Improving convolutional neural network design via variable neighborhood search

dc.contributor.author Teresa Finisterra Araújo en
dc.contributor.author Guilherme Moreira Aresta en
dc.contributor.author Bernardo Almada-Lobo en
dc.contributor.author Ana Maria Mendonça en
dc.contributor.author Aurélio Campilho en
dc.date.accessioned 2018-01-14T17:05:02Z
dc.date.available 2018-01-14T17:05:02Z
dc.date.issued 2017 en
dc.description.abstract An unsupervised method for convolutional neural network (CNN) architecture design is proposed. The method relies on a variable neighborhood search-based approach for finding CNN architectures and hyperparameter values that improve classification performance. For this purpose, t-Distributed Stochastic Neighbor Embedding (t-SNE) is applied to effectively represent the solution space in 2D. Then, k-Means clustering divides this representation space having in account the relative distance between neighbors. The algorithm is tested in the CIFAR-10 image dataset. The obtained solution improves the CNN validation loss by over 15% and the respective accuracy by 5%. Moreover, the network shows higher predictive power and robustness, validating our method for the optimization of CNN design. © Springer International Publishing AG 2017. en
dc.identifier.uri http://repositorio.inesctec.pt/handle/123456789/6074
dc.identifier.uri http://dx.doi.org/10.1007/978-3-319-59876-5_41 en
dc.language eng en
dc.relation 6381 en
dc.relation 6071 en
dc.relation 5428 en
dc.relation 6321 en
dc.relation 6320 en
dc.rights info:eu-repo/semantics/embargoedAccess en
dc.title Improving convolutional neural network design via variable neighborhood search en
dc.type conferenceObject en
dc.type Publication en
Files
Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
P-00M-WJ8.pdf
Size:
887.12 KB
Format:
Adobe Portable Document Format
Description: