Description | Document | Platform | Program |
Music composer based on
two-layer artificial neural network. Music composition is a time-series
prediction task. Net is trained on a melody from which it extracts
regularitites of note progressions. Then it tries to create a new melody.
Error backpropagation network is used.
Keywords: Artificial neural networks, music. |
PDF /
PS Doc |
Java |
Screenshot Download Music |
GenAMusic is a user-friendly
interactive genetic algorithm based music composer. Here, the GA generates
short melodies or you can give allready generated melodies and the user assigns
the fitnesses. The iterative stepping, mutation frequency and number of
crossover points are all controlled by the user. Each series of musical notes
is represented in binary form. This provides a solution space with
approximately 13^NumOfNotesInMelody possible melodies. An unintelligent
search for an ideal melody would require testing every possible solution.
In theory, the GA should be able to find a near optimum melody after searching
only a fraction of the solution space. As the result new simple melodies and a
new variation of Beethoven's "Moonlight Sonata" has been produced.
Keywords: Genetic algrithms, music. |
PDF /
PS Doc |
Java |
Screenshot Download Music |
Food gathering robot. It is
walking in the two-dimensional grid world. Each cell can be the wall, the
grass, the food or the mine. Each program takes as input the sensory data and
can do only one step per execution. So to make robot walking we have to
execute each program many times. Fitness: the total count of gathered foods
during programs runs. In particular wall-following robot is considered.
Keywords: Genetic programming. |
PDF /
PS Doc |
Java |
Download Applet |
Ant Colony Optimizations (ACO).
ACO simulates artificial ant colony behavior for searching the shortest path
to food. Each ant doesn't have any memory. It has only Tabu List for storing
the previous ant's state (Tabu Search). Ants communicate with each other by
laying pheromones on the ground.
Keywords: Ant colony optimization. |
- | Win32/C++ |
Screenshot Download |
Tic Tac Toe (Naughts and Crosses)
game (3x3 board) playing based on classifier system (LCS). Classifier system
can play versus another LCS, versus expert system, versus random player and
versus human. Different stratagies can be assign during LCS learning due to
two levels of rewards (win and draw). Bucket brigade algorithm is used for
rules chains learning.
Keywords: Learning classifier systems. |
PDF /
PS Doc |
Win32,Unix/ Ada95 |
Screenshot Download |