在電腦作曲中創作悅耳的音樂是一項困難的任務，因為需要考慮許多音樂元素，例如音高，和弦，節奏，曲式和音階, 在音樂即興演奏中，需要豐富的音樂理論知識來規範旋律的進行方式，而不是隨機增減音符。由於旋律應以和諧的方式進行並且在正確的和弦架構下進行，因此音樂即興演奏可以視為基於規則的作曲系統。在基因演算法（GA）中，“物競天擇，適者生存”的概念也可以應用於音樂即興演奏, 由於適應函數是一種有助於提供染色體演化方向的機制, 在這個研究中，我們採用基因演算法即興創作給定的音樂片段，並設計了三種衡量方法來指導旋律以和諧的方式發展，使旋律有更多的變化，但仍保持與原始歌曲相似的結構以便聽眾可以識別其來源。

Composing melodious music is a challenging task since lots of musical elements need to be considered, such as pitch, chord, rhythm, musical form, and scale. In music variation, it requires strong music theory knowledge to constraint how the melody is being progressed instead of adding note randomly. Since the melody should be progressed in a harmony way and under a reasonable chord hierarchy, music variation can be considered as a rule based composing system. And in genetic algorithm (GA), the concept, "Nature selects, the fittest survives.", can also be applied in music variation. Since fitness function is a mechanism that help to direct the way how chromosome evolves. In this research, we adapt genetic algorithm to improvise given music piece and design three measurements to direct the melody evolving in a harmony way, giving it more variations but still keep the
melody structure similar to the original song so that the listeners still can identify its source.

Introduction -1
Related Work -4
Genetic Algorithm Overview-8
Methodology-11
Experiment-27
Conclusion-52

