Time filter

Source Type

Flexer A.,Austrian Research Institute for Artificial Intelligence
IEEE International Conference on Data Mining Workshops, ICDMW | Year: 2017

Outlier detection is the task of automatic identification of unknown data not covered by training data (e.g. a previously unknown class in classification). We explore outlier detection in the presence of hubs and anti-hubs, i.e. data objects which appear to be either very close or very far from most other data due to a problem of measuring distances in high dimensions. We compare a classic distance based method to two new approaches, which have been designed to counter the negative effects of hubness, on six high-dimensional data sets. We show that mainly anti-hubs pose a problem for outlier detection and that this can be improved by using a hubness-Aware approach based on re-scaling the distance space. © 2016 IEEE.


Gligorijevic V.,Jozef Stefan Institute | Skowron M.,Austrian Research Institute for Artificial Intelligence | Tadic B.,Jozef Stefan Institute
Physica A: Statistical Mechanics and its Applications | Year: 2013

High-resolution data of online chats are studied as a physical system in the laboratory in order to quantify collective behavior of users. Our analysis reveals strong regularities characteristic of natural systems with additional features. In particular, we find self-organized dynamics with long-range correlations in user actions and persistent associations among users that have the properties of a social network. Furthermore, the evolution of the graph and its architecture with specific k-core structure are shown to be related with the type and the emotion arousal of exchanged messages. Partitioning of the graph by deletion of the links which carry high arousal messages exhibits critical fluctuations at the percolation threshold. © 2012 Elsevier B.V. All rights reserved.


Flexer A.,Austrian Research Institute for Artificial Intelligence | Schnitzer D.,Johannes Kepler University
Computer Music Journal | Year: 2010

The effects of album and artist filters in audio similarity computed for very large music databases were reported. A data set D(ALL) of S = 254,398 song excerpts (30 seconds each) from a popular Web store selling music was used. The mel-frequency cepstral coefficients (MFCC) was a relatively direct representation of the spectral information of a signal and therefore of the specific sound of a song, fluctuation patterns (FP) are a more abstract kind of feature describing the amplitude modulation of the loudness per frequency band. The 30-second song excerpts in MP3 format were recomputed to 22,050 Hz mono audio signals. The frame size for computation of MFCC was 46.4 msec (1,024 samples). The resulting FP was a 12 × 30 (modulation frequencies, ranging from 0 to 10 Hz) matrix for each song. The first nearest neighbor for both methods Gaussians (G1) and FP was evaluated for every song in the database D(ALL). The first nearest neighbor was the song with minimum Kullback-Leibler divergence from the query song for method G1.


Schedl M.,Johannes Kepler University | Flexer A.,Austrian Research Institute for Artificial Intelligence
Proceedings of the 13th International Society for Music Information Retrieval Conference, ISMIR 2012 | Year: 2012

Personalized and context-aware music retrieval and recommendation algorithms ideally provide music that perfectly fits the individual listener in each imaginable situation and for each of her information or entertainment need. Although first steps towards such systems have recently been presented at ISMIR and similar venues, this vision is still far away from being a reality. In this paper, we investigate and discuss literature on the topic of user-centric music retrieval and reflect on why the breakthrough in this field has not been achieved yet. Given the different expertises of the authors, we shed light on why this topic is a particularly challenging one, taking a psychological and a computer science view. Whereas the psychological point of view is mainly concerned with proper experimental design, the computer science aspect centers on modeling and machine learning problems. We further present our ideas on aspects vital to consider when elaborating user-aware music retrieval systems, and we also describe promising evaluation methodologies, since accurately evaluating personalized systems is a notably challenging task. © 2012 International Society for Music Information Retrieval.


Niedermayer B.,Johannes Kepler University | Widmer G.,Austrian Research Institute for Artificial Intelligence
Proceedings of the 11th International Society for Music Information Retrieval Conference, ISMIR 2010 | Year: 2010

Most current audio-to-score alignment algorithms work on the level of score time frames; i.e., they cannot differentiate between several notes occurring at the same discrete time within the score. This level of accuracy is sufficient for a variety of applications. However, for those that deal with, for example, musical expression analysis such microtimings might also be of interest. Therefore, we propose a method that estimates the onset times of individual notes in a post-processing step. Based on the initial alignment and a feature obtained by matrix factorization, those notes for which the confidence in the alignment is high are chosen as anchor notes. The remaining notes in between are revised, taking into account the additional information about these anchors and the temporal relations given by the score. We show that this method clearly outperforms a reference method that uses the same features but does not differentiate between anchor and non-anchor notes. © 2010 International Society for Music Information Retrieval.


