Wolfram Computation Meets Knowledge

Wolfram Summer School


Ed Meier

Summer School

Class of 2005


Ed Meier was born in Detroit, Michigan, and raised in its suburb of Southfield. After an “adventure” of living in Huntsville, Alabama, Ed currently resides in Springfield, Ohio. Ed is married and has a two-year-old daughter. Ed spent 20 years as a professional software developer: Ed has written banking software, geographic information systems, terrain-modeling software, ATM (automatic teller machine) driver software (including the finite-state machine), and Air Force logistics systems. Ed’s proudest accomplishment is that at each company he has worked for, he has solved the “impossible” problem–the one that others shied away from due to complexity or length, or it was just inconceivable that it could be done. Ed is the one who attempted to use CA to beat the Mega Millions lotto game. (He didn’t win.) Ed applied genetics-based machine learning to search for the proper sequence of CA rules to find a specific output bit string, and continues to explore NKS as a hobby. Ed received his bachelor’s degree from Central Michigan University in 1984.

Project: Using NKS to Visualize Documents

The goal of this project was to explore NKS-based approaches to visualize and interpret documents. Finding one methodology to visualize all types of documents is challenging. Not only do types of documents differ in their style, but the English language is fraught with inferences. Context-dependent meanings are difficult for a computer to process efficiently.

The first experiment was to see the similarity between the plots of CA output interpreted as words and actual documents. The next experiment was to explore the plots of documents where each sentence was plotted as its own graph. Words in common across sentences were plotted as intersection points of the graphs. The last experiment explored the possibility of using cellular automata that produce structure from random initial conditions to organize documents.

Favorite Four-Color, Nearest-Neighbor, Totalistic Rule

Rule chosen: 49998

While this pattern does show a lot of symmetry of the right-left kind, it has a good spread of numbers, is a class 3 (borderline 4), and appears to have a minority of white spaces (which are not useful to this problem), so it has the better chance of emulating a lotto game compared to others turned up in this search. Plus it is visually interesting in the “traditional” visual analysis of CAs.