There are several methods for creating music artificially, the best known being GAN, LSTM and VAE. Let’s explore the last two!
LSTM for Long Short-Term Memory: This is a stack of layers of neurons. Each neuron modifies the data it receives and transfers it to the next layer. So as not to lose context, all the data is analyzed together by all the neurons in a layer.
VAE for Variational Auto Encoder: Auto encoding is the combination of two antithetical algorithms. The first learns to compress data into latent code, while the second learns to decompress this latent code. Learning is said to be successful when the margin of error is considered negligible. In the field of music, VAEs can be used to create AIs in the style of certain composers.
