I am currently reading:
Stephen Jose Hanson: Meiosis Networks, 1990.
and I stumbled about this:
It is possible to precisely characterize the search problem in terms of the resources or degress of freedom in the learning model. If the task the learning system is to perform is classification then the system can be analyzed in terms of its ability to dichotomize stimulus points in feature space.
Dichotomization Capability: Network Capacity Using a linear fan-in or hyperplane type neuron we can characterize the degrees of freedom inherent in a network of units with thresholded output. For example, with linear boundaries, consider 4 points, well distributed in a 2-dimensional feature space. There are exactly 14 linearly separable dichotomies that can be formed with the 4 target points. However, there are actually 16 ($2^4$) possible dichotomies of 4 points in 2 dimensions consequently, the number of possible dichotomies or arbitrary categories that are linearly implementable can be thought of as a capacity of the linear network in $k$ dimensions with $n$ examples.
What is a "dichonomy" in this case?
(Side questions: what is a fan-in type neuron?)