Description

Genomata is an application that spans different topics, including Music Information Retrieval (exploiting the Spotify API for Developers), Genetic Algorithms, 3D Cellular Automata and Particle Systems.

The initial population is chosen by randomly selecting 10 songs from the tracks in a user-selected playlist. The fitness function is computed for every track in the population. A new population is generated, by applying the next steps until the population is completed. Parents are selected according to their fitness function: the more the song has been listened to, the higher the fitness function, the higher the probability of that song to be selected as a parent for the generation of a new child. The new child is the result of the crossover between the two selected parents: since we are using track’s features as genes, we use an arithmetic crossover function, which generates the new child song by computing the arithmetic mean of each of the parents’ features. To introduce some variability into the new population, mutation is performed. In this case, since we are working with real values, mutation consists in adding/subtracting a small number to each gene of the new child, with a defined (small) mutation probability.

Challenges, accomplishment and lessons learned

Technology

Python, Processing

Quotes

Media