Music Generation Using Recurrent Neural Network

This project employs a form of LSTM networks as the special variant of Recurrent Neural Networks for producing musical compositions by learning from patterns in pitch, rhythm, and duration. A trained model on MIDI data produces a complete sequence given a seed, with this type of model always capable of producing stylistically coherent, creative outputs. Focus includes the fine-tuning parameters in such a way that there would be a minimal repetition and logical progression in compositions. The effectiveness of the approach may be assessed through quantitative criteria such as accuracy and loss, but also through feedback from listeners, which validates the capability of AI models to enrich music composition or creativity.