使用RNN-RBM和基因演算法創作複音音樂

Polyphonic Music Generation by RNN-RBM and Genetic Algorithm

陳竹軒(Zhuxuan Chen)

指導教授 ： 蔡志忠

國立中正大學/工學院/資訊工程學系/碩士(2015年)

摘要

在資訊科學中，自動生成音樂是一個相當有趣的議題，在過去幾十年中，有數百篇論文在談論自動生成音樂，但大多數都是在單音的範疇裡討論。這篇論文主要是討論複音的自動生成音樂，也就是說，允許同一時刻裡出現多於一個的音符。我們使用了RNN-RBM這個模型產生一些音樂片段，並將這些音樂片段當作基因演算法中的初始族群，再經由與樂理相關的適應函數使之演化，演化後，我們調整了這些音樂片段使之符合樂理中的三段體曲式。在論文的最後，我們比較了不同的適應函數以及訓練資料庫的效果。

關鍵字

自動音樂生成 ； 基因演算法

並列摘要

Automatic music generation is an interesting issue in computer science. During the past few decades, numerous papers were discussed this issue, but most of them focused on monophonic music at the present stage. This thesis studies the problem of Automatic generation of Polyphonic music, which allows plurality of notes existing in a time slot. We adopt pieces of music generated by Recurrent Neural Networks - Restricted Boltzmann Machine (RNN-RBM) and take these licks to serve as an initial population in our genetic system. The fitness function of our system basically employs music theory, i.e. hardwired fitness functions. After evolution, we adjust some piece to fit Ternary form (a kind of music form). In the end of thesis, we made a comparison between different fitness and dataset.

並列關鍵字

genetic algorithm ； algorithmic music ； rnn-rbm

參考文獻

[3] Hiller, Lejaren Arthur, and Leonard M. Isaacson. Experimental Music; Composition with an electronic computer. Greenwood Publishing Group Inc., 1979.

[5] Ames, Charles. "The Markov process as a compositional model: a survey and tutorial." Leonardo (1989): 175-187.

[6] Hinton, Geoffrey, Simon Osindero, and Yee-Whye Teh. "A fast learning algorithm for deep belief nets." Neural computation 18.7 (2006): 1527-1554.

[11] Biles, John. "GenJam: A genetic algorithm for generating jazz solos."Proceedings of the International Computer Music Conference. INTERNATIONAL COMPUTER MUSIC ACCOCIATION, 1994.

[18] Marques, M., et al. "Music composition using genetic evolutionary algorithms." (2000): 714-719.