Unsupervised feature learning for audio classification using convolutional deep belief networks

By Honglak Lee, Peter T. Pham, Yan Largman, Andrew Y. Ng. 2009

1,175 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Better Mixing via Deep Representations (2012, methodology): “We train a linear SVM on the concatenation of the raw input with the upper layers representations (which worked better than using only the top layer, a setup already used successfully when there is no supervised fine-tuning (Lee et al., 2009)).”
Representation Learning: A Review and New Perspectives (2012, methodology): “…include the following: efficient GPU training allowing one to train longer (more than 100 million visits of examples), an aspect first reported by Lee et al. (2009a); Ciresan et al. (2010), large number of labeled examples, artificially transformed examples (see Section 11.1), a large number of…”
Deep Speech: Scaling up end-to-end speech recognition (2014, background): “[27] who applied early unsupervised feature learning techniques to replace hand-built speech features.”
SoundNet: Learning Sound Representations from Unlabeled Video (2016, background): “Even though there are unsupervised [20] and supervised [27, 23, 6, 12] deep learning methods applied to sound classification, the models are limited by the amount of available labeled natural sound data.”
UvA-DARE (Digital Academic Repository) Self-Supervised Video Representation Learning With Odd-One-Out Networks (neutral): “The most common techniques studied include autoencoders [6, 20], restricted Boltzmann machines [19], convolutional deep belief networks [28], LSTMs and recurrent neural networks [36].”
Marginalized Denoising Autoencoders for Domain Adaptation (2012, methodology): “The outputs of their intermediate layers are then used as input features for SVMs (Lee et al., 2009).”
Deep Learning: Methods and Applications (2014, background): “A further example of hybrid deep networks is the use of generative models of DBNs to pre-train deep convolutional neural networks (deep CNNs) [215, 216, 217].”
SampleRNN: An Unconditional End-to-End Neural Audio Generation Model (2016, neutral): “Chung et al. (2015) also attempt to model raw audio waveforms which is in contrast to traditional approaches which use spectral features as in Tokuda et al. (2013), Bertrand et al. (2008), and Lee et al. (2009).”
Compressing Convolutional Neural Networks in the Frequency Domain (2015, background): “In the recent years convolutional neural networks (CNN) have led to impressive results in image classification [15,30,48,63], object detection [19, 58], image retrieval [43], image caption generation [26,37,58], face verification [47,55], video understanding [27] and audio classification [35].”
Deep learning for detecting robotic grasps (2013, neutral): “Recent methods based on deep learning [1] have demonstrated state-of-the-art performance in a wide variety of tasks, including visual recognition [35, 60], audio recognition [39, 41], and natural language processing [12].”
Learning Invariant Representations with Local Transformations (2012, neutral): “5% MFCC + CDBN (Lee et al., 2009) 80 .”
Demystifying Parallel and Distributed Deep Learning (2018, neutral): “Consequently, deep learning managed to outperform all existing approaches in speech recognition [152] and image”
Unsupervised Learning of Semantic Audio Representations (2017, background): “[11] applied convolutional deep belief networks to extract a representation for speech and music, but not general purpose nonspeech audio.”
Biological Sequence Modeling with Convolutional Kernel Networks (2017, neutral): “This has been the case for computer vision [22], natural language processing [10], audio signals [23], and more recently biological sequences [46, 1, 2].”
A tutorial survey of architectures, algorithms, and applications for deep learning (2014, methodology): “A further example of the hybrid deep architecture is the use of the generative model of DBN to pretrain deep convolutional neural networks (deep DNN) [123, 144, 145]).”
Computational Phenotype Discovery Using Unsupervised Feature Learning over Noisy, Sparse, and Irregular Clinical Data (2013, background): “For example, deep learning produces edge and spot detectors as translation-invariant features for photographic images [23], stereotyped object parts as features for images of objects [23], and phonemes as features from speech recordings [24].”