Algorithmic Music Generation
Music is mainly an artistic act of inspired creation and is unlike some of the traditional math problem. Music cannot be solved by a simple set of formulae. The most interesting and challenging part is producing unique music without infringing the copyright. The generated music has to sound good, and what sounds good is very subjective. The model is mainly built for producing Indian music.
Artificial Neural Network/Deep Learning has wide range of application, such as in Image processing, Natural language processing, Time series prediction, etc. But what about its usage in art? This talk shows how deep learning was used to generate music.
Deep learning attempts to model high-level abstractions in data by using multiple processing layers whereas the traditional Machine learning algorithms emphasizes on what work the computer program must do after it is given a data set. The basic idea in deep learning is to perform an unsupervised learning procedure on every single layer in addition to using gradient descent for the network as a whole. The goal of the unsupervised learning is to make each single layer extract characteristic features out of its input that can be used by subsequent layers. The neural network architecture makes use of numerous amount of Indian music to train the model. After adequate number of iterations and training time, this model generates music that is unique and original.
This talk will also cover some of the challenges and trade-offs made for algorithmic music generation.
Padmaja V Bhagwat is currently pursuing 3rd year of B.Tech in Information Technology at National Institute of Technology Karnataka. The academic projects undertaken by her include developing a Job portal system, developing a chat application and implemented handwritten digits recognizer using neural networks. Currently as a part of her summer internship she is working on the Algorithmic music generation project.