<abstract xmlns="http://www.w3.org/1999/xhtml"><p>A novel technique for deep learning of image classifiers is presented. The learned CNN models higher offer better separation of deep features (also known as embedded vectors) measured by Euclidean proximity and also no deterioration of the classification results by class membership probability. The latter feature can be used for enhancing image classifiers having the classes at the model’s exploiting stage different from from classes during the training stage. While the Shannon information of SoftMax probability for target class is extended for mini-batch by the intra-class variance, the trained network itself is extended by the Hadamard layer with the parameters representing the class centers. Contrary to the existing solutions, this extra neural layer enables interfacing of the training algorithm to the standard stochastic gradient optimizers, e.g. AdaM algorithm. Moreover, this approach makes the computed centroids immediately adapting to the updating embedded vectors and finally getting the comparable accuracy in less epochs.</p></abstract>

A novel technique for deep learning of image classifiers is presented. The learned CNN models higher offer better separation of deep features (also known as embedded vectors) measured by Euclidean proximity and also no deterioration of the classification results by class membership probability. The latter feature can be used for enhancing image classifiers having the classes at the model’s exploiting stage different from from classes during the training stage. While the Shannon information of SoftMax probability for target class is extended for mini-batch by the intra-class variance, the trained network itself is extended by the Hadamard layer with the parameters representing the class centers. Contrary to the existing solutions, this extra neural layer enables interfacing of the training algorithm to the standard stochastic gradient optimizers, e.g. AdaM algorithm. Moreover, this approach makes the computed centroids immediately adapting to the updating embedded vectors and finally getting the comparable accuracy in less epochs.

On Intra-Class Variance for Deep Learning of Classifiers

Foundations of Computing and Decision Sciences

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

{"article-title":"On Intra-Class Variance for Deep Learning of Classifiers"}

A novel technique for deep learning of image classifiers is presented. The learned CNN models higher offer better separation of deep features (also known as...

On Intra-Class Variance for Deep Learning of Classifiers

Data publikacji: 28 sie 2019

Zakres stron: 285 - 301

Otrzymano: 29 sty 2019

Przyjęty: 23 kwi 2019

DOI: https://doi.org/10.2478/fcds-2019-0015

Słowa kluczowe
deep learning of image classifier, convolutional neural network, Shanon information measure, intra-class variance, image embedding

© 2019 Rafał Pilarczyk et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

On Intra-Class Variance for Deep Learning of Classifiers

Data publikacji: 28 sie 2019

Zakres stron: 285 - 301

Otrzymano: 29 sty 2019

Przyjęty: 23 kwi 2019

DOI: https://doi.org/10.2478/fcds-2019-0015

Słowa kluczowedeep learning of image classifier, convolutional neural network, Shanon information measure, intra-class variance, image embedding

© 2019 Rafał Pilarczyk et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Słowa kluczowe
deep learning of image classifier, convolutional neural network, Shanon information measure, intra-class variance, image embedding