Cagri Erdem

Cagri Erdem

Post-doctoral research fellow, Department for Informatics, University of Oslo

MishMash role: Member · WP1 , WP2 , WP3 , WP4 , WP7

MishMash Projects

Other projects

Latest results

Type

Book chapter

  1. Book chapter, 2026

    Modeling Relations Between Musical Events in Continuous Time with Transformer Models for Live Co-Improvisational Interactions | Zenodo

    Vincenzo Madaghiele ; Stefano Fasciani ; Tejaswinee Kelkar ; Cagri Erdem

    This paper presents a live semi-autonomous system that co-improvises with a live musician using a model learned from the relationship observed in paired recordings of an improvising duo. This responsive system is based on a deep learning approach that models the relations between sequences of events produced by co-playing musicians in continuous time, using transformer models. We present the architecture for simultaneous sequences of sonic events and provide a quantitative evaluation on several representative tasks. Results are compared against a canonical transformer and a multichannel Factor Oracle, the latter being a widely used model for live sequence-based symbolic music generation. We detail a live implementation employing concatenative synthesis and introduce a customization procedure in which the generative model is iteratively retrained on curated, satisfactory sections from sound recordings of actual human-system co-improvisational interactions. This iterative fine-tuning enables the model’s stylistic output to diverge from the original training corpus. Two musical use cases demonstrate the application of this technique.

  2. Book chapter, 2026

    A Visualization and Control Interface for Rhythmic Relations

    Cagri Erdem ; Davide Rocchesso

    RhyGlyph is a radial visualization system that represents rhythmic interaction as three pairwise spacetime trajectories within a compact three-wedge glyph, supporting both overview and detailed reading. Building on this representation, RhyDiff is a generative framework that learns aligned latent spaces for symbolic drum patterns and the encoded trajectory features through paired VQ-VAEs and FiLM-conditioned diffusion. Users edit relational trajectories and metadata to generate new drum patterns by treating visualization itself as a control surface.

Journal article

  3. Journal article, 2026

    Inverse and indirect mappings in embodied AI systems in everyday environments

    Maham Riaz ; Cagri Erdem ; Alexander Refsum Jensenius

    This paper explores how musicking technologies—interactive systems with musical properties—can enhance everyday public environments. We are particularly interested in investigating the effects of musical interactions in non-musical settings, such as offices, meeting rooms, and social work areas. Traditional music technologies (such as instruments) are built for goal-directed, conscious, and voluntary interactions. We propose a new perspective on embodied AI through systems that utilize indirect, inverse, unconscious, and, at times, involuntary interactions. Four different sound/music systems are examined and discussed with regard to their activity level: a reactive “birdbox,” a reactive painting, active self-playing guitars, and interactive music balls. All these systems are multimodal, containing sensors that detect various physical inputs to produce sound and light, and having varying levels of perceived agency. The paper explores differences between direct/indirect and regular/inverse embodied AI paradigms. This study demonstrates how minimalistic interactions have the potential to yield complex and engaging musicking experiences, challenging the norms of overly intricate AI implementations.

More results in NVA…