Neil Parker

This was an open-ended project for my Artificial Intelligence class in Fall 2013. I worked on this project throughout the semester with two partners. The premise of this project was to optimize the placement of defensive units in a virtual population, given a fixed tile-based map that spawned enemies. Specifically, the environment consisted of gatherers that collected food for their colony, guards that defended these gatherers against enemies, enemies that spawned from fixed sources, food… Show more

This was an open-ended project for my Artificial Intelligence class in Fall 2013. I worked on this project throughout the semester with two partners. The premise of this project was to optimize the placement of defensive units in a virtual population, given a fixed tile-based map that spawned enemies. Specifically, the environment consisted of gatherers that collected food for their colony, guards that defended these gatherers against enemies, enemies that spawned from fixed sources, food sources, and walls. On simulation, gatherers moved between food sources and their “base”, enemies chased and killed gatherers, and guards killed enemies that came within their fixed range circle. We created a GUI to run the simulation and algorithm, view results, and create custom maps.

To optimize guard placement, we used a genetic algorithm that evaluated the fitness of a particular placement scheme and evolved successive generations by crossing the most successful placements in each generation. We customized the algorithm to fit this specific case by defining fitness, crossing, and mutation functions, and incorporating simulated annealing to reduce the magnitude of mutation across generations to speed up the convergence guard placement.

Our algorithm produced surprisingly successful results, showing a clear convergence towards the optimal guard placement for any given map, including hand made and randomly generated maps. For more details, see the project report below.

Project Report:
https://drive.google.com/file/d/0B6J1RIx1oLu4UHBTXy1fS3dEMlU/edit?usp=sharing

Github:
https://github.com/pr342/StrategyEvolution Show less

We predict IMDB film ratings using a number of regression models, including Ordinary Least Squares, Lasso Regression, Ridge Regression, Logistic Regression, and Support Vector Regression. We also explore kernels and several feature selection techniques such as recursive feature selection and a custom feature scoring method.

We collected and cleaned data from three online movie databases. Movies features are comprised of attributes such as director, cast, budget, runtime, and… Show more

We predict IMDB film ratings using a number of regression models, including Ordinary Least Squares, Lasso Regression, Ridge Regression, Logistic Regression, and Support Vector Regression. We also explore kernels and several feature selection techniques such as recursive feature selection and a custom feature scoring method.

We collected and cleaned data from three online movie databases. Movies features are comprised of attributes such as director, cast, budget, runtime, and genre.

Our results indicate that Ridge Regression using a polynomial kernel performed the best, giving an mean average prediction error of 0.65 on IMDB's 1-10 scale. We were also able to discover which attributes were the most influential; interestingly, runtime was the attribute most positively correlated with ratings.

Project report:
https://drive.google.com/file/d/0BwfNnba-6tCEUGFLUFF3M3dwclk/view?usp=sharing

*Web app soon to come Show less

- Implemented a computer game in C# with Microsoft's XNA framework
- Created Web-based Level Designer

Pathogenesis is a unique action-arcade game in which the player takes on the roles of a virus that has entered a host and seeks to infect it while combatting the body's white blood cells. The player controls the virus and converts enemies by firing infectious particles, building a horde of virus minions that attack other enemies. The player must traverse the body in various stages… Show more

- Implemented a computer game in C# with Microsoft's XNA framework
- Created Web-based Level Designer

Pathogenesis is a unique action-arcade game in which the player takes on the roles of a virus that has entered a host and seeks to infect it while combatting the body's white blood cells. The player controls the virus and converts enemies by firing infectious particles, building a horde of virus minions that attack other enemies. The player must traverse the body in various stages, infecting the boss of each level to complete it, eventually infecting the host's brain for victory! Show less