Zero-bias autoencoders and the benefits of co-adapting features

Abstract

Regularized training of an autoencoder typically results in hidden unit biases that take on large negative values. We show that negative biases are a natural result of using a hidden layer whose responsibility is to both represent the input data and act as a selection mechanism that ensures sparsity of the representation. We then show that negative biases impede the learning of data distributions whose intrinsic dimensionality is high. We also propose a new activation function that decouples the two roles of the hidden layer and that allows us to learn representations on data with very high intrinsic dimensionality, where standard autoencoders typically fail. Since the decoupled activation function acts like an implicit regularizer, the model can be trained by minimizing the reconstruction error of training data, without requiring any additional regularization.

References in corpus (5)

• Improving neural networks by preventing co-adaptation of feature detectors
• Fast Inference in Sparse Coding Algorithms with Applications to Object Recognition
• k-Sparse Autoencoders
• On the number of response regions of deep feed forward networks with piece-wise linear activations
• Saturating Auto-Encoders

Cited by in corpus (13)

• A survey on modern trainable activation functions
• Learning Combinations of Activation Functions
• Automated Pruning for Deep Neural Network Compression
• Scheduled denoising autoencoders
• Learning Deep ${\ell }_0$ Encoders
• Denoising without access to clean data using a partitioned autoencoder
• Estimation of Tissue Microstructure Using a Deep Network Inspired by a Sparse Reconstruction Framework
• Autonomous Deep Learning: A Genetic DCNN Designer for Image Classification
• Autoencoders Learn Generative Linear Models
• Decomposition-Based Domain Adaptation for Real-World Font Recognition
• On Binary Classification with Single-Layer Convolutional Neural Networks
• Fiber Orientation Estimation Guided by a Deep Network
• Unsupervised Deep Transfer Feature Learning for Medical Image Classification