Bryan's Academic Life

Currently, I am a graduate student at Carnegie Mellon University working towards a Ph.D. in Computer Science, under the guidance of Manuela Veloso.

You can find my CV here.

You can also find information about my thesis.

I am working with the SPIRAL group to see how machine learning can be applied to optimizing signal processing code. For details on this work, see the following papers:

• "Learning to Generate Fast Signal Processing Implementations" which was presented at the 2001 International Conference on Machine Learning (ICML-2001).
  This paper discusses our method for generating fast WHT formulas. To do this, we also learned to accurately predict the number of level 1 data cache misses that formulas incur. See also my presentation slides.

• "Stochastic Search for Signal Processing Algorithm Optimization" which wiil be presented at Super Computing 2001 (SC'01).
  This paper discusses STEER, an evolutionary algorithm for finding fast signal processing implementations.

• "Fast Automatic Generation of DSP Algorithms" which was presented at ICCS-2001 (International Conference on Computational Science).
  This paper discusses the formula generator and its relation to the search engine in the SPIRAL system.

• "Automatic Generation and Adaptation of High-Performance Implementations of Signal Processing Transforms".
  This paper discusses the entire SPIRAL system and gives some results.

• "Learning to Predict Performance from Formula Modeling and Training Data" which was presented at ICML-2000.
  This paper discusses our work in learning to predict performance of formulas, including discussion and evaluation of feature sets.

• "Automated Formula Generation and Performance Learning for the FFT" which is a technical report of older work similar to that presented in the ICML-2000 paper.
  This paper also includes information on our early formula generation work and on FFTs.

I have also investigated one particular method for taking advantage of solutions to previous problems to help solve new problems in reinforcement learning. For a brief overview of this method, see my AAAI-99 student abstract; for a longer version, see the technical report "Learning State Features From Policies to Bias Exploration in Reinforcement Learning"; and for some of my earlier work along these lines, see this paper.

I used to have a very out-of-date copy of my resume on-line, but it so out of date it isn't worth having it around.

In May of 1997, I graduated from Purdue University majoring in computer engineering with a minor in mathematics. While at Purdue, I had been very active in the Purdue Christian Campus House, in the IEEE (Institute of Electrical and Electronics Engineers) student branch at Purdue, as well as the Purdue Mathematics Club.

During the summer of 1996, I was out at the University of Iowa for their REU (Research Experience for Undergraduates) program in Computer Science. where I worked on Probabilistic Theory Revision with Alberto Segre.