Abstract
Algorithmic music composition is a way of composing musical pieces with minimal to no human intervention. While recurrent neural networks are traditionally applied to many sequence-to-sequence prediction tasks, including successful implementations of music composition, their standard supervised learning approach based on input-to-output mapping leads to a lack of note variety. These models can therefore be seen as potentially unsuitable for tasks such as music generation. Generative adversarial networks learn the generative distribution of data and lead to varied samples. This work implements and compares adversarial and non-adversarial training of recurrent neural network music composers on MIDI data. The resulting music samples are evaluated by human listeners, and their preferences are recorded. The evaluation indicates that adversarial training produces more aesthetically pleasing music.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Yamshchikov, I.P., Tikhonov, A.: Music generation with variational recurrent autoencoder supported by history. SN Appl. Sci. 2(12), 1–7 (2020). https://doi.org/10.1007/s42452-020-03715-w
Alfonseca, M., Cebrian, M., Puente, O.: A simple genetic algorithm for music generation by means of algorithmic information theory. In: Proceedings of the Institute of Electrical and Electronics Engineers Congress, Evolutionary Computation, pp. 25–28 (2007)
António, R., et al. The freesound loop dataset and annotation tool. In arXiv, (2020)
Arjovsky, M., Chintala, S., Bottou, L.: Wasserstein gan. In: Proceedings of the 34th International Conference on Machine Learning, vol. 70, pp. 214–223 (2017)
Bretan, M., Weinberg, G., Heck, L.: A unit selection methodology for music generation using deep neural networks. In: Proceedings of the International Conference on Computational Creativity (2017)
Chen, B., Hsu, E., Liao, W., Ramírez, M., Mitsufuji, Y., Yang, Y.: Automatic DJ transitions with differentiable audio effects and generative adversarial networks. In: IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2022, Virtual and Singapore, 23–27 May 2022, pp. 466–470. IEEE (2022)
Cho, K., et al.: Learning phrase representations using RNN encoder-decoder for statistical machine translation. In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, pp. 1724–1734 (2014)
Chongsheng, Z., et al.: An empirical study on the joint impact of feature selection and data resampling on imbalance classification. Appl. Intell. 53, 5449–5461 (2022)
Chu, H., Urtasun, R., Fidler, S.: Song from pi: a musically plausible network for pop music generation. In: Proceedings of the 5th International Conference on Learning Representations (2017)
Chung, J., Gulcehre, C., Cho, K., Bengio, Y.: Empirical evaluation of gated recurrent neural networks on sequence modeling. In: Proceedings of the 27th international conference on Neural Information Processing Systems, Deep Learning and Representation Learning (2014)
Colombo, F., Gerstner, W.: Bachprop: Learning to compose music in multiple styles. Arxiv, volume abs/1802.05162 (2018)
Cui, Z., Ke, R., Wang, Y.: Deep stacked bidirectional and unidirectional LSTM recurrent neural network for network-wide traffic speed prediction. In: 6th International Workshop on Urban Computing (2017)
Denil, M., Bazzani, L., Larochelle, H., de Freitas, N.: Learning where to attend with deep architectures for image tracking. IEEE Neural Comput. 24, 2151–2184 (2012)
Dong, H., Hsiao, W., Yang, L., Yang, Y.: Musegan: multi-track sequential generative adversarial networks for symbolic music generation and accompaniment. In: Proceedings of the the 30th international conference on Innovative Applications of Artificial Intelligence, pp. 34–41 (2018)
Edirisooriya, S., Dong, H.W., McAuley, J., Berg-Kirkpatrick, T.: An empirical evaluation of end-to-end polyphonic optical music recognition. In: International Society for Music Information Retrieval (2021)
