Location
https://ccse.kennesaw.edu/computing-showcase/cday-programs/fall23program.php
Document Type
Event
Start Date
30-11-2023 4:00 PM
Description
Generative AI has transformed music creation, blending human and machine artistry. This study presents a neural network model trained on piano MIDI files for music generation, utilizing LSTM and self-attention mechanisms to capture music's complexity. Bayesian optimization with Tree-structured Parzen Estimator (TPE) refines the model's hyperparameters. The architecture includes bidirectional GRUs and self-attention layers, trained on the extensive Magenta MAESTRO dataset. The model, bettered by TPE over conventional tuning, is assessed for accuracy and expressiveness. The paper details the model's design and validates TPE's efficiency, marking progress in AI's creative application in music.
Included in
UR-445 Symphony of Silicon: Rethinking Music Creation through Deep Learning Models
https://ccse.kennesaw.edu/computing-showcase/cday-programs/fall23program.php
Generative AI has transformed music creation, blending human and machine artistry. This study presents a neural network model trained on piano MIDI files for music generation, utilizing LSTM and self-attention mechanisms to capture music's complexity. Bayesian optimization with Tree-structured Parzen Estimator (TPE) refines the model's hyperparameters. The architecture includes bidirectional GRUs and self-attention layers, trained on the extensive Magenta MAESTRO dataset. The model, bettered by TPE over conventional tuning, is assessed for accuracy and expressiveness. The paper details the model's design and validates TPE's efficiency, marking progress in AI's creative application in music.