Emmanouil Karystinaios

Postdoctoral Researcher in Artificial Intelligence

Johannes Kepler University

Biography

Emmanouil Karystinaios is a Postdoctoral researcher in artificial intelligence working for the institute of Computational Perception and the Human-centered Artificial Intelligence (HCAI) group at Johannes Kepler university in Austria. His research focuses on Music Information Research, Graph Neural Networks, and Computational Musicology. Currently, he is working on the Automatic Analysis of Symbolic Music using Graph Neural Networks (GNNs) and on Generative Music Medicine, exploring the use of AI-generated music in therapeutic contexts.

His past and ongoing work includes Music Analysis, Structure Segmentation, Generative Music models, and developing Python packages such as Partitura and GraphMuse for symbolic music processing.

Download my resumé.

Interests

Graph Neural Networks
Music Generation and Control
Music Information Retrieval

Education

PhD in Artificial Intelligence, 2020-2024
Johannes Kepler University Austria
M.Sc. in Mathematical Logic (Mathematics), 2018-2020
Université Paris 7 Diderot
M.A. in Musical Programming and Composition, 2017-2019
Université Paris 8 Vincennes
B.A. in Musicology and Composition, 2012-2017
Aristotle University of Thessaloniki

Recent Posts

For more posts, please visit my Medium blog.

Language Models for Music Medicine Generation

This research introduces a novel model for generating music for music therapy purposes, named Language Models for Music Medicine Generation. The paper explores how AI-driven music can support mental health by guiding listeners through emotional transitions using generative music designed to align with established therapeutic principles.

Emmanouil Karystinaios

Nov 11, 2024 3 min read Large Language Models, Music Medicine, Generation

Language Models for Music Medicine Generation

From Notes to Insights - Visualizing Graph Neural Network Explanations with SMUG-Explain

Visualizing Graph Neural Network Explanations for music.

Emmanouil Karystinaios

Jun 27, 2024 7 min read Graph Neural Networks, Cadence, Interpretability

From Notes to Insights - Visualizing Graph Neural Network Explanations with SMUG-Explain

Harmonic Trajectories in the Tonnetz

Introducing the Tonnetz and related applications for composition and musical analysis.

Emmanouil Karystinaios

Dec 13, 2020 11 min read Symbolic Representations

Harmonic Trajectories in the Tonnetz

Recent Publications

Quickly discover relevant content by filtering publications.

Emmanouil Karystinaios (2024). Symbolic Music Analysis with Graph Neural Networks. PhD Thesis at JKU.

Featured Publications

Cluster and Separate: a GNN Approach to Voice and Staff Prediction for Score Engraving

In this work, we present a framework for separating the notes from a quantized symbolic music piece into multiple voices and staves. We propose an end-to-end system based on graph neural networks that clusters notes that belong to the same chord and connects them with edges if they are part of a voice.

Francesco Foscarin, Emmanouil Karystinaios, Eita Nakamura, Gerhard Widmer

Cluster and Separate: a GNN Approach to Voice and Staff Prediction for Score Engraving

GraphMuse: A Library for Symbolic Music Graph Processing

GraphMuse is a graph processing framework and library that facilitates efficient music graph processing and GNN training for symbolic music tasks. Central to our contribution is a new neighbor sampling technique specifically targeted toward meaningful behavior in musical scores. Additionally, GraphMuse integrates hierarchical modeling elements that augment the expressivity and capabilities of graph networks for musical tasks.

Emmanouil Karystinaios, Gerhard Widmer

GraphMuse: A Library for Symbolic Music Graph Processing

Perception-Inspired Graph Convolution for Music Understanding Tasks

In this paper we propose a new graph convolutional block, called MusGConv, specifically designed for the efficient processing of musical score data and motivated by general perceptual principles. We evaluate our approach on four different musical understanding problems, such as monophonic voice separation, harmonic analysis, cadence detection, and composer identification which, in abstract terms, translate to different graph learning problems, namely, node classification, link prediction, and graph classification. Our experiments demonstrate that MusGConv improves the performance on three of the aforementioned tasks while being conceptually very simple and efficient.

Emmanouil Karystinaios, Francesco Foscarin, Gerhard Widmer

Perception-Inspired Graph Convolution for Music Understanding Tasks

See all publications

Contact