EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Applied Research on Deep Generative Modeling for Automated Music Composition

Junyi Han

Quantum Information & Computation · 01.06.2026 · 26 (2) · 388-403

DOI: 10.2478/qic-2026-0022 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Quantum Information & Computation, online erschienen am 30.06.2026. Verfasst von Junyi Han. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Music is an essential element in entertainment life, which can resonate in human physiology and emotion. However, with the continuous pioneering and development of deep learning technology, people’s demand for music has increased dramatically. The traditional way of human creation is no longer compatible with it. To this end, this paper proposes an end-to-end music generation model, MIDI-GAN, which introduces a multi-scale feature extraction module based on null convolution to improve the discriminator’s long-range feature extraction capability for music data. In order to solve the problem that cyclic music generation leads to the accumulation of errors, which is easy to lead to the poor effect of long-time music generation, the semi-recurrent music generation model LMV-GAN based on the long-time memory structure is further proposed, which gives different weights to the data of previous time steps and inputs them into the encoder, which prompts the encoder to construct a richer coding space and improves the effect of the model in the long-time music generation. In the experiments, the effects generated by the LM-VGAN model are higher than the metric scores of the TrMTMG and R-MTMG models in the whole, and the melody score on the MUTMIDI dataset is the highest of 8.95, which verifies that this paper’s model improves the overall effects of the music to a certain extent.

Zitieren

APA

Han, J. (2026). Applied Research on Deep Generative Modeling for Automated Music Composition. Quantum Information & Computation, 26(2), 388-403. https://doi.org/10.2478/qic-2026-0022

Vancouver

Han J. Applied Research on Deep Generative Modeling for Automated Music Composition. Quantum Information & Computation. 2026;26(2):388-403. doi:10.2478/qic-2026-0022

BibTeX
@article{Han2026,
  author = {Han, Junyi},
  title = {Applied Research on Deep Generative Modeling for Automated Music Composition},
  journal = {Quantum Information & Computation},
  year = {2026},
  volume = {26},
  number = {2},
  pages = {388--403},
  doi = {10.2478/qic-2026-0022}
}
Kostenlos starten →