Re:Training a Recurrent Neural Network to Compose Music
by lopatin on Sunday August 21, 2016 @10:07PM
(#12333696)
Noob question: It intuitively seems to me that by feeding raw text of a structured format (such as the music notation in the article) we're making the algorithm unnecessarily learn the syntax in addition to interesting stuff, which is the high level musical patterns. What kind of results would you expect from running the same experiment, but with an input encoding more specialized to the problem domain? Would the performance benefits be significant?
[ Parent
| on HN ]
Re:Training a Recurrent Neural Network to Compose Music
by jcranmer on Sunday August 21, 2016 @10:51PM
(#12333790)
As far as music notation goes, this is very unstructured notation--asides from the initial metadata, the data boils down to hitting one of 7 notes of the scale across 2 octaves, with an optional length parameter. Reading the full documentation indicates there's more complexity, but the RNN appears to have used none of it.
That the result is largely consonant appears to primarily due to the structure of the format: if you don't know that you can create accidentals off of a normal scale, play chords, create multiple competing melodies, etc., then you can't use them to create dissonance. This means that a pure random string largely amounts to hitting random white keys on a piano, which never really sounds grating (although it doesn't sound good either, just bland)... which also happens to describe the samples that have been selected for us!
Re:Training a Recurrent Neural Network to Compose Music
by adamnemecek on Sunday August 21, 2016 @10:18PM
(#12333723)
Part of the appeal of ml is that you don't have to care about the underlying processes.
Re:Training a Recurrent Neural Network to Compose Music
by lopatin on Sunday August 21, 2016 @10:35PM
(#12333756)
We want the model to be better right? I just want to express input data using as little bytes as possible. Do you mean that if we are able to somehow express in 1 byte what used to take 100, that you wouldn't expect it to noticeably improve the model?
Re:Training a Recurrent Neural Network to Compose Music
by azernik on Monday August 22, 2016 @12:26AM
(#12334099)
Not appreciably. Neural Networks' job (in some configurations) is to approximate their input in as few bytes as possible. There's a vague equivalence between machine learning and lossy compression.
Re:Training a Recurrent Neural Network to Compose Music
by Scea91 on Monday August 22, 2016 @08:05AM
(#12335675)
There's even vague equivalence between statistics and lossy compression.
Re:Training a Recurrent Neural Network to Compose Music
by azernik on Monday August 22, 2016 @08:42AM
(#12335888)
And most modern machine learning is referred to in academia as "statistical learning".
Equivalences all the way down!
Re:Training a Recurrent Neural Network to Compose Music
by deckar01 on Sunday August 21, 2016 @11:48PM
(#12333989)
I wondered the same thing. If the encoding was completely dense such that all possible outputs were valid, would the network express more complex compositional behavior by avoiding the need for binary feedback about syntax validity?
I think the main benefits of the abc midi syntax are that it is already a very dense encoding and has an established library of music to train on.