XAI-Driven Loss Optimization via Activation Map Alignment in Multi-Task CNNs: A Systematic Literature Review
Resumo
Este artigo apresenta uma revisão sistemática da literatura sobre a integração de técnicas de XAI em funções de perda de CNNs. Seguindo as diretrizes de Kitchenham e Charters, foram analisados estudos publicados entre 2020 e 2026 nas bases ACM Digital Library, IEEE Xplore e Scopus. Os trabalhos selecionados foram avaliados quanto aos seus mecanismos de alinhamento, sinais de supervisão, estratégias de otimização e desempenho. Os resultados indicam uma transição da interpretabilidade passiva para abordagens de otimização guiadas por explicações. A revisão também destaca desafios relacionados ao custo computacional, à sensibilidade, à hiperparametrização e ao equilíbrio entre a restrição da atenção do modelo e a preservação de informações contextuais.Referências
Amyar, A., Modzelewski, R., Vera, P., Morard, V., and Ruan, S. (2022). Weakly supervised tumor detection in pet using class response for treatment outcome prediction. Journal of Imaging, 8(5):130.
Aravinda, C., Shaffi, N., Nair, S. S. K., Khan, M. S., Viswan, V., and Arjun, B. (2026). More-xai: Moment-based rigid registration and multi-perspective explainability for fracture localization in x-ray imaging. IEEE Access.
Caruana, R. (1997). Multitask learning. Machine learning, 28(1):41–75.
Dai, G., Wang, C., Tang, Q., Zhang, Y., Dai, D., Qiao, L., Yan, J., and Chen, H. (2025). Interpretable breast cancer diagnosis using histopathology and lesion mask as domain concepts conditional simulation ultrasonography. Information Fusion, 123:103343.
de Araujo Morais, L. R., de Melo Fragoso, G. A., Ludermir, T. B., and Monteiro, C. L. A. (2024). Explainable ai for the brazilian stock market index: A post-hoc approach to deep learning models in time-series forecasting. In Encontro Nacional de Inteligência Artificial e Computacional (ENIAC), pages 436–447. SBC.
de Araujo Morais, L. R. and Ludermir, T. (2026). Explainability and uncertainty for time-series classification in cryptocurrency data: a hybrid xai methodology based on comte and leftist. International Journal of Data Science and Analytics, 22(1):127.
Hu, W. (2025). Instance-level human parsing based on semantic-edge fusion in crowded scenes. In 2025 7th International Conference on Robotics, Intelligent Control and Artificial Intelligence (RICAI), pages 1095–1098. IEEE.
Kitchenham, B., Charters, S., et al. (2007). Guidelines for performing systematic literature reviews in software engineering.
Li, W., Yang, W., Zhang, W., Liu, T., Liu, Y., and Liu, L. (2023). Hierarchical disentanglement-alignment network for robust sar vehicle recognition. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 16:9661–9679.
Liu, L., Tong, Z., Cai, Z., Wu, H., Zhang, R., Le Bris, A., and Olteanu-Raimond, A.-M. (2024). Hieru-net: A hierarchical semantic segmentation method for land cover mapping. IEEE Transactions on Geoscience and Remote Sensing, 62:1–14.
Ludermir, T. B. (2021). Inteligência artificial e aprendizado de máquina: estado atual e tendências. Estudos avançados, 35:85–94.
Misra, I., Shrivastava, A., Gupta, A., and Hebert, M. (2016). Cross-stitch networks for multi-task learning. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3994–4003.
Mukherjee, S., Templeton, K., Tindimwebwa, S., Lin, I., Sutin, J., Yu, M., Peterson, M., Truwit, C., Schiff, S. J., and Monga, V. (2025). Glapal-h: Global, local, and parts aware learner for hydrocephalus infection diagnosis in low-field mri. IEEE Transactions on Biomedical Engineering.
Rieger, L., Singh, C., Murdoch, W., and Yu, B. (2020). Interpretations are useful: penalizing explanations to align neural networks with prior knowledge. In International conference on machine learning, pages 8116–8126. PMLR.
Ross, A. S., Hughes, M. C., and Doshi-Velez, F. (2017). Right for the right reasons: Training differentiable models by constraining their explanations. arXiv preprint arXiv:1703.03717.
Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., and Batra, D. (2017). Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision, pages 618–626.
Shi, L. (2025). Medically explainable deep learning-based glaucoma diagnosis. In 2025 47th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 1–6. IEEE.
Xu, G., Feng, C., Guo, X., Zhu, Z., and Wang, W. (2025). A joint learning framework for bridging defect prediction and interpretation. IEEE Transactions on Reliability.
Yu, T. and Liu, J. (2024). Multi-task learning of visual attributes for image aesthetics assessment. In 2024 IEEE 36th International Conference on Tools with Artificial Intelligence (ICTAI), pages 350–354. IEEE.
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., and Torralba, A. (2016). Learning deep features for discriminative localization. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2921–2929.
