Embedded Files
Disclaimer: this website is not updated much. Our list of publications on arXiv is a much better way to get an up-do-date glimpse at what we do.
Presentations
Presentations
David Kozak
Stochastic subspace approach to optimization in high dimensions, INFORMS 2019
Scaling up gradient free minimization, SIAM sectional meeting 2019
Farhad Pourkamali-Anaraki, Randomized Algorithms for Large-Scale Data Analysis
March 22, 2017, Ph.D. Thesis Defense Presentation
Farhad Pourkamali-Anaraki, A Randomized Approach to Efficient Kernel Clustering
slides, IEEE GlobalSIP 2016
Farhad Pourkamali-Anaraki, Randomized Dimension Reduction for Large-Scale Data Analysis
SIGAI Career Network and Conference, Northeastern University, October 20, 2016
Stephen Becker, Sampling large datasets via random mixing.
Colorado School of Mines, Applied Math and Statistics colloquium, Sept. 2 2016
Farhad Pourkamali-Anaraki, Preconditioned Data Sparsification for Big Data with Applications to PCA and K-Means
2016 SIAM Front Range Conference
Abstract: We analyze a compression scheme for large data sets that randomly keeps a small percentage of the components of each data sample. The benefit is that the output is a sparse matrix and therefore subsequent processing, such as PCA or K-means, is significantly faster, especially in a distributed-data setting. Furthermore, the sampling is single-pass and applicable to streaming data. The sampling mechanism is a variant of previous methods proposed in the literature combined with a randomized preconditioning to smooth the data. We provide guarantees for PCA in terms of the covariance matrix, and guarantees for K-means in terms of the error in the center estimators at a given step. We present numerical evidence to show both that our bounds are nearly tight and that our algorithms provide a real benefit when applied to standard test data sets, as well as providing certain benefits over related sampling approaches.
James Folberth, Application of Adjoint Operators in Gradient Computations
2016 SIAM Front Range Conference
Abstract: When using first-order optimization algorithms, it is often the case that the user must supply the gradient of the differentiable terms in the objective function. We consider two example problems that have a Euclidean error term involving a linear operation on the problem variables. The gradient of the Euclidean error term involves both the linear operator and its adjoint, which, in our examples, is not known in the literature. The first example is an image deblurring problem, where the linear operation is multi-stage wavelet synthesis. Our formulation of the adjoint holds for a variety of boundary conditions, which allows the formulation to generalize to a larger class of problems. The second example is a blind channel estimation problem taken from the convex optimization literature; the linear operation is convolution, but with a slight twist. In each example, we show how the adjoint operator can be applied efficiently.
Derek Driggs, Stable Principal Component Pursuit
2016 SIAM Front Range Conference
Other Projects
Other Projects
Perceptual Learning Website
Jeff Everett's DLA Project (2017/2018)
PerceptLearn.com website
Fast Hadamard Transform Code
Jeff Everett's DLA Project (2017/2018)
Implements fast Hadamard Transform in C using SIMD (AVX), with Matlab wrappers
Fastest known implementation (a bit faster than Spiral code)
Report abuse