Most Magenta models make use of LTSMs which are a type of RNN that are better at learning long term structure as present in music. Recurrent neural networks (RNN) are a popular choice for sequence generation but they can be used for classification and prediction as well. A priming sequence (a piece of music) is provided as an input/seed to condition the model and it attempts further sequence generation in that style. A model is trained on a large number of musical sequences and becomes capable of generating music on its own. Generation of music can be thought of in AI terms as a sequence generation task. It is built using TensorFlow (TF), a popular Python-based open source library for machine intelligence.
_1599372487656_1599372494361.jpg "google project magenta")
Magenta is a Google project whose goal is to use machine learning for generating compelling music and art.
0 kommentar(er)
