MLDemos visualizes what classifiers do
MLDemos is based on a really nice idea – to visualize how different classifiers construct the decision boundaries around arbitrary sets of data points. I had of course seen the concept of decision boundaries before; in many machine-learning classes you will draw or at least get to see boundaries or surfaces that delineate the parts of the sample space where a classifier will yield different predictions. In MLDemos, you get to draw the points in the (2-D) sample space by hand, and you can choose between a variety of different algorithms. Or if you want, you can upload your own data sets. The software doesn’t just do decision boundaries, it also visualizes regression, clustering and dynamical systems in cool and downright beautiful ways.
