Gerhard Widmer

Exploring Music Collections by Browsing Different Views

The availability of large music collections calls for ways to efficiently access and explore them. We present a new approach which combines descriptors derived from audio analysis with meta-information to create different views of a collection. Such views can have a focus on timbre, rhythm, artist, style or other aspects of music. For each view the pieces of music are organized on a map in such a way that similar pieces are located close to each other. The maps are visualized using an Islands of Music metaphor where islands represent groups of similar pieces. The maps are linked to each other using a new technique to align self-organizing maps. The user is able to browse the collection and explore different aspects by gradually changing focus from one view to another. We demonstrate our approach on a small collection using a meta-information-based view and two views generated from audio analysis, namely, beat periodicity as an aspect of rhythm and spectral information as an aspect of timbre.

Computational Models of Expressive Music Performance: The State of the Art

This contribution gives an overview of the state of the art in the field of computational modeling of expressive music performance. The notion of predictive computational model is briefly discussed, and a number of quantitative models of various aspects of expressive performance are briefly reviewed. Four selected computational models are reviewed in some detail. Their basic principles and assumptions are explained and, wherever possible, empirical evaluations of the models on real performance data are reported. In addition to these models, which focus on general, common principles of performance, currently ongoing research on the formal characterisation of differences in individual performance style are briefly presented.

Effective Learning in Dynamic Environments by Explicit Context Tracking

Daily experience shows that in the real world, the meaning of many concepts heavily depends on some implicit context, and changes in that context can cause radical changes in the concepts. This paper introduces a method for incremental concept learning in dynamic environments where the target concepts may be context-dependent and may change drastically over time. The method has been implemented in a system called FLORA3. FLORA3 is very flexible in adapting to changes in the target concepts and tracking concept drift. Moreover, by explicitly storing old hypotheses and re-using them to bias learning in new contexts, it possesses the ability to utilize experience from previous learning. This greatly increases the systems effectiveness in environments where contexts can reoccur periodically. The paper describes the various algorithms that constitute the method and reports on several experiments that demonstrate the flexibility of FLORA3 in dynamic environments.

Tracking Context Changes through Meta-Learning

The article deals with the problem of learning incrementally (‘on-line’) in domains where the target concepts are context-dependent, so that changes in context can produce more or less radical changes in the associated concepts. In particular, we concentrate on a class of learning tasks where the domain provides explicit clues as to the current context (e.g., attributes with characteristic values). A general two-level learning model is presented that effectively adjusts to changing contexts by trying to detect (via ‘meta-learning’) contextual clues and using this information to focus the learning process. Context learning and detection occur during regular on-line learning, without separate training phases for context recognition. Two operational systems based on this model are presented that differ in the underlying learning algorithm and in the way they use contextual information: METAL(B) combines meta-learning with a Bayesian classifier, while METAL(IB) is based on an instance-based learning algorithm. Experiments with synthetic domains as well as a number of ‘real-world’ problems show that the algorithms are robust in a variety of dimensions, and that meta-learning can produce substantial increases in accuracy over simple object-level learning in situations with changing contexts.

Prediction of Ordinal Classes Using Regression Trees

This paper is devoted to the problem of learning to predict ordinal (i.e., ordered discrete) classes using classification and regression trees. We start with S-CART, a tree induction algorithm, and study various ways of transforming it into a learner for ordinal classification tasks. These algorithm variants are compared on a number of benchmark data sets to verify the relative strengths and weaknesses of the strategies and to study the trade-off between optimal categorical classification accuracy (hit rate) and minimum distance-based error. Preliminary results indicate that this is a promising avenue towards algorithms that combine aspects of classification and regression.

Machine Discoveries: A Few Simple, Robust Local Expression Principles

The paper presents a new approach to discovering general rules of expressive music performance from real performance data via inductive machine learning. A new learning algorithm is briefly presented, and then an experiment with a very large data set (performances of 13 Mozart piano sonatas) is described. The new learning algorithm succeeds in discovering some extremely simple and general principles of musical performance (at the level of individual notes), in the form of categorical prediction rules. These rules turn out to be very robust and general: when tested on performances by a different pianist and even on music of a different style (Chopin), they exhibit a surprisingly high degree of predictive accuracy.

A music search engine built upon audio-based and web-based similarity measures

An approach is presented to automatically build a search engine for large-scale music collections that can be queried through natural language. While existing approaches depend on explicit manual annotations and meta-data assigned to the individual audio pieces, we automatically derive descriptions by making use of methods from Web Retrieval and Music Information Retrieval. Based on the ID3 tags of a collection of mp3 files, we retrieve relevant Web pages via Google queries and use the contents of these pages to characterize the music pieces and represent them by term vectors. By incorporating complementary information about acous tic similarity we are able to both reduce the dimensionality of the vector space and improve the performance of retrieval, i.e. the quality of the results. Furthermore, the usage of audio similarity allows us to also characterize audio pieces when there is no associated information found on the Web.

Discovering simple rules in complex data: a meta-learning algorithm and some surprising musical discoveries

This article presents a new rule discovery algorithm named PLCG that can find simple, robust partial rule models (sets of classification rules) in complex data where it is difficult or impossible to find models that completely account for all the phenomena of interest. Technically speaking, PLCG is an ensemble learning method that learns multiple models via some standard rule learning algorithm, and then combines these into one final rule set via clustering, generalization, and heuristic rule selection. The algorithm was developed in the context of an interdisciplinary research project that aims at discovering fundamental principles of expressive music performance from large amounts of complex real-world data (specifically, measurements of actual performances by concert pianists). It will be shown that PLCG succeeds in finding some surprisingly simple and robust performance principles, some of which represent truly novel and musically meaningful discoveries. A set of more systematic experiments shows that PLCG usually discovers significantly simpler theories than more direct approaches to rule learning (including the state-of-the-art learning algorithm RIPPER), while striking a compromise between coverage and precision. The experiments also show how easy it is to use PLCG as a meta-learning strategy to explore different parts of the space of rule models.

Playing Mozart by Analogy: Learning Multi-level Timing and Dynamics Strategies

The article describes basic research in the area of machine learning and musical expression. A first step towards automatic induction of multi-level models of expressive performance (currently only tempo and dynamics) from real performances by skilled pianists is presented. The goal is to learn to apply sensible tempo and dynamics “shapes” at various levels of the hierarchical musical phrase structure. We propose a general method for decomposing given expression curves into elementary shapes at different levels, and for separating phrase-level expression patterns from local, note-level ones. We then present a hybrid learning system that learns to predict, via two different learning algorithms, both note-level and phrase-level expressive patterns, and combines these predictions into complex composite expression curves for new pieces. Experimental results indicate that the approach is generally viable; however, we also discuss a number of severe limitations that still need to be overcome in order to arrive at truly musical machine-generated performances.

An innovative three-dimensional user interface for exploring music collections enriched

We present a novel, innovative user interface to music repositories. Given an arbitrary collection of digital music files, our system creates a virtual landscape which allows the user to freely navigate in this collection. This is accomplished by automatically extracting features from the audio signal and training a Self-Organizing Map (SOM) on them to form clusters of similar sounding pieces of music. Subsequently, a Smoothed Data Histogram (SDH) is calculated on the SOM and interpreted as a three-dimensional height profile. This height profile is visualized as a three-dimensional island landscape containing the pieces of music. While moving through the terrain, the closest sounds with respect to the listeners current position can be heard. This is realized by anisotropic auralization using a 5.1 surround sound model. Additionally, we incorporate knowledge extracted automatically from the web to enrich the landscape with semantic information. More precisely, we display words and related images that describe the heard music on the landscape to support the exploration.