tapanauti/msc_thesis

The earliest research can be found around the 1940s and the majority of the Research that emerged over Neural Network and its application can be found after the 1980s. From then to the present, there have been significant changes in the applications and use of Neural Network and the concepts of Deep Learning. Many kinds of research emerged in all the fields about machines showing cognitive abilities, which is not new. The capabilities of performing human tasks and cognitive thinking shown by machines is what debatably AI can achieve and researches in. Over the years the research gave rise to new concepts of reproducing various activities performed by humans, one such being Music. The concept of music reproduction with the help of deep learning is as well not new, but what is new is the quality of regeneration, its similarities to original work, and also the uniqueness of melodies generated. We can say the machine truly learned and performed, along with all these things one crucial thing that is also new in today’s research is the availability of resources. This thesis has tried to achieve one similar thing where I have tried to reproduce Irish traditional folk music. Generative Adversarial Network model has been proposed for the generation where Long Short-Term Memory (LSTM) will be used as the generator and as the discriminator. ABC files have been used to train the model for music generation. This thesis aims to generate tunes that will be similar to Irish Folk Music with a model that is not specifically tuned for Irish Folk but can later be used on any type of music and generate similar tunes.

Follow the link to listen to the generated music - https://drive.google.com/drive/folders/14EMRpjFe9Wrr2mMnjLbV1q08sLtNSFwh?usp=sharing