Algorithm Optimization
This one's not quite as easy to see.
Any algorithm or function, however arbitrary, can be represented as a simple mapping. The input parameters can be thought of as an input vector, which is mapped to an output vector. Therefore, the mapping exists in a vector space with the number of dimensions equal to the number of input parameters. Each point in the vector space represents one input vector. Thus, at each point there is an output vector. So it's just a big matrix.
To evaluate the algorithm or function, we need to look up the
output vector corresponding to the current input vector. So, we need to represent the matrix somehow.
In other words, algorithm optimization is simply representation with the goal being to have as fast a representation as is
feasible. See