Classification of Noisy Weizmann HOF snippets

We created noisy versions of our HOF versions of the Weizmann snippets and tested the ability of the model to correctly classify the noisy versions, having been trained (single-trial) only on the orginals. Basically, the model gets at least 95% accuracy in this case, which is a "train != test" condition. Below we show an original HOF snippet (for daria_bend) and below it, a 40% perturbed version of it. As you can see, these are very different. Nevertheless, Sparsey correctly classifies the noisy versions almost perfectly. This gives us much increased confidence that merely developing a better preprocessing filter, i.e., that produces somewhat higher ratios of between-class to within-class variance, will enable Sparsey to do well on the benchmark Weizmann classification task. Apologies for low video quality...these the frames are 24x120 binary pixels.