Bryan E. Smith

Soar is a "general cognitive architecture for developing systems that exhibit intelligent behavior" (source), with features such as reinforcement learning, semantic memory, and episodic memory.

My work has exclusively focused on semantic memory, which is the storage and retrieval of facts about the world and their meaning. I have designed tools to use ontologies such as SUMO, and have designed test agents to demonstrate the respective semantic memory knowledge base. Additionally, I have worked on performance tests to prepare for the use of very large knowledge bases.

My short term goals include completing a comprehensive performance profile of the semantic memory module as well as porting the OpenCyc ontology and WordNet lexical database. My long term goal is to find suitable methods for performing reasoning tasks with the semantic memory knowledge bases.

To identify unknown proteins, tandem mass spectrometry-based proteomics uses instruments to fragment proteins into shorter peptides, which are they fragmented into series of smaller molecules (which are subsequently ionized so they can be measured). From these ion series, the instrument produces spectra with masses (molecular mass over charge) and the associated intensities of the recorded signals. The problem is that these spectra are often noisy; furthermore, though a great deal of fragmentation research has produced useful heuristics and models of how spectra should be produced, software applications have not been particularly successful in reliably producing appropriate sequences.

My first attempt to produce de novo sequences used genetic algorithms. Though this approach generally worked for smaller peptides of fairly good quality, the results were less predictable for longer sequences, and the run times were not appropriate for high throughput data processing.

My second approach produced the set of all candidate sequences described as a Cartesian product of all potential sub-sequences. (This approach relies on the problem exhibiting optimal substructure, which is does if the ions can be correctly classified by their series.) Combined with some optimizations that were guided by fragmentation research (limited to tryptic digests) as well as some dynamic parameters to increase effeciency, I was able to sequence longer peptides in more reasonable times, measured in minutes instead of hours. However, for longer proteins, this approach produced increasingly larger sets of high scoring, tightly clustered candidates, making it unsuitable for fully automated batch processing.

One potential application would be to produce a plugin for a spectra visualization tool (such as Expedite) to identify candidate solutions, which experts could then review, potentially reducing the time spent identifying peptides. One benefit of using this tool is that the results are justified (i.e., labeled ion peaks and information about support for ion series classifications), and with a little more development, could produce meta information such as the quality of the spectra and the confidence of the tool. Alternative chemistries could also be supported, but would greatly increase the search space for each identification.

Though I worked independently, I received feedback from several research scientists in the field, and collaborated closely with a researcher who produced the Expedite visualization tool.

Java developer employed to work as a core developer on the open source Tranche project, which is a secure, P2P distributed file system. We are working to build new tools and integrate existing tools to build a scientific pipeline for analyzing and sharing Proteomics data.

As of January 2008, I was promoted to the position of Intermediate Application Developer.

Taught four courses — INP 150 (Web Coding I), INP 153 (Designing for the User Experience), INP 160 (Internet Technology), and INP 170 (Web Coding II).

Designing for the User Experience was primarily an information architecture course, though we briefly covered other topics like accessibility, usability engineering and SEO.

Internet Technology was a cursory overview to the technologies that make the Internet work.

Web Coding I & II explored web page design with XHTML 1.0 Strict/CSS with modularity and ease of maintenance in mind, along with a little XML/XSL, WML and SGML.

(734) 764-4481

Redesigned the undergraduate website, as well as updated subsites [e.g., Searchpath], as well as some information architecture and graphic design [for website, fliers and posters]. I also conducted user testing for the new website I created.

(734) 647-5836

Assisted patrons with general technology issues such as web design, basic javascript coding, graphic arts applications like Illustrator and Photoshop, and other issues related to web and electronic-document creation.

Tutored adult students who were learning basic reading and pronunciation skills.