Abstract:
In this paper, we explores the use of machine learning algorithms for music generation and retrieval. Specifically, Long Short-Term Memory (LSTM) and Variational Autoenco...Show MoreMetadata
Abstract:
In this paper, we explores the use of machine learning algorithms for music generation and retrieval. Specifically, Long Short-Term Memory (LSTM) and Variational Autoencoders (VAEs) are employed to generate music based on a monodic sound composition (MIDI file). LSTM, a neural network commonly used in deep learning and artificial intelligence, is utilized to train the initial music set, evaluate and filter the music collection, and compose music by minimizing the loss function. VAEs, which are nonlinear unsupervised dimensionality reduction techniques, compress a music dataset into a low-dimensional representation. The paper discusses the limitations of conventional algorithmic composition techniques and presents the advantages of using LSTM and VAEs for music generation. The methodology section provides a detailed explanation of the dataset loading, exploration, cleaning, and preprocessing processes. Furthermore, the paper presents related works in the field of music generation using CNN, LSTM, and other neural networks. The results indicate the effectiveness of LSTM and VAEs in generating music and melodies, surpassing traditional methods in terms of melody clarity and complexity. The proposed models offer promising outcomes in automatic music generation and demonstrate the potential of deep learning algorithms in the field of music composition.
Published in: 2023 14th International Conference on Computing Communication and Networking Technologies (ICCCNT)
Date of Conference: 06-08 July 2023
Date Added to IEEE Xplore: 23 November 2023
ISBN Information: