This is how I plan to do it.

Inputs: virtual memory layout, current context, and current instruction
Output execution state.

avoid clarifying the change and just declare the action. ie teach it how to emulate the instruction.

mov ecx,eax

This could translate to [2] - > [1] or some other representation of your choice.

A random instruction generator could then be used to generate all known instructions and collect execution maps of each instruction using a single stepping approach.

Currently I am wring a NLP categorization tool that looks up a verifies with references what it thinks a item should be categorized under, I have been studying machine learning by

taking classes and reading as much as I can get my hands on. Which isn't to say I am a expert, I am just beginning to understand the math and algorithms involved and would

greatly appreciate your input and idea's on implementing something such as this.