Consider a vector of real-valued features. A decision tree is a classification structure where each node represents a feature of the input data and edges to children nodes represent a comparison...

Consider a vector of real-valued features. A decision tree is a classification structure where each node represents a feature of the input data and edges to children nodes represent a comparison operation against the feature for each datum. Leaf nodes represent classes. Thus, to classify a numerical vector, the tree is traversed from root node to leaf node, following edges based on the comparison between vector components and node feature, until a leaf node is reached. Assuming a binary, less-than operator at each node, against some given constant, what is the geometric interpretation of decision trees, and how does it compare to (non-kernelized) SVMs?