fractal analysis doesn't work (simple lack of nonzero autoco

Questions, answers, discussions related to the FIND Technologies Inc. challenge

fractal analysis doesn't work (simple lack of nonzero autoco

Postby ponec » Wed Nov 16, 2011 7:51 pm

fractal analysis doesn't work (simple lack of nonzero autocorrelation), random forest (based on descriptive statistics like quantiles, variance and more complicaded ones) doesn't work (35% accuracy on training set), k-nearest neighbor (with clustering) based on p-values of Kolmogorov–Smirnov test (null hypothesis is that two samples are drawn from the same distribution) doesn'r work. The last approach is quite interesting because the lack of significant autocorrelation structure is the clue that information about the classes is coded in shapes of empirical distributions if that is true then Kolmogorov–Smirnov test should tell us if two observations are from the same class, unfortunately that approach doesn't work
ponec
 
Posts: 10
Joined: Fri Feb 25, 2011 11:28 am

Return to Materials Identification Based on Measurements of Electromagnetic Radiation

Who is online

Users browsing this forum: No registered users and 1 guest

cron