Feature learning and deep architectures: new directions for music informatics

By Eric J. Humphrey, Juan Pablo Bello, Yann LeCun. 2013

139 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Signal Processing

How researchers see this paper

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

In Cognitive Neuroscience

In Signal Processing

Citations in the corpus

Deep Learning: Methods and Applications (2014, background): “In the general field of audio and music processing, the impacted areas by deep learning include mainly music signal processing and music information retrieval [15, 22, 141, 177, 178, 179, 319].”
Automatic music transcription: challenges and future directions (2013, background): “(2011) proposed a classification-based approach for piano transcription using features learned from deep belief networks (Humphrey et al. 2013) for computing a mid-level time-pitch representation.”
Deep Learning Techniques for Music Generation - A Survey (2017, background): “study by integrating as an additional dimension the challenge (after having analyzed it in [16]). 6 In [65], Graves presented an analysis focusing on recurrent neural networks and text generation. In [90], Humphrey et al. presented another analysis, sharing some issues about music representation (se”
DeepEar: robust smartphone audio sensing in unconstrained acoustic environments using deep learning (2015, background): “Understanding why deep learning models are able to outperform alternatives has been an area of considerable study [13, 27, 45, 40].”
Improving Content-based and Hybrid Music Recommendation using Deep Learning (2014, background): “In [36, 37], Humphrey et al. proposed that the traditional two-stage machine learning process — feature extraction and classi-ﬁcation/regression — should be conducted simultaneously.”
A review of unsupervised feature learning and deep learning for time-series modeling (2014, background): “The motivation for using deep networks is that music itself is structured hierarchically by a combination of chords, melodies and rhythms that creates motives, phrases, sections and finally entire pieces [56].”
An End-to-End Neural Network for Polyphonic Piano Music Transcription (2015, background): “Their ability to jointly learn feature transformations and classiﬁers makes neural networks particularly well suited to problems in MIR [30].”
Music transcription modelling and composition using deep learning (2016, methodology): “Deep learning is now being applied to music data, from analysing and modelling the content of sound recordings [22,23,26,32–34, 40, 41 In particular, we build long short-term memory (LSTM) networks having three hidden layers of 512 LSTM blocks each, and train them using approximately 23,000 music…”
A Simple Method to Determine if a Music Information Retrieval System is a “Horse” (2014, background): “Many researchers have observed a “performance limit” to MIR systems, which has been called a “glass ceiling,” and attributed to “signal only approaches” and the “semantic gap” [45]–[47], the lack considering the involvement of humans [44], or the inappropriateness of standard features and machine le”