Simple Pattern Spectrum Estimation for Fast Pattern Filtering with CoCoNAD
- Cite this paper as:
- Borgelt C., Picado-Muiño D. (2014) Simple Pattern Spectrum Estimation for Fast Pattern Filtering with CoCoNAD. In: Blockeel H., van Leeuwen M., Vinciotti V. (eds) Advances in Intelligent Data Analysis XIII. IDA 2014. Lecture Notes in Computer Science, vol 8819. Springer, Cham
CoCoNAD (for Continuous-time Closed Neuron Assembly Detection) is an algorithm for finding frequent parallel episodes in event sequences, which was developed particularly for neural spike train analysis. It has been enhanced by so-called Pattern Spectrum Filtering (PSF), which generates and analyzes surrogate data sets to identify statistically significant patterns, and Pattern Set Reduction (PSR), which eliminates spurious induced patterns. A certain drawback of the former is that a sizable number of surrogates (usually several thousand) have to be generated and analyzed in order to achieve reliable results, which can render the analysis process slow (depending on the analysis parameters). However, since the structure of a pattern spectrum is actually fairly simple, we propose a simple estimation method, with which (an approximation of) a pattern spectrum can be derived from the original data, bypassing the time-consuming generation and analysis of surrogate data sets.
Unable to display preview. Download preview PDF.