References
[1] Manuel Alfonseca, Manuel Cebri ́an Ramos, and Alfonso Ortega. Testing geneticalgorithm, recombination strategies and the normalized compression distance forcomputer-generated music. 2006.
[2] John A Biles et al. Genjam: A genetic algorithm for generating jazz solos.
[3] Jean-Pierre Briot, Ga ̈etan Hadjeres, and Franc ̧ois Pachet.Deep learning techniquesfor music generation, volume 10. Springer, 2019
[4] Gino Brunner, Yuyi Wang, Roger Wattenhofer, and Jonas Wiesendanger. Jambot:Music theory aware chord based generation of polyphonic music with lstms. InToolswith Artificial Intelligence (ICTAI), 2017 IEEE 29th International Conference on,pages 519–526. IEEE, 2017.
[5] C-CJ Chen and Risto Miikkulainen. Creating melodies with evolving recurrent neuralnetworks. InNeural Networks, 2001. Proceedings. IJCNN’01. International JointConference on, volume 3, pages 2241–2246. IEEE, 2001.
[6] Elaine Chew and Yun-Ching Chen. Real-time pitch spelling using the spiral array.Computer Music Journal, 29(2):61–76, 2005.55
[7] Hang Chu, Raquel Urtasun, and Sanja Fidler. Song from pi: A musically plausiblenetwork for pop music generation.arXiv preprint arXiv:1611.03477, 2016.
[8] Hao-Wen Dong, Wen-Yi Hsiao, Li-Chia Yang, and Yi-Hsuan Yang. Musegan: Multi-track sequential generative adversarial networks for symbolic music generation andaccompaniment. InThirty-Second AAAI Conference on Artificial Intelligence, 2018.
[9] Swati Gaur, Aastha Maheshwari, Leena Dhruwa, and Amit Upadhyay.Hiddenmarkov model and genetic algorithm based credit card fraud detection. 2017.
[10] Ali Asghar Rahmani Hosseinabadi, Javad Vahidi, Behzad Saemi, Arun Kumar San-gaiah, and Mohamed Elhoseny. Extended genetic algorithm for solving open-shopscheduling problem.Soft computing, 23(13):5099–5116, 2019.
[11] Vasanth Kalingeri and Srikanth Grandhe. Music generation with deep learning.arXivpreprint arXiv:1612.04928, 2016.
[12] Gullapalli Keerti, AN Vaishnavi, Prerana Mukherjee, A Sree Vidya, Gattineni SaiSreenithya, and Deeksha Nayab. Attentional networks for music generation.arXivpreprint arXiv:2002.03854, 2020.
[13] Oliver Kramer.Genetic algorithm essentials, volume 679. Springer, 2017.
[14] Chien-Hung Liu and Chuan-Kang Ting. Fusing flamenco and argentine tango byevolutionary composition. InEvolutionary Computation (CEC), 2017 IEEE Congresson, pages 2645–2652. IEEE, 2017.
[15] I Liu, Bhiksha Ramakrishnan, et al. Bach in 2014: Music composition with recurrentneural network.arXiv preprint arXiv:1412.3191, 2014.
[16] Jing Luo, Xinyu Yang, Shulei Ji, and Juan Li. Mg-vae: Deep chinese folk songsgeneration with specific regional styles. InProceedings of the 7th Conference onSound and Music Technology (CSMT), pages 93–106. Springer, 2020.
[17] Sanjay Majumder and Benjamin D Smith. Music recombination using a genetic al-gorithm. ICMA, 2018.
[18] Dragan Mati ́c. A genetic algorithm for composing music.Yugoslav Journal of Oper-ations Research, 20(1):157–177, 2010.
[19] Ryan A McIntyre. Bach in a box: The evolution of four part baroque harmony usingthe genetic algorithm. InEvolutionary Computation, 1994. IEEE World Congresson Computational Intelligence., Proceedings of the First IEEE Conference on, pages852–857. IEEE, 1994.
[20] Ender ̈Ozcan and T ̈urker Erc ̧al. A genetic algorithm for generating improvised mu-sic. InInternational Conference on Artificial Evolution (Evolution Artificielle), pages266–277. Springer, 2007.
[21] Jos ́e Maria Sim ̃oes, Penousal Machado, and Ana Cl ́audia Rodrigues. Deep learningfor expressive music generation. InProceedings of the 9th International Conferenceon Digital and Interactive Arts, pages 1–9, 2019.
[22] Bob L Sturm, Joao Felipe Santos, Oded Ben-Tal, and Iryna Korshunova.Mu-sic transcription modelling and composition using deep learning.arXiv preprintarXiv:1604.08723, 2016.
[23] Eric Thul and Godfried T Toussaint. On the relation between rhythm complexitymeasures and human rhythmic performance. InProceedings of the 2008 C3S2E con-ference, pages 199–204, 2008.
[24] Harun U ̆guz. A two-stage feature selection method for text categorization by usinginformation gain, principal component analysis and genetic algorithm.Knowledge-Based Systems, 24(7):1024–1032, 2011.
[25] Michael D Vose.The simple genetic algorithm: foundations and theory. MIT press,1999.
[26] Li-Chia Yang, Szu-Yu Chou, and Yi-Hsuan Yang. Midinet: A convolutional gen-erative adversarial network for symbolic-domain music generation.arXiv preprintarXiv:1703.10847, 2017.
[27] Lantao Yu, Weinan Zhang, Jun Wang, and Yong Yu. Seqgan: Sequence generativeadversarial nets with policy gradient. InAAAI, pages 2852–2858, 2017.
[28] Yi Yu and Simon Canales. Conditional lstm-gan for melody generation from lyrics.arXiv preprint arXiv:1908.05551, 2019.
[29] Damon Daylamani Zad, Babak Nadjar Araabi, and Caru Lucas. A novel approach toautomatic music composing: Using genetic algorithm. InICMC, 2006.
[30] Xiaomei Zheng, Dongyang Li, Lei Wang, Lin Shen, Yanyuan Gao, and YuanyuanZhu. An automatic composition model of chinese folk music. InAIP ConferenceProceedings, volume 1820, page 080003. AIP Publishing, 2017.