Companion Losses for Deep Neural Networks

  1. David Díaz-Vico 1
  2. Angela Fernández 1
  3. Dorronsoro, José R. 11
  1. 1 Universidad Autónoma de Madrid
    info

    Universidad Autónoma de Madrid

    Madrid, España

    ROR https://ror.org/01cby8j38

Livre:
Hybrid Artificial Intelligent Systems: 16th International Conference, HAIS 2021. Bilbao, Spain. September 22–24, 2021. Proceedings
  1. Hugo Sanjurjo González (coord.)
  2. Iker Pastor López (coord.)
  3. Pablo García Bringas (coord.)
  4. Héctor Quintián (coord.)
  5. Emilio Corchado (coord.)

Éditorial: Springer International Publishing AG

ISBN: 978-3-030-86271-8,  978-3-030-86270-1

Année de publication: 2021

Pages: 538-549

Congreso: Hybrid Artificial Intelligent Systems (HAIS) (16. 2021. Bilbao)

Type: Communication dans un congrès

DIALNET GOOGLE SCHOLAR

Résumé

Modern Deep Neuronal Network backends allow a great flexibility to define network architectures. This allows for multiple outputs with their specific losses which can make them more suitable for particular goals. In this work we shall explore this possibility for classification networks which will combine the categorical cross-entropy loss, typical of softmax probabilistic outputs, the categorical hinge loss, which extends the hinge loss standard on SVMs, and a novel Fisher loss which seeks to concentrate class members near their centroids while keeping these apart.

La source de données: Dialnet