Two projects in music and artificial intelligence

By Søren Tjagvad Madsen.

The Palestrina project (2002)

An article about making music in the style og G. da Palestrina. The system is able to make two part 1st, 2nd and 4th species counterpoint. It uses an evolutionary algorithm for evolving the counterpoint candidates. The fitness function is based on Knud Jeppesen's book, "Counterpoint, vocal polyphony". Download the article here: pdf.

The harm project (2002)

A paper about training a neural network for learning chord progressions, and voice leading rules. The network was trained with hundreds of Bach chorales, to learn the art of chorale harmonisation! Read about the experiment here: pdf. In collaboration with Martin Elmer Jørgensen.


Machine Learning in medicine

Den Århusianske Fødselskohorte (2004, in Danish)

Questionnaires from about 15000 pregnant women and data about their children were mined for finding relations between the condition af the mother and the newborn child. Download pdf.

The data set was kindly lend out by Perinatal Epidemiologisk forskningsenhed, Skejby Sygehus, The University Hospital of Århus.