3 ms·

Representing music in all of its relevant dimensions (scale, time, polyphony, melodic patterns, dynamics, etc.) is a very interesting and still not optimally so

by Rochus 5y ago

Representing music in all of its relevant dimensions (scale, time, polyphony, melodic patterns, dynamics, etc.) is a very interesting and still not optimally solved problem. It is e.g. relevant for machine learning based systems for musical composition or improvisation; with transformers there was a big progress made compared to nearly thirty years of stagnation. And this is not about "exotic" compositions associated with the composers you mentioned, but about any kind of popular or classic music. The "necklace" representation of scales (also used in the referenced site) is an interesting one since it allows to e.g. represent modulation by rotation and/or mirroring. You rarely meet it in traditional music education. Such representations lead to useful insights about music if one wants to deal with it algorithmically. And there are many people here (including me) who do.

thenoblesunfish 5y ago
I completely agree that it's very interesting. But it's a Glass Bead Game you're playing. A fun game, a game which can make you a better composer, even but one should not be fooled into thinking that this is really essentially to how the music you love was created. One of the few things I think I know about music is that good way to make it is to take whatever system you have and subvert it a little bit.
- Rochus 5y ago
  I'm a musician and composer myself (even used to work as a professional musician a couple of years). There are totally different approaches to composing music, from totally intuitive/non-intellectual to totally formal/rule-based. I would say that today the majority of composers compose music intuitively and directly on the instrument, without being aware of the exact processes involved. Formalizations in music are therefore mostly academic, arising from analysis of the compositions of others, as an attempt to explain how these compositions work. For algorithmic composition, such formalizations are necessary, but so far not sufficient to create a high-quality composition (just as one could not become a good musician just by studying music theory). Nevertheless, it is necessary to study such formalizations if one wants to advance the field of algorithmic composition. However, I consider most "classical" approaches to formalizing music obsolete for this purpose.

by Rochus 5y ago

thenoblesunfish 5y ago

I completely agree that it's very interesting. But it's a Glass Bead Game you're playing. A fun game, a game which can make you a better composer, even but one should not be fooled into thinking that this is really essentially to how the music you love was created. One of the few things I think I know about music is that good way to make it is to take whatever system you have and subvert it a little bit.

Rochus 5y ago
I'm a musician and composer myself (even used to work as a professional musician a couple of years). There are totally different approaches to composing music, from totally intuitive/non-intellectual to totally formal/rule-based. I would say that today the majority of composers compose music intuitively and directly on the instrument, without being aware of the exact processes involved. Formalizations in music are therefore mostly academic, arising from analysis of the compositions of others, as an attempt to explain how these compositions work. For algorithmic composition, such formalizations are necessary, but so far not sufficient to create a high-quality composition (just as one could not become a good musician just by studying music theory). Nevertheless, it is necessary to study such formalizations if one wants to advance the field of algorithmic composition. However, I consider most "classical" approaches to formalizing music obsolete for this purpose.