Postdoctoral Position in Machine Learning for Computational Music

We are recruiting a post-doctoral researcher in the area of machine learning, specifically, to develop algorithms for mining and automatic composition of successful musical structures using complex networks. The position is located at the School of Philosophy, Science and Literature of Ribeirão Preto (FFCLRP), University of São Paulo (USP), Ribeirão Preto ? SP, in Brazil.
More information of our research activities can be found at: http://dcm.ffclrp.usp.br/~zhao/
Required skills and expertise:
- Very good knowledge of written and spoken English (Português is not required);
- Strong background knowledge in machine learning, data mining and graph theory;
- Background knowledge in music theory;
- Goog knowlegde of languages and tools, such as C, C++, Java, Python, and Matlab.
Eduaction: a PhD degree in computer science, electrical engineering, computer engineering, or a similar area with strong publication record.
Mission: In computer music field, automatic composition aims to develop algorithms that allow automatically compose musical pieces using mathematical or computational methods, trying to preserve musical aesthetic in a coherent manner. Recently, a new subfield called Hit Song Science (HSS that aims to characterize and predict commercially successful songs has been emerged. The inverse problem is the automatic generation of musical structures aiming to commercial success. Due to the high dimensionality of this type of data, this project proposes the use of complex networks to address the following goals: analysis, prediction, and automatic composition of musical structures. Firstly, complex networks are constructed from the musical structure of each song considering pitch, rhythm, harmony and musical form, thereby obtaining four subnetworks of different levels. Then, the subnetworks are joined to generate a larger network (metanetwork). After that, the complex network measures are used to characterize musical structures. Subsequently, techniuqes of classification and clustering are applied to categorize musical success according to the popularity from specialized websites in music, such as Last.fm and Youtube. Finally, extracted musical features from different classes of successful musics will be used as target characteristics in the composition of new songs by artificial intelligence techniques, for example, Recurrent Neural Networks (RNN). We hope that the methods to be developed can be used in the phonographic industry and broadcasters aiming to predict potential success of songs.
Annual salary: US$32.400,00 + US$4.860,00 (technical reserve) + airplane tickets + US$2.700,00 (installation aid)
Starting date: July 20, 2014
If you are interested in it and believe that you qualify, please send a cover letter, a résumé with a list of publications, and the names, e-mail addresses and phone numbers of at least three references to:
zhao-AT-usp.br. Please mention “Application to Post-Doctoral Position” in the title of your e-mail.