Approximate programming by Chris Kiefer

Coding by weird interfaces.

Approximate coding. Take a number generator and turn it into code. Any number generator will work. Use gene expression programming. It’s fast but not precise.

What is the value of code as a medium? How precise does code need to be in a live situation? Is the keyboard the best way to write code?

Genetic programming uses a tree representation. Numbers are translated into the tree. Component functions plus arrays of numbers make new code to run.

He’s got jit nodes in SuperCollider and a multislider. Changing the faders changes the synthdef.  You can live code the Synth functions.

A controller could give you your arrays.

There are two levels of live coding. You can create your autonomic functions and the trees created by generating numbers.

You can discover relationships between sound and gesture. The code is kind of incomprehensible but can be out into trees.

Smoothing out parameter space. Small changes can cause static output changes due to non linear ness.

Future work involves smoothing landscapes. He wants some bidirectional feedback.  Could code output make new code?
Can this system do anything precise? Could you find a gradual quick sorry  algorithm?

Q: is there room for genetic processes via evolution?
A:

Q: how duo you compare two trees? How do make the minimal changes on transition?
A: this may be a good way to smooth a transition
Q: diffmatch patching.

Published by

Charles Céleste Hutchins

Supercolliding since 2003

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.