Much of the research reported in this thesis was carried out or initiated while I was a visiting student in the “Foundations of Machine Learning” program at the Simons Institute for the Theory of Computing at UC Berkeley. I would like to explicitly thank Peter Bartlett.
Based on stochastic gradient Hamiltonian Monte Carlo, we develop a fully Bayesian approach to scalable GP and deep GP models, and demonstrate its competitive performance through an extensive.I am interested in the interplays between the numerical schemes of stochastic differential equations (SDEs) and data science. My research mainly focused on the construction of higher order schemes and on the analysis of the stochastic gradient Langevin dynamics (SGLD) algorithm in the non-convex learning problem.Novel ultrasound features for the identification of the vulnerable carotid plaque. Grigoris Makris, MD. A thesis submitted for the Degree of Doctor of Philosophy. Vascular Surgery Department. Imperial College of London. United Kingdom. October 2014.
I Novel adaptation of SGLD to infer covariance parameters in Gaussian processes 3 Accurate in characterizing the posterior distribution over covariance parameters 3 Scales with O(n) in space and with O(n2) in time 3 Massively parallelizable 3 Without assuming factorization of the likelihood (mini-batches).
PIGlets. The purpose of the session is for PhD students interested in machine learning to get to grips with advanced material sitting in difficulty above the taught courses here and below the discussions of cutting-edge material in the usual PIGS meetings. Anyone else who's interested in joining us for the same is welcome, but the point is to get a solid handle on the techniques rather than.
Bayesian Deep Learning for Integrated Intelligence: Bridging the Gap between Perception and Inference by Hao Wang This is to certify that I have examined the above PhD thesis and have found that it is complete and satisfactory in all respects, and that any and all revisions required by the PhD qualifying examination committee have been made.
On Thursday morning we talked to three of the PhD students about what life is like at STOR-i. They were very helpful and answered a lot of questions that we had! This week’s STOR-i forum involved five pi-minute theses as opposed to the general half-hour presentation of a single thesis. We heard short presentations on a range of topics, and.
PhD Thesis Chapter 3 contains unpublished results on combining selective inference with the debiased lasso, knockoff, and proposes a general algorithm for selective inference. Matrix Completion and Low-Rank SVD via Fast Alternating Least Squares. Trevor Hastie, Rahul Mazumder, Jason D. Lee, and Reza Zadeh. Journal of Machine Learning Research 2015. (Spark Implementation) (R Implementation) On.
Abstract: This talk is all about my PhD thesis for Bayesian supervised learning. I focus on the development of methodologies that help undertake learning of functional relationships between variables, given high-dimensional observations. The probabilistic learning of the functional relation between these variables is done by modelling this.
In this paper, we present a robust real-time face detection algorithm. We improved the conventional face detection algorithms for three different steps. For preprocessing step, we revise the modified.
CHARMM (Chemistry at HARvard Molecular Mechanics) is a highly versatile and widely used molecular simulation program. It has been developed over the last three decades with a primary focus on molecules of biological interest, including proteins, peptides, lipids, nucleic acids, carbohydrates and small molecule ligands, as they occur in solution, crystals, and membrane environments.
College of Health Sciences’ Research Dean, Professor Moses Chimbari recently hosted a successful Dissertation and Thesis Guidelines workshop for the College’s postgraduate students and supervisors. Attended by 60 delegates the one day workshop looked at CHS thesis and dissertation guidelines. The structure of the CHS thesis and dissertation.
The automated method to segment radiodense tissue presented in this thesis incorporates the effects of tissue compression from the mammography procedure. A survey of existing literature regarding the automated segmentation of radiodense tissue infers that no other researchers are including the effects created by tissue compression.
Rank aggregation based on pairwise comparisons over a set of items has a wide range of applications. Although considerable research has been devoted to the development of rank agg.
HEp-2 