Locally-connected transformations for deep GMMs
- Author
- Aäron van den Oord (UGent) and Joni Dambre (UGent)
- Organization
- Abstract
- Recently there has been a lot of interest in density modeling with Deep Generative Models. So far, these models have arguably been relatively successful on tasks such as modeling handwritten digits (MNIST), small image patches (e.g., 8 by 8 pixels) and other low-dimensional datasets. However, convincingly modeling higher dimensional data such as small images (e.g., 32 by 32 pixels and higher) is still a big unsolved problem. In this work we will extend and apply Deep Gaussian Mixture Models (DGMMs) to this task, by introducing locally connected transformations. Similarly to convolutions in deep neural networks, local connectivity in DGMMs allow us to train faster and with less overfitting than fully connected networks when applied to images. Our experiments show the benefits of using locally-connected Deep GMMs and give new insights on modeling higher dimensional images.
- Keywords
- deep learning, GMM, gaussian mixture models, machine learning, generative models
Downloads
-
Aaron dgmm locally connected.pdf
- full text
- |
- open access
- |
- |
- 1.13 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-7028865
- MLA
- van den Oord, Aäron, and Joni Dambre. “Locally-Connected Transformations for Deep GMMs.” International Conference on Machine Learning (ICML) : Deep Learning Workshop, Abstracts, 2015, pp. 1–8.
- APA
- van den Oord, A., & Dambre, J. (2015). Locally-connected transformations for deep GMMs. International Conference on Machine Learning (ICML) : Deep Learning Workshop, Abstracts, 1–8.
- Chicago author-date
- Oord, Aäron van den, and Joni Dambre. 2015. “Locally-Connected Transformations for Deep GMMs.” In International Conference on Machine Learning (ICML) : Deep Learning Workshop, Abstracts, 1–8.
- Chicago author-date (all authors)
- van den Oord, Aäron, and Joni Dambre. 2015. “Locally-Connected Transformations for Deep GMMs.” In International Conference on Machine Learning (ICML) : Deep Learning Workshop, Abstracts, 1–8.
- Vancouver
- 1.van den Oord A, Dambre J. Locally-connected transformations for deep GMMs. In: International Conference on Machine Learning (ICML) : Deep learning Workshop, Abstracts. 2015. p. 1–8.
- IEEE
- [1]A. van den Oord and J. Dambre, “Locally-connected transformations for deep GMMs,” in International Conference on Machine Learning (ICML) : Deep learning Workshop, Abstracts, Lille, France, 2015, pp. 1–8.
@inproceedings{7028865,
abstract = {{Recently there has been a lot of interest in density modeling with Deep Generative Models. So far, these models have arguably been relatively successful on tasks such as modeling handwritten digits (MNIST), small image patches (e.g., 8 by 8 pixels) and other low-dimensional datasets.
However, convincingly modeling higher dimensional data such as small images (e.g., 32 by 32 pixels and higher) is still a big unsolved problem.
In this work we will extend and apply Deep Gaussian Mixture Models (DGMMs) to this task, by introducing locally connected transformations. Similarly to convolutions in deep neural networks, local connectivity in DGMMs allow us to train faster and with less overfitting than fully connected networks when applied to images. Our experiments show the benefits of using
locally-connected Deep GMMs and give new insights on modeling higher dimensional images.}},
author = {{van den Oord, Aäron and Dambre, Joni}},
booktitle = {{International Conference on Machine Learning (ICML) : Deep learning Workshop, Abstracts}},
keywords = {{deep learning,GMM,gaussian mixture models,machine learning,generative models}},
language = {{eng}},
location = {{Lille, France}},
pages = {{1--8}},
title = {{Locally-connected transformations for deep GMMs}},
url = {{https://sites.google.com/site/deeplearning2015/20.pdf?attredirects=0}},
year = {{2015}},
}