Lab Meetings
From NEBL
This is the speaker schedule for the joint NEBL - ECSL laboratory meetings.
Spring 2008
Meetings are in the AGN from 11:30-1:00, unless noted otherwise.
(Those marked as "short" will only run from 11:30-12:00.)
May 2 - No meeting (CSE Open House - have a demo ready!)
Apr 25 - Mark Harmer
Apr 18 - short
Apr 11 - Ryan Leigh, "Using Coevolution to Understand and Validate Game Balance in Continuous Games"
Apr 4 - Nathan Penrod, "Neuro-evolving Maintain-Station Behavior for Realistically Simulated Boats" (CEC'08 practice talk); also, everyone will show their demos on the new big-screen system in the lab.
Mar 28 - No meeting (Spring break)
Mar 21 - Matt Parker, "Neuro-visual Control in the Quake II Environment" (IJCNN'08 practice talk)
Mar 14 - short; Plan Engineers' Day Demos for Thursday, April 10, 9:30-1:30
Mar 7 - Plan Engineers' Day demos
Feb 29 - No presentation.
Feb 22 - short, followed by colloquium at noon -
- Evolutionary Learning Algorithms for Intelligent Systems
- Gordon K. Lee, San Diego State University
- MS 227
Feb 15 - short
Feb 8 - Amit Banerjee - "A Computational Model for Creative Design using Collaborative IGAs"
Feb 1 - short
Jan 25 - short (Plan semester schedule)
Fall 2007
| December 21 | - | Open (last day of semester)
|
| December 14 | - | Open (finals week)
|
| December 7 | - | Open (CS&E videogame party day)
|
| November 23 | - | Meeting Cancelled (Thanksgiving Holiday)
|
| November 16 - |
| Justin Schonfeld
|
| October 5 | - | Meeting Cancelled (Nevada Day)
|
| October 19 - |
| Nathan Penrod
|
| Title of Presentation : Creating a maintain station behavior using neuroevolution.
|
- Nathan's presentation was about creating fitness functions and sensor configurations to simplify the problem of evolving a neurocontroller capable of performing the maintain station task.
|
| October 12 - |
| Anil Shankar
|
| Title of Presentation : Sycophant A Context Based Generalized User Modeling Framework for Interactive Interfaces
|
- We present a user-context based personalization framework, Sycophant, for interactive interfaces. Sycophant employs simple sensors to enable an interface interface such as Google Calendar for learning a user's alarm preferences. results from user studies indicate Sycophant can successfully predict user preferences for different application actions. Our investigation also highlights the importance of external contextual information from a user's environment for interface personalization.
|
| October 5 | - | Meeting Cancelled
|
| September 28 - |
| Shane Warren
|
| Title of Presentation : EPSCoR Grant Proposal - "Contextual Control System"
|
- Use of artificial neural networks (ANN) in real-world applications has shown limited success. One of these reasons is due to how much a single ANN can generalize. Shane suggests the need for a higher level control network that could be trained to pick a previously trained network given the context of the input. This approach would allow for incorporating ANNs with multiple behaviors without the risk of approaching our limit of generalization.
|
| September 21 | - | Meeting Cancelled
|
| September 14 - |
| Amit Banerjee
|
| Title of Presentation : A Recursive Clustering Methodology using a Genetic Algorithm
|
- Amit presented a recursive clustering scheme that uses a GA-based search in a dichotomous partition space for data clustering. The proposed algorithm makes no assumption on the number of clusters in the data set, instead it recursively uncovers subsets in the data until all isolated and compact regions in the data have been identified. A test of spatial randomness serves as the termination criteria for the recursive process, and a silhouette width-based distance function is used as the fitness function for the GA. It was also shown that results of clustering on test cases ranging from simple 2-D to large multidimensional data sets compare favorably with state-of-the-art techniques in data clustering. The presentation was based on my CEC 2007 paper by the same name.
|
| September 7 - |
| Ryan Leigh
|
| Title of Presentation : Ph.D. Proposal
|
- A.I. in many games provides a static opponent, in that the strategies used by the A.I. doesn't change. I propose a system in which the A.I. learns how the player plays and then can develop new strategies to counter them.
|
| August 31 | - | Chris Miles
|
| Title of Presentation : From the Apocalypse to Entropy, a user's manual
|
- Chris presented a brief overview of the design justifications and some basic implementation within the entropy game engine.
|
| August 22 | - | Matt Parker
|
| Title of Presentation : An Overview of Xpilot-AI
|
- Xpilot is a 2-D multiplayer space-combat game that has been modified to allow for artificial intelligence control. Xpilot-AI has been used by researchers as a platform to test various controllers such as neural networks and cyclic genetic algorithms. Because the game cannot be speeded up to use the maximum CPU rate of a machine without dropping frames and altering gameplay, a Queue Genetic Algorithm was developed to easily distribute the evolution, while still yielding similar evolution results as a regular genetic algorithm. The Core was also developed, which is a system in which all the agents in a population are fighting in a large map, and whenever an agent is killed it becomes a new agent that is a crossover between its killer's genes and its own genes.
|
![[Main Page]](http://nebl.cse.unr.edu/logo.png){kind=logo}