamical gaussian systems","volume":"91","author":"Barrett","year":"2015","journal-title":"Phys. Rev. E"},{"key":"ref_26","unstructured":"Hendrycks, D., and Dietterich, T. (2019, January 6\u20139). Benchmarking neural network robustness to common corruptions and perturbations. Proceedings of the International Conference on Learning Representations, New Orleans, LA, USA."},{"key":"ref_27","doi-asserted-by":"crossref","unstructured":"Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., and Chen, L.-C. (2018, January 18\u201323). MobilenetV2: Inverted residuals and linear bottlenecks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA.","DOI":"10.1109\/CVPR.2018.00474"},{"key":"ref_28","unstructured":"Krizhevsky, A., Sutskever, I., and Hinton, G.E. (2012). Imagenet classification with deep convolutional neural networks. Adv. Neural Inf. Process. Syst., 25, Available online: https:\/\/proceedings.neurips.cc\/paper_files\/paper\/2012\/file\/c399862d3b9d6b76c8436e924a68c45b-Paper.pdf."},{"key":"ref_29","unstructured":"Simonyan, K., and Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv."},{"key":"ref_30","doi-asserted-by":"crossref","unstructured":"Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., and Wojna, Z. (2016, January 27\u201330). Rethinking the inception architecture for computer vision. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA.","DOI":"10.1109\/CVPR.2016.308"},{"key":"ref_31","doi-asserted-by":"crossref","unstructured":"He, K., Zhang, X., Ren, S., and Sun, J. (2016, January 27\u201330). Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA.","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref_32","doi-asserted-by":"crossref","unstructured":"Huang, G., Liu, Z., Maaten, L.V.D., and Weinberger, K.Q. (2017, January 21\u201326). Densely connected convolutional networks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA.","DOI":"10.1109\/CVPR.2017.243"},{"key":"ref_33","doi-asserted-by":"crossref","first-page":"1291","DOI":"10.1109\/TIP.2020.3042083","article-title":"Interpreting and improving adversarial robustness of deep neural networks with neuron sensitivity","volume":"30","author":"Zhang","year":"2020","journal-title":"IEEE Trans. Image Process."},{"key":"ref_34","doi-asserted-by":"crossref","first-page":"11539","DOI":"10.1523\/JNEUROSCI.23-37-11539.2003","article-title":"Synergy, redundancy, and independence in population codes","volume":"23","author":"Schneidman","year":"2003","journal-title":"J. Neurosci."},{"key":"ref_35","doi-asserted-by":"crossref","unstructured":"Venkatesh, P., Gurushankar, K., and Schamberg, G. (2023, January 25\u201330). Capturing and interpreting unique information. Proceedings of the 2023 IEEE International Symposium on Information Theory (ISIT), Taipei, Taiwan.","DOI":"10.1109\/ISIT54713.2023.10206597"}],"container-title":["Entropy"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1099-4300\/28\/1\/50\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,12,31]],"date-time":"2025-12-31T16:12:07Z","timestamp":1767197527000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1099-4300\/28\/1\/50"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,12,31]]},"references-count":35,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,1]]}},"alternative-id":["e28010050"],"URL":"https:\/\/doi.org\/10.3390\/e28010050","relation":{},"ISSN":["1099-4300"],"issn-type":[{"value":"1099-4300","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,12,31]]}}}