Week 1
We've spent our first week getting acquainted with the campus, meeting fellow members of our research group, and soaking up the greens and grays of the Pacific Northwest. We are currently discussing ideas for summer research topics, including studying the game of Go, or some application of computational biology.
We will also join the EECS department's graphics research reading group, which meets weekly on Fridays. This week's discussion was a continuation of a presentation about Discrete Exterior Calculus.
Week 2
This week, we met with members of the Planteome Project to discuss potential summer research topics involving computational biology. We were given a brief introduction to the field of biological ontolgy, and were told about several portions of the Planteome Project to which we might be able to contribute. The most promising candidate project involves redesigning the database used for AmiGo, a gene ontology analysis tool. We have tentatively decided that I will focus my research on the game of Go, and the other student under Dr. Zhang's supervision will focus his efforts on the planteome database project.
This week's reading group discussion introduced the concept of Mixed Integer Quadrangulation. A cornerstone of computational geometry, Mixed Integer Quadrangulatoin provides a method for transforming triangular meshes into quad meshes.
Week 3
Dr. Zhang and I have finalized our topic. An important skill for developing as a Go player is the ability to visualize move patterns and recognize common sequences of play. Our research intends to observe the effect of presenting multiple perspectives of a game (e.g. the player’s view and the opponent’s view) on the playing habits of Go players. We hope to observe which views are the most beneficial (or distracting) to the players, and whether players of varying skill levels interact with the interfaces differently. I will be creating this desired interface, connecting it to a popular online Go playing server, and observing how users interact with our model.
During this week's reading group meeting, we continued our discussion of Mixed Integer Quadrangulation.
Week 4
We have begun the process of building a tool to collect data that will help us test our hypothesis. I have designed a mock-up interface to get a better idea of what the final product will look like.
Week 5
Many go players engage in their hobby by connecting to online go servers, such as IGS or KGS. These servers allow players to connect with opponents from all over the world. In order to collect as much data as possible, our go
playing interface should allow players to log in to an online go server, and play games with live opponents, and we are currently working on adding this functionality.
Week 6
So far, we have generated four views of the same Go game, and connected the interface to an artificial intelligence computer player. Initial testing of the interface seems promising. However, there are still many features we would like to add to our game before we release it to the public.
Week 7
This week, we got some initial feedback about our interface design. We provided the link to a prototype site to a few go players. The feedback they gave us was promising and encouraging. They noticed that the rotation of the board actually affected the way they played the game. For example, in one rotation of a game, they felt like they were in a strong position. However, a different rotation of the exact same board gave a different vibe, and caused them to see a different set of strengths and weaknesses. One potential problem we noticed was that, without prompting to switch, users might tend to use one board only. It is natural to want to finish a subgame before switching the board. Switching boards in the middle of a battle would be similar to reading 4 books in parallel by switching books every sentence. Promising results. It will be interesting to figure out exactly why the users see such drastically different points in the game.
Week 8
This week, we were assigned a paper by our professor to present at our weekly research group meeting. The paper, "Scalable Visualization of Semantic Nets Using Power Law Graphs," (Hussain) can be found here. It described the traditional methods of laying out node-and-edge graphs. A semantic net, also known as an ontology, is a system of organizing and codifying all of the known information about a particular domain. The nodes can be thought of as entities and the edges as relationships between the entities. This topic relates to our work with the Planteome project, which is focused on organizing biological ontology information. In order to see the structure of ontologies (e.g. determine related groups and the associated hierarchies), we often want to display them as graphs. Because most ontoliges are assosciated with very large graphs, visualizations often suffer from issues with readability and compuational complexity. The paper suggested taking advantage of the power law to partition graphs in to power nodes (nodes with a high degree) and non-power nodes. Because power nodes usually have a low frequency in ontology graphs, we are able to improve computational complexiy, allowing us to work with larger ontology graphs in a reasonable amount of time. Though the paper lacked the mathematical rigor expected of an academic publication, the ideas presented in it were novel. Overall, the paper sparked my curiosity in the field and the presentation was a great opportunity to practice my public speaking skills.
We also presented another novel game visualization for chess, found here.
Links to our presentation can be found Visualizing Semantic Nets Thinking Machine
Week 9
As our research experience nears its finale, we are beginning to prepare our final deliverables. We updated our website with several features that most Go players are accustomed to, such as highlighting the previous move, and playing on a realistic-looking wooden boards. I have begun drafting a final report for the Go project. Along the way, I've been reading lots of interesting papers about computer Go research, and learned about some interesting computer science techniques that have spawned from this research and have real-world applications. We've also begun work on a final design document for the Planteome database. We reviewed the original grant proposal, and crafted our final design based on the intial requirements, the meetings we've had with planteome project personnel, and feedback we've received on previous designs.
Week 10
Our last week with the DREU has consisted of finalizing our design documents and meeting with our advisor. We met with Eugene over lunch and discussed grad school, his personal experiences with research, and potential career paths. Overall, this summer has been full of valuable experience. I've gotten a real taste for what the research process is like, and how it will differ from work in industry. I'm extremely thankful for the opportunity provided to me by the DREU.