%0 Journal Article %1 KUMAR2024107900 %A Kumar, Mohit %A Singh, Sukhvir %A Bowles, Juliana %D 2024 %J Engineering Applications of Artificial Intelligence %K bayesian deep-learning eaai fuzzy mohit neural-networks %P 107900 %R https://doi.org/10.1016/j.engappai.2024.107900 %T Variational Bayesian deep fuzzy models for interpretable classification %U https://www.sciencedirect.com/science/article/pii/S0952197624000587 %V 132 %X This paper makes a direct contribution to the mathematical theory and practical implementation of interpretable deep fuzzy learning whilst addressing the model interpretability challenge posed by the typical non-interpretable nature of training data. A deep autoencoder, composed of a finite number of stochastic fuzzy filters, is learned using variational Bayes for an efficient representation of high-dimensional feature vectors at different levels of abstraction. The robustness of the deep model towards data outliers is achieved by incorporating heavy-tailed distributed noises in the inference mechanism. The proposed deep stochastic fuzzy autoencoder makes it possible to have an interpretable classification by means of an induced mapping from a non-interpretable feature space to an interpretable parameter space. The mapping induced by our autoencoder can be used to explain the classifier’s predictions by interpreting the non-interpretable high-dimensional feature vectors. The experimental results obtained on various benchmark datasets (e.g., Freiburg Groceries, Caltech-101, Caltech-256, Adobe Panoramas, and MNIST) show that our proposed approach not only outperforms widely used machine learning methods in classification tasks but brings the added value of interpretability as a further advantage.

@article{KUMAR2024107900, abstract = {This paper makes a direct contribution to the mathematical theory and practical implementation of interpretable deep fuzzy learning whilst addressing the model interpretability challenge posed by the typical non-interpretable nature of training data. A deep autoencoder, composed of a finite number of stochastic fuzzy filters, is learned using variational Bayes for an efficient representation of high-dimensional feature vectors at different levels of abstraction. The robustness of the deep model towards data outliers is achieved by incorporating heavy-tailed distributed noises in the inference mechanism. The proposed deep stochastic fuzzy autoencoder makes it possible to have an interpretable classification by means of an induced mapping from a non-interpretable feature space to an interpretable parameter space. The mapping induced by our autoencoder can be used to explain the classifier’s predictions by interpreting the non-interpretable high-dimensional feature vectors. The experimental results obtained on various benchmark datasets (e.g., Freiburg Groceries, Caltech-101, Caltech-256, Adobe Panoramas, and MNIST) show that our proposed approach not only outperforms widely used machine learning methods in classification tasks but brings the added value of interpretability as a further advantage.}, added-at = {2024-03-27T10:28:59.000+0100}, author = {Kumar, Mohit and Singh, Sukhvir and Bowles, Juliana}, biburl = {https://www.bibsonomy.org/bibtex/22a7ad74381bf4484a09a864a5fcff4e3/scch}, doi = {https://doi.org/10.1016/j.engappai.2024.107900}, interhash = {5bb5badaedbc214fdc7ddaa257fd7707}, intrahash = {2a7ad74381bf4484a09a864a5fcff4e3}, issn = {0952-1976}, journal = {Engineering Applications of Artificial Intelligence}, keywords = {bayesian deep-learning eaai fuzzy mohit neural-networks}, pages = 107900, timestamp = {2024-03-27T10:28:59.000+0100}, title = {Variational Bayesian deep fuzzy models for interpretable classification}, url = {https://www.sciencedirect.com/science/article/pii/S0952197624000587}, volume = 132, year = 2024 }