Reproduce “Algorithmic Composition of Melodies with Deep Recurrent Neural Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Florian Colombo, Samuel P. Muscinelli, Alexander Seeholzer, Johanni Brea, Wulfram Gerstner
Researchers at EPFL developed a deep recurrent neural network model utilizing Gated Recurrent Units (GRUs) to compose monophonic melodies, demonstrating its ability to learn long-range temporal dependencies from a large corpus of Irish folk music. The model generates new, stylistically coherent melodies and provides musically plausible continuations for existing fragments, capturing intricate rhythmic and melodic patterns.
Key insights
- Gated Recurrent Units (GRUs) effectively capture long-range temporal dependencies in music, enabling the model to learn complex stylistic rules without explicit musical knowledge.
- Explicitly modeling the dependence of upcoming pitch on upcoming duration allows the network to better reflect the intrinsic relationship between these musical features.
- Training on a large, real-world music corpus enables the model to generalize and generate novel compositions that maintain the stylistic fidelity of the learned genre.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.