Catherine Sibert, RPI Graduate Student

 

Catherine Sibert, RPI Graduate Student

Sage 4101

March 1, 2017 12:00 PM - 1:00 PM

 

The game of Tetris is one of the most widely played games in the world, and in addition to providing entertainment to its many fans, its nature as a complex, dynamic, decision-making task makes it an ideal testbed for a variety of research domains. Machine learning Tetris research has found that very simple models are capable of high level performance, but do so by adopting a particular, and often unhumanlike, behavior. Previous work found that the behavior of these models could be affected by altering the objective function, or goal, of the task but did so by artificially limiting the length of the test games, meaning that the full capabilities of the models were never completely explored. This research aimed to determine the how game length and other changes to the task environment affected the behavior of models, as well as how the best models reacted to those changes.

 

 

Add to calendar
Share|