An Information Theoretic Approach to the Autoencoder
| Autor: | Vincenzo Crescimanna, Bruce P. Graham |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Computer Science::Machine Learning
business.industry Computer science Noise reduction Computer Science::Neural and Evolutionary Computation Pattern recognition Data_CODINGANDINFORMATIONTHEORY Mutual information Autoencoder Statistics::Machine Learning Computer Science::Computer Vision and Pattern Recognition Artificial intelligence Infomax Representation (mathematics) business Feature learning MNIST database |
| Zdroj: | Proceedings of the International Neural Networks Society ISBN: 9783030168407 INNSBDDL |
| DOI: | 10.1007/978-3-030-16841-4_10 |
| Popis: | We present a variation of the Autoencoder (AE) that explicitly maximizes the mutual information between the input data and the hidden representation. The proposed model, the InfoMax Autoencoder (IMAE), by construction is able to learn a robust representation and good prototypes of the data. IMAE is compared both theoretically and then computationally with the state of the art models: the Denoising and Contractive Autoencoders in the one-hidden layer setting and the Variational Autoencoder in the multi-layer case. Computational experiments are performed with the MNIST and Fashion-MNIST datasets and demonstrate particularly the strong clusterization performance of IMAE. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|