Summer Institute |
Summer Institute 2003• Overview
• Participants • Projects
• Overview Students gained an increase in computer science knowledge, new friendships, an enhanced ability to work on a team, a unique opportunity to use state-of-the-art supercomputers, and exposure to campus life. Students left SI2003 with a greater confidence in their individual learning abilities and their ability to adapt while working and living in a new environment with new friends of similar interests.
Projects The Mechanical Engineering project consisted of modeling the
stress response of cars and the impact upon their passengers (crash dummies) upon impact. The students built
the cars with a finite element model with the Altair Hyperworks software package. The Parallel Processing project was designed
mainly to teach the parallel processing language needed to split up a problem and assign different tasks to
different processors. Their parallel code was run on supercomputers containing 128 and 256 separate processors.
They simulated the genetic evolution of organisms using random gene mutations, mating, survival, and conflict.
The parallel processing allowed simulation of 30,000 generations in just one hour. In its second year, the Neural Network project combined planning, computer programming, image processing, and Artificial Neural Network pattern recognition
techniques. The group of three students and two teachers first were given tools they needed to complete their
project: they learned the C programming language and the basic principles of Artificial Neural Networks,
specifically, the mathematical equations and steps used in a pattern learning technique known as "Feed
Forward Training" and those used in an error reducing algorithm known as "Back Error Propagation." The
students built an application capable of pattern recognition that works in ways very similar to how neurons
in the human brain interact in learning. The group wrote a program that recognized only the letters in one
of the group member's name, 'T' 'o' 'n' 'y'. To verify later that the program worked, several other SI
attendees provided samples of their 'T' 'o' 'n' and 'y' to be tested against the neural network trained
only to recognize Tony's handwriting. To accomplish this task, the group broke into two teams; one team
wrote the C code, the other took writing samples and processed the images. Once work from both teams was
completed, they collaborated to train the network and test it with their samples. The project was successful
since the neural network only slightly recognized others' handwriting as Tony's The VRML / MOCAP project gave the group the
opportunity to work in the high tech Motion Capture Laboratory at the Advanced Computing Center for the Arts
and Design. The project team scripted a short story about a tour through a haunted house. The environment
involved the exterior and interior of the house. The keyframes of the story were rigorously defined. In the
MOCAP Lab one team member wore a special suit, which had 41 reflecting balls attached at critical points of
the body. This one actor performed all the motions for all the characters. 14 specially designed cameras
captured the action of the reflective balls. The position of the individual reflective balls was calculated
and placed in virtual space. The team designed the haunted house using VRML. Other programming tools were used
to put the marker positions in the house. The marker locations were then associated with a character, a skeleton,
which jumped out during the tour. Click here to download a detailed schedule. |
