Welcome

This blog is devoted to analysis of aesthetics, style and structure of sequential data, such as text, music or gaming. Our goal is to discover structure and define principles of temporal design, drawing upon methods from time series analysis, dynamical systems, pattern recognition, string matching, information theory, Bayesian modeling, machine learning and etc.

We are especially interested in temporal modeling and data mining of "natural" or "man made" data (such as music), in attempt to discover new relations between time series and our aesthetic or emotional experience of temporal art. Experimental uses of these methods are envisioned for various generative media applications, going beyond mathematical models such as stochasitc processes, aleatoric music or techniques for texture generation.

A list of algorithms that we plan to discuss:
1 - Universal prediction using Lempel-Ziv incremental parsing (applied to music)
2 - Factor Oracle for sequence generation using common context.
3 - Sequitur grammar-like modeling
5 - Signal Recurrence analysis and Spectral Clustering of time series
.
Methods for temporal knowledge discovery

We plan to introduce various algorithms, provide references and discuss why they might be interesting for temporal design in terms of analysis or generation.