A dynamical system as the source of augmentation in a deep learning problem
In this work we build a convolutional neural network capable of identifying individual birds by their songs. Since the actual data available from each individual is very limited, we use a dynamical system capable of synthesizing realistic songs, to generate surrogate-training data. The different syn...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2019-06-01
|
| Series: | Chaos, Solitons & Fractals: X |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590054419300107 |
| _version_ | 1818198129941413888 |
|---|---|
| author | P.L. Tubaro G.B. Mindlin |
| author_facet | P.L. Tubaro G.B. Mindlin |
| author_sort | P.L. Tubaro |
| collection | DOAJ |
| description | In this work we build a convolutional neural network capable of identifying individual birds by their songs. Since the actual data available from each individual is very limited, we use a dynamical system capable of synthesizing realistic songs, to generate surrogate-training data. The different synthetic songs are the result of integrating the dynamical system with slightly varied parameters. We show that a data set built in this way allows us to train the network to successfully identify the different individuals in our study. In this way, we present a novel way to perform data augmentation using dynamical systems. Keywords: Birdsong, Deep learning, Individual recognition |
| first_indexed | 2024-12-12T02:00:58Z |
| format | Article |
| id | doaj.art-50163f91b0bc46b790727c8c3a99e370 |
| institution | Directory Open Access Journal |
| issn | 2590-0544 |
| language | English |
| last_indexed | 2024-12-12T02:00:58Z |
| publishDate | 2019-06-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Chaos, Solitons & Fractals: X |
| spelling | doaj.art-50163f91b0bc46b790727c8c3a99e3702022-12-22T00:42:10ZengElsevierChaos, Solitons & Fractals: X2590-05442019-06-012A dynamical system as the source of augmentation in a deep learning problemP.L. Tubaro0G.B. Mindlin1División Ornitología, Museo Argentino de Ciencias Naturales Bernardino Rivadavia, ArgentinaIFIBA, CONICET and Departamento de Física, FCEyN, UBA, Buenos Aires, Argentina; Corresponding author.In this work we build a convolutional neural network capable of identifying individual birds by their songs. Since the actual data available from each individual is very limited, we use a dynamical system capable of synthesizing realistic songs, to generate surrogate-training data. The different synthetic songs are the result of integrating the dynamical system with slightly varied parameters. We show that a data set built in this way allows us to train the network to successfully identify the different individuals in our study. In this way, we present a novel way to perform data augmentation using dynamical systems. Keywords: Birdsong, Deep learning, Individual recognitionhttp://www.sciencedirect.com/science/article/pii/S2590054419300107 |
| spellingShingle | P.L. Tubaro G.B. Mindlin A dynamical system as the source of augmentation in a deep learning problem Chaos, Solitons & Fractals: X |
| title | A dynamical system as the source of augmentation in a deep learning problem |
| title_full | A dynamical system as the source of augmentation in a deep learning problem |
| title_fullStr | A dynamical system as the source of augmentation in a deep learning problem |
| title_full_unstemmed | A dynamical system as the source of augmentation in a deep learning problem |
| title_short | A dynamical system as the source of augmentation in a deep learning problem |
| title_sort | dynamical system as the source of augmentation in a deep learning problem |
| url | http://www.sciencedirect.com/science/article/pii/S2590054419300107 |
| work_keys_str_mv | AT pltubaro adynamicalsystemasthesourceofaugmentationinadeeplearningproblem AT gbmindlin adynamicalsystemasthesourceofaugmentationinadeeplearningproblem AT pltubaro dynamicalsystemasthesourceofaugmentationinadeeplearningproblem AT gbmindlin dynamicalsystemasthesourceofaugmentationinadeeplearningproblem |