Sigmoidal Resource Allocation on MIND

This is a simulation/experiment of using sigmoidal programming for resource allocation optimization on MIND

This solves the problem on a single resource (cache), assuming that throughput = f(cache) can be modelled as a sigmoidal function, thus permitting simulation as a sigmoidal programming problem. A more complete writeup can be found in the pdf, but the gist is that we use SigmoidalProgramming.jl to solve problems of the form:

$$\begin{align*} \max \sum_{i=1}^n f_i(c_i) \\\ \text{subject to } \sum_{i=1}^n c_i = C \end{align*}$$

where $f_i$ is sigmoidal.

The simulation is currently running on simple sigmoids, not a differnece of sigmoids of of a sigmoid minus convex function.

Usage

The simulation is found in src/MIND_sigmoidal_programming.ipynb and needs a julia installation to run. The current version 1.6 will suffice, and can be run as any other julia notebook. Instructiosn can be found here: https://julialang.github.io/IJulia.jl/stable/manual/running/