Scheduled denoising autoencoders
This addresses the challenge of improving feature learning for image classification tasks, particularly in unsupervised settings, though it is incremental as it builds on existing denoising autoencoder methods.
The paper tackled the problem of learning multi-scale features in unsupervised representation learning by introducing scheduled denoising autoencoders, which adjust noise levels during training, and achieved the lowest ever reported error on CIFAR-10 among permutation-invariant methods after fine-tuning.
We present a representation learning method that learns features at multiple different levels of scale. Working within the unsupervised framework of denoising autoencoders, we observe that when the input is heavily corrupted during training, the network tends to learn coarse-grained features, whereas when the input is only slightly corrupted, the network tends to learn fine-grained features. This motivates the scheduled denoising autoencoder, which starts with a high level of noise that lowers as training progresses. We find that the resulting representation yields a significant boost on a later supervised task compared to the original input, or to a standard denoising autoencoder trained at a single noise level. After supervised fine-tuning our best model achieves the lowest ever reported error on the CIFAR-10 data set among permutation-invariant methods.