Fig. 3.

Combining multiple similarity scores with a quadratic classifier. Each point represents a comparison between 2 units on 2 different days. We computed 4 similarity scores as described in Tracking the same neurons. These plots show projections of 2 scores at a time. Points are labeled according to whether they were classified as the same neuron. The same neuron/different neuron Gaussians estimated from the data are shown as contour plots. A 2-dimensional slice of the decision boundary that is used by the classifier is shown as a black line. The contours correspond to 25%, 50%, 75%, and 95% of the distribution. There are 6 unique combinations that could be shown; we chose the pairwise/wave scatterplot because they are the 2 most informative features, and we chose the mean rate/wave scatterplot because it illustrates the unique characteristics of the change-in-mean-rate feature.