Schedl M.,Johannes Kepler University | Schnitzer D.,Austrian Research Institute for Artificial Intelligence
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2014

Current advances in music recommendation underline the importance of multimodal and user-centric approaches in order to transcend limits imposed by methods that solely use audio, web, or collaborative filtering data. We propose several hybrid music recommendation algorithms that combine information on the music content, the music context, and the user context, in particular integrating geospatial notions of similarity. To this end, we use a novel standardized data set of music listening activities inferred from microblogs (MusicMicro) and state-of-the-art techniques to extract audio features and contextual web features. The multimodal recommendation approaches are evaluated for the task of music artist recommendation. We show that traditional approaches (in particular, collaborative filtering) benefit from adding a user context component, geolocation in this case. © 2014 Springer International Publishing.


Holighaus N.,Austrian Academy of Sciences | Dorfler M.,University of Vienna | Velasco G.A.,University of the Philippines at Diliman | Grill T.,Austrian Research Institute for Artificial Intelligence
IEEE Transactions on Audio, Speech and Language Processing | Year: 2013

Audio signal processing frequently requires time-frequency representations and in many applications, a non-linear spacing of frequency bands is preferable. This paper introduces a framework for efficient implementation of invertible signal transforms allowing for non-uniform frequency resolution. Non-uniformity in frequency is realized by applying nonstationary Gabor frames with adaptivity in the frequency domain. The realization of a perfectly invertible constant-Q transform is described in detail. To achieve real-time processing, independent of signal length, slice-wise processing of the full input signal is proposed and referred to as sliCQ transform. By applying frame theory and FFT-based processing, the presented approach overcomes computational inefficiency and lack of invertibility of classical constant-Q transform implementations. Numerical simulations evaluate the efficiency of the proposed algorithm and the method's applicability is illustrated by experiments on real-life audio signals. © 2006-2012 IEEE.


Skowron M.,Austrian Research Institute for Artificial Intelligence
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2010

We present the concept and motivations for the development of Affect Listeners, conversational systems aiming to detect and adapt to affective states of users, and meaningfully respond to users' utterances both at the content- and affect-related level. In this paper, we describe the system architecture and the initial set of core components and mechanisms applied, and discuss the application and evaluation scenarios of Affect Listener systems. © 2010 Springer-Verlag.


Flexer A.,Austrian Research Institute for Artificial Intelligence | Schnitzer D.,Austrian Research Institute for Artificial Intelligence
Proceedings - IEEE 13th International Conference on Data Mining Workshops, ICDMW 2013 | Year: 2013

'Hubness' is a recently discovered general problem of machine learning in high dimensional data spaces. Hub objects have a small distance to an exceptionally large number of data points, and anti-hubs are far from all other data points. It is related to the concentration of distances which impairs the contrast of distances in high dimensional spaces. Computation of secondary distances inspired by shared nearest neighbor (SNN) approaches has been shown to reduce hubness and concentration and there already exists some work on direct application of SNN in the context of hubness in image recognition. This study applies SNN to a larger number of high dimensional real world data sets from diverse domains and compares it to two other secondary distance approaches (local scaling and mutual proximity). SNN is shown to reduce hubness but less than other approaches and, contrary to its competitors, it is only able to improve classification accuracy for half of the data sets. © 2013 IEEE.


Dixon S.,Austrian Research Institute for Artificial Intelligence
Proceedings of the 9th International Conference on Digital Audio Effects, DAFx 2006 | Year: 2013

Various methods have been proposed for detecting the onset times of musical notes in audio signals. We examine recent work on onset detection using spectral features such as the magnitude, phase and complex domain representations, and propose improvements to these methods: a weighted phase deviation function and a half-wave rectified complex difference. These new algorithms are compared with several state-of-the-art algorithms from the literature, and these are tested using a standard data set of short excerpts from a range of instruments (1060 onsets), plus a much larger data set of piano music (106054 onsets). Some of the results contradict previously published results and suggest that a similarly high level of performance can be obtained with a magnitude-based (spectral flux), a phase-based (weighted phase deviation) or a complex domain (complex difference) onset detection function.

Loading Austrian Research Institute for Artificial Intelligence collaborators
Loading Austrian Research Institute for Artificial Intelligence collaborators