Dienstag, 05. Mai 2026, 17:30 Uhr – 20:30 Uhr

Computational Creativity and Artificial Intelligence

Beschreibung:	This course provides an overview of computational creativity and its integration with machine learning systems, with an emphasis on how structured knowledge representations can guide and enhance generative models. The course explores probabilistic models, graph-based representations, and transformer architectures, focusing on their role in creative generation. While symbolic music generation is used as the primary application domain, the concepts and methodologies discussed are broadly applicable to other structured generative tasks in artificial intelligence, such as language generation, planning, and design. The course is theory-oriented and example-driven (hands-off), aiming to expose students to core ideas, design principles, and current research directions rather than implementation details. Intended Audience: Upper-level undergraduate and graduate students with an interest in artificial intelligence, machine learning, or computational creativity.
Art der Veranstaltung:	Seminar/Workshop
Referent:	Maximos Kaliakatsos-Papakostas, Associate Professor & Head of the Music Technology and Acoustics Department, Hellenic Mediterranean University, Greece
Voraussetzungen	Basic probability and linear algebra Introductory machine learning concepts Familiarity with neural networks is helpful but not required By the end of the course, students will be able to: Describe key paradigms in computational creativity and their historicalevelopment Compare probabilistic, graph-based, and transformer-based generative models Explain how structured knowledge can guide and constrain creative AI systems Critically assess strengths and limitations of data-driven versus knowledge-driven creativity Identify open research challenges in computational creativity and generative AI Assessment No formal assessment. Active participation and discussion are encouraged.
Kursprogramm	Day 1 – 5:30 to 8:30 p.m. Computational Creativity in Music: Pre-Transformer Approaches Overview of computational creativity in symbolic music Hidden Markov Models (HMMs) for melodic harmonization Conceptual Blending (CB) theory Computational implementations of CB Creativity in probabilistic models via CB mechanisms ------------------------------------------------------------------------------- Day 2 – 5:30 to 8:30 p.m. From HMMs to Transformers Limitations of probabilistic sequence models Transformer architectures and generative modeling Incorporating “hard” constraints in transformer-based generation Encoder-only transformer architectures Training curricula for effective constrained generation Interpreting attention patterns during inference ----------------------------------------------------------------------------------------------------------------- Day 3 – 9:00 to 11:00 a.m. Future Directions in Computational Creativity Conceptual relations between HMMs and Graph Neural Networks (GNNs) GNNs as structured knowledge representations for generation Guiding transformer generation using graph-based information Conceptual Blending on graph representations Discussion: evaluation, limitations, and future challenges in creative AI

Zurück