Archiv der Kategorie: MIR Research

Presentation of tempo recognition thesis at AES conference

The paper which summarizes the results of the diploma thesis „Tempo recognition based on audio signal for slow music pieces“ is accepted for the proceedings of the AES conference and will be presented as a poster. (Thorsten Deinert, Igor Vatolkin, … Weiterlesen

Veröffentlicht unter Publications | Schreib einen Kommentar

GECCO paper accepted

A paper is accepted for the GECCO conference: I. Vatolkin, M. Preuß, G. Rudolph – Multi-Objective Feature Selection in Music Genre and Style Recognition Tasks

Veröffentlicht unter General, Publications | Schreib einen Kommentar

New EA in AMUSE

Several new evolutionary strategies developed for the publication mentioned in http://sig-ma.de/2010/87 are now available in AMUSE for optimization of feature selection. It is possible to adjust the mutation probability based on the correlation with the label and to manage the … Weiterlesen

Veröffentlicht unter AMUSE & MIR Software | Schreib einen Kommentar

Application of SMS-EMOA for recall/specificity optimization

After SMS-EMOA [1] became a part of the actual (currently unreleased) AMUSE version, I’ve run a couple of experiments optimizing both recall and specificity. Here is an example for the performance progress for Jazz music category (blue circles: solutions from evaluations 1-500; green: … Weiterlesen

Veröffentlicht unter MIR Research | Schreib einen Kommentar

RapidMiner R plugin

Data Mining framework RapidMiner has now connection to R! Many further statistical methods and even the application of any R scripts becomes possible. For further details see: http://rapid-i.com/content/view/219/1/lang,en/ In future we are going to integrate the new RapidMiner into AMUSE.

Veröffentlicht unter AMUSE & MIR Software, MIR Research | Schreib einen Kommentar

PPSN publication about feature selection by EA

The joint publication of the Faculty of Computational Statistics and Chair of the Algorithm Engineering (TU Dortmund) [1] was recently presented as poster at the PPSN conference in Krakow. Evolutionary strategy driven by asymmetric mutation and several hybridization methods were developed and compared for the feature selection task for … Weiterlesen

Veröffentlicht unter Publications | Schreib einen Kommentar

New book about MIR

A new interesting book has been published recently at Springer: Ras, Zbigniew W.; Wieczorkowska, Alicja (Eds.): Advances in Music Information Retrieval (http://www.springer.com/engineering/mathematical/book/978-3-642-11673-5). It will be soon available in our library at Chair of Algorithm Engineering.

Veröffentlicht unter Publications | Schreib einen Kommentar

Bug fixed

The mapping of song partitions to songs was not done correctly during the n-fold cross-validation; now this bug has been fixed. Please checkout the new AMUSE version!

Veröffentlicht unter AMUSE & MIR Software | Schreib einen Kommentar

AMUSE beta 0.1 released

Just for the ISMIR 2010 conference we have uploaded to SourceForge the 0.1 beta version of AMUSE. It is available here: http://sourceforge.net/projects/amuse-framework/ Here is a screenshot with the AMUSE experiment table:

Veröffentlicht unter AMUSE & MIR Software | Schreib einen Kommentar

Project on multi-objective optimization of music classification started

The 3-year research project „Multi-objective optimization of music classification based on high-level features with Computational Intelligence methods“ was started on the 1st April 2009. The project is funded by Klaus Tschira Foundation.

Veröffentlicht unter Project News | Schreib einen Kommentar