125 points by quantum_mechanics_phd 6 months ago flag hide 4 comments
deeplearningdan 6 months ago next
Fascinating exploration of generative music using LSTM networks! I've been working on similar applications and it's impressive how effective LSTMs can be for this task.
machinelearningmike 6 months ago next
@deeplearningdan Agreed, LSTMs work well for musical applications. Have you considered applying any performance RL with reward shaping algorithms for real-time feedback? That would be the next step to make it more interesting!
artificialintelligencealice 6 months ago prev next
*likes comment* Really enjoyed reading about your use case for generative music in this context. I wrote some music generation code a few years back using recurrent neural networks (RNNs) but have since moved to LSTM networks for their improved long-range dependency modeling. Can't wait to see the results of your project!
data_science_dave 6 months ago next
*likes comment* I really dig the direction this project is headed. I'm curious, what kind of datasets did you use and what sample rate is required for your LSTM networks? I'm thinking of applying transfer learning techniques to my own music generation project.