 |
ML Newsnew | past | best | rss | submit | about |
| A Recurrent Neural Network Based Alternative to Convolutional Networks (arxiv.org) | |
| 5 points by dribnet on May 5, 2015 | hide | past | pdf | discuss on HN | |
|
In plain words: Instead of the usual convolution and pooling layers, four small recurrent nets sweep across the image in both directions, left-to-right and top-to-bottom, to build up a picture of what is there. On three image-recognition tests it performed about as well as the usual design. Abstract · ReNet: A Recurrent Neural Network Based Alternative to Convolutional NetworksIn this paper, we propose a deep neural network architecture for object recognition based on recurrent neural networks. The proposed network, called ReNet, replaces the ubiquitous convolution+pooling layer of the deep convolutional neural network with four recurrent neural networks that sweep horizontally and vertically in both directions across the image. We evaluate the proposed ReNet on three widely-used benchmark datasets; MNIST, CIFAR-10 and SVHN. The result suggests that ReNet is a viable alternative to the deep convolutional neural network, and that further investigation is needed. | |
| add comment on HN |