Zhu, H., Salcudean, S., and Rohling, R. (2022). Gaze-guided class activation mapping: Leverage human visual attention for network attention in chest x-rays classification. In Proceedings of the 15th International Symposium on Visual Information Communication and Interaction, pages 1–8.
Aravinda, C., Shaffi, N., Nair, S. S. K., Khan, M. S., Viswan, V., and Arjun, B. (2026). More-xai: Moment-based rigid registration and multi-perspective explainability for fracture localization in x-ray imaging. IEEE Access.
Caruana, R. (1997). Multitask learning. Machine learning, 28(1):41–75.
Dai, G., Wang, C., Tang, Q., Zhang, Y., Dai, D., Qiao, L., Yan, J., and Chen, H. (2025). Interpretable breast cancer diagnosis using histopathology and lesion mask as domain concepts conditional simulation ultrasonography. Information Fusion, 123:103343.
de Araujo Morais, L. R., de Melo Fragoso, G. A., Ludermir, T. B., and Monteiro, C. L. A. (2024). Explainable ai for the brazilian stock market index: A post-hoc approach to deep learning models in time-series forecasting. In Encontro Nacional de Inteligência Artificial e Computacional (ENIAC), pages 436–447. SBC.
de Araujo Morais, L. R. and Ludermir, T. (2026). Explainability and uncertainty for time-series classification in cryptocurrency data: a hybrid xai methodology based on comte and leftist. International Journal of Data Science and Analytics, 22(1):127.
Hu, W. (2025). Instance-level human parsing based on semantic-edge fusion in crowded scenes. In 2025 7th International Conference on Robotics, Intelligent Control and Artificial Intelligence (RICAI), pages 1095–1098. IEEE.
Kitchenham, B., Charters, S., et al. (2007). Guidelines for performing systematic literature reviews in software engineering.
Li, W., Yang, W., Zhang, W., Liu, T., Liu, Y., and Liu, L. (2023). Hierarchical disentanglement-alignment network for robust sar vehicle recognition. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 16:9661–9679.
Liu, L., Tong, Z., Cai, Z., Wu, H., Zhang, R., Le Bris, A., and Olteanu-Raimond, A.-M. (2024). Hieru-net: A hierarchical semantic segmentation method for land cover mapping. IEEE Transactions on Geoscience and Remote Sensing, 62:1–14.
Ludermir, T. B. (2021). Inteligência artificial e aprendizado de máquina: estado atual e tendências. Estudos avançados, 35:85–94.
Misra, I., Shrivastava, A., Gupta, A., and Hebert, M. (2016). Cross-stitch networks for multi-task learning. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3994–4003.
Mukherjee, S., Templeton, K., Tindimwebwa, S., Lin, I., Sutin, J., Yu, M., Peterson, M., Truwit, C., Schiff, S. J., and Monga, V. (2025). Glapal-h: Global, local, and parts aware learner for hydrocephalus infection diagnosis in low-field mri. IEEE Transactions on Biomedical Engineering.
Rieger, L., Singh, C., Murdoch, W., and Yu, B. (2020). Interpretations are useful: penalizing explanations to align neural networks with prior knowledge. In International conference on machine learning, pages 8116–8126. PMLR.
Ross, A. S., Hughes, M. C., and Doshi-Velez, F. (2017). Right for the right reasons: Training differentiable models by constraining their explanations. arXiv preprint arXiv:1703.03717.
Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., and Batra, D. (2017). Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision, pages 618–626.
Shi, L. (2025). Medically explainable deep learning-based glaucoma diagnosis. In 2025 47th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 1–6. IEEE.
Xu, G., Feng, C., Guo, X., Zhu, Z., and Wang, W. (2025). A joint learning framework for bridging defect prediction and interpretation. IEEE Transactions on Reliability.
Yu, T. and Liu, J. (2024). Multi-task learning of visual attributes for image aesthetics assessment. In 2024 IEEE 36th International Conference on Tools with Artificial Intelligence (ICTAI), pages 350–354. IEEE.
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., and Torralba, A. (2016). Learning deep features for discriminative localization. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2921–2929.
Zhu, H., Salcudean, S., and Rohling, R. (2022). Gaze-guided class activation mapping: Leverage human visual attention for network attention in chest x-rays classification. In Proceedings of the 15th International Symposium on Visual Information Communication and Interaction, pages 1–8.
Publicado
19/10/2026
Como Citar
FRAGOSO, Gabriel; LUDERMIR, Teresa; FRAGOSO, Júlia; CALIXTO, Diego; VITAL, Gabriel.
XAI-Driven Loss Optimization via Activation Map Alignment in Multi-Task CNNs: A Systematic Literature Review. In: ENCONTRO NACIONAL DE INTELIGÊNCIA ARTIFICIAL E COMPUTACIONAL (ENIAC), 23. , 2026, Cuiabá/MT.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 1628-1639.
ISSN 2763-9061.
DOI: https://doi.org/10.5753/eniac.2026.31260.
