- 期刊
Study of Generation based on Recurrent Neural Networks
摘要
This article mainly uses the Magenta framework to explore the two models of Melody RNN and Performance RNN, They are all recurrent neural networks based on LSTM (Long short-term memory).. The experiment uses a data set containing 120 songs. I set the Melody RNN model is trained for approximately 20,000 steps, while the Performance RNN model is trained for approximately 40,000 steps. The results show that the accuracy of Melody RNN is 80%, while that of Performance RNN is 90%.
參考文獻
CG;, Kapur SS;Stebbins GT;Goetz. “Vibration Therapy for Parkinson's Disease: Charcot's Studies Revisited.” Journal of Parkinson's Disease, U.S. National Library of Medicine, 2012, https://pubmed. ncbi.nlm.nih.gov/23939405/.
Trahan, Tabitha, et al. “The Music That Helps People Sleep and the Reasons They Believe It Works: A Mixed Methods Analysis of Online Survey Reports.” PloS One, Public Library of Science, 14 Nov. 2018, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6235300/#pone.0206531.ref015.
Harmat, László, et al. “Harmat - 2008 - Wiley Online Library.” Music Improves Sleep Quality in Students, 18 Apr. 2008, https://onlinelibrary.wiley.com/doi/10.1111/j.1365-2648.2008.04602.x.
Wang, Chun-Fang, et al. “Music Therapy Improves Sleep Quality in Acute and Chronic Sleep Disorders: A Meta-Analysis of 10 Randomized Studies.” International Journal of Nursing Studies, Pergamon, 9 Apr. 2013, https://www.sciencedirect.com/science/article/abs/pii/S0020748913000965.
Mao, Huanru Henry, et al. “DeepJ: Style-Specific Music Generation.” ArXiv.org, 3 Jan. 2018, https:// arxiv.org/abs/1801.00887.