Elgammal, A., Liu, B., Elhoseiny, M., Mazzone, M.: CAN: creative adversarial networks generating "art" by learning styles and deviating from style norms. In: Proceedings of the International Conference on Computational Creativity (2017)
Goodfellow, I., et al.: Generative adversarial nets. In: Proceedings of the 27th International Conference on Neural Information Processing Systems, pp. 2672–2680 (2014)
Henning, B., Soon, O., Enno, S., Joachim, B.M.: The balanced accuracy and its posterior distribution. In: 20th International Conference on Pattern Recognition, pp. 3121–3124 (2010)
Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9, 1735–1780 (1997)
Huang, A., Wu, R.: Deep learning for music. Arxiv, volume abs/1606.04930 (2016)
Hui, J.: GAN-wasserstein GAN and WGAN-GP. Medium (2018)
Hung, T., Chen, B., Yeh, Y., Yang, Y.: A benchmarking initiative for audio-domain music generation using the freesound loop dataset. In: Proceedings of the 22nd International Society for Music Information Retrieval Conference, ISMIR 2021, Online, November 7–12, 2021, pp. 310–317 (2021)
Juefei-Xu, F., Boddeti, V.N., Savvides, M.: Gang of GANs: generative adversarial networks with maximum margin ranking. ArXiv, volume abs/1704.04865 (2017)
Kim, J.: Using keras and theano for deep learning driven jazz generation (2017). https://deepjazz.io
Lackner, K.: Composing a melody with long-short term memory (LSTM) recurrent neural networks, Bachelor’s thesis, Technische Universität München, Munich, Germany (2016)
Lerdahl, F., Jackendoff, R.: On the Algorithmic Description of the Process of Composing Music. MIT Press (1983)
Liu, I. Randall, R.: Predicting missing music components with bidirectional long short-term memory neural networks. In: Proceedings of the 17th International Society on Music Information Retrieval Conference (2016)
Lopyrev, K.: Generating news headlines with recurrent neural networks. ArXiv, volume abs/1512.01712 (2015)
Mogren, O.: C-RNN-GAN: continuous recurrent neural networks with adversarial training. In: Proceedings of the 30th International Conference on Neural Information Processing Systems, Constructive Machine Learning Workshop (2016)
Mukkamala, M.C., Hein, M.: Variants of RMSProp and adagrad with logarithmic regret bounds. In: Proceedings of the 34th International Conference on Machine Learning, vol. 70, pp. 2545–2553 (2017)
Oord, A.,et al.: WaveNet: a generative model for raw audio. In: Proceedings of the 9th International Speech Communication Association, Speech Synthesis Workshop, pp. 125–125 (2016)
Oord, A., Kalchbrenner, N., Kavukcuoglu, K.: Pixel recurrent neural networks. In: Proceedings of the 33rd International Conference on Machine Learning, vol. 48, pp. 1747–1756 (2016)
Raffel, C.: Learning-based methods for comparing sequences, with applications to audio-to-midi alignment and matching (2018). https://colinraffel.com/projects/lmd/
Roberts, A., Engel, J., Eck, D.: Hierarchical variational autoencoders for music. NIPS Workshop on Machine Learning for Creativity and Design, vol. 3 (2017)
Roberts, A., Engel, J., Raffel, C., Hawthorne, C., Eck, D.: A hierarchical latent vector model for learning long-term structure in music. In: Proceedings of the 35th International Conference on Machine Learning, Machine Learning Research, vol. 80, pp. 4364–4373 (2018)
Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., Chen, X.: Improved techniques for training GANs. In: Proceedings of the 30th International Conference on Neural Information Processing Systems, pp. 2234–2242 (2016)
Schuster, M., Paliwal, K.: Bidirectional recurrent neural networks. IEEE Signal Process. 45, 2673–2681 (1997)
Shiebler, D.: Music RNN RBM. github.com/dshieble/Music_RNN_RBM (2017)
Shuqi, D., Zeyu, J., Celso, G., Roger, D.: Controllable deep melody generation via hierarchical music structure representation. arXiv preprint arXiv:2109.00663 (2021)
Srivastava, N., Hinton, G.E., Krizhevsky, A., Sutskever, I., Salakhutdinov, R.: Dropout: a simple way to prevent neural networks from overfitting. J. Mach. Learn. Res. 15(1), 1929–1958 (2014)
Sutskever, I., Vinyals, O., Le, Q.: Sequence to sequence learning with neural networks. Adv. Neural Inf. Process. Syst. 4, (2014)
Vaswani, A., et al.: Attention is all you need. In: Proceedings of the 31st International Conference on Neural Information Processing Systems, pp. 6000–6010 (2017)
García, V., Mollineda, R.A., Sánchez, J.S.: Index of balanced accuracy: a performance measure for skewed class distributions. In: Araujo, H., Mendonça, A.M., Pinho, A.J., Torres, M.I. (eds.) IbPRIA 2009. LNCS, vol. 5524, pp. 441–448. Springer, Heidelberg (2009). https://doi.org/10.1007/978-3-642-02172-5_57
Vincent, P., Lewandowski, N., Bengio, Y.: Modeling temporal dependencies in high-dimensional sequences: application to polyphonic music generation and transcription. In: Proceedings of the 27th International Conference on Machine Learning (2012)
Waite, E.: Generating long-term structure in songs and stories. Lookback-RNN-attention-RNN (2016)
Weel, J.: RoboMozart: generating music using LSTM networks trained per-tick on a midi collection with short music segments as input (2017)
Werbos, P.: Backpropagation through time: what it does and how to do it. Proc. IEEE 78, 1550–1560 (1990)
Wilcoxon, F.: Individual comparisons by ranking methods. Biometrics Bull. 1, 80–83 (1945)
Xenakis, I.: Formalized Music: Thought and Mathematics in Music. Pendragon Press (1992)
Yang, L., Chuo, S., Yang, Y.: MidiNet: a convolutional generative adversarial network for symbolic-domain music generation. In: Proceedings of the 18th International Society on Music Information Retrieval Conference, pp. 1–12 (2017)
Yu, L., Zhang, W., Wang, J., Yu, Y.: Seqgan: Sequence generative adversarial nets with policy gradient. In: Proceedings of the 31st AAAI Conference on Artificial Intelligence, vol. 31 (2017)
Zaripov, R.: On the algorithmic description of the process of composing music. USSR Acad. Sci. 132, 1283–1286 (1960)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
A Appendix :Listener Comments
A Appendix :Listener Comments
-
“I couldn’t get a feel of where the encoder-decoder song is going, the WGAN sample has a nice classical feel to it.”
-
“They both sound musical but the sound quality is bad.”
-
“I like the pace of the WGAN sample.”
-
“The encoder-decoder sample had too many silent spaces, the quiet spots emphasize the loud spots , which sounded like rambling.”
-
“The WGAN sample is a little chaotic but generally creates a good atmosphere, the encoder-decoder song has a good rhythm but no actual melody.
-
“I’d say the encoder-decoder song is better, has more rhythm, the WGAN sample sounds too fast and just noise to me.”
-
“The WGAN Sample sounds like me when I am under dissertation stress, terrible.”
-
“The WGAN sample sounds more creative, the encoder-decoder sample wins on rhythm although it takes so long to get there.”
-
“What’s important to me is they all sound musical.”
-
“I don’t listen to this genre so my view may be way off.”
-
“Most of the songs sound similar to me.”
-
“Wow, can’t believe the WGAN sample was generated by a machine, though it’s funny at the end LOL.”
-
“Can’t you get them to generate longer songs? perhaps with words? I like the encoder-decoder sample.”
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Mots’oehli, M., Bosman, A.S., De Villiers, J.P. (2023). Comparison of Adversarial and Non-Adversarial LSTM Music Generative Models. In: Arai, K. (eds) Intelligent Computing. SAI 2023. Lecture Notes in Networks and Systems, vol 711. Springer, Cham. https://doi.org/10.1007/978-3-031-37717-4_28
Download citation
DOI: https://doi.org/10.1007/978-3-031-37717-4_28
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-37716-7
Online ISBN: 978-3-031-37717-4
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)