The paper discusses the idea of a system that is given a small set of
vector images, and proceeds to find new images based on the input images
and a few answers from the user to "yes/no" questions.

The system depends on the fact that all input images are "vector" images,
and that images have the same number of curves and control points.

The basic idea is to do interpolation between the input images. And then,
the user answers are used to "cut off" parts of the interpolation space.
For each "no" ansewr the user gives (the image doesn't look good), a chunk
of the interpolation space is cut off. An algorithm similar to binary
search is used to minimize the number of questions you need to ask the
user. Once the user gets bored and refuses to answer any further
questions, the system used the set of answers given to construct the space
of legal positions (The simplicial complex). Finally, this can be used to
construct some cool tricks. Like morphing from one image to the other
(using Dijkstra's shortest path algorithm to find the shortes "distance"
between the 2 images. There is an interesting definition for "distance"
in the paper). Other applications include generating new
images, finding the closest match to a given vector drawing, and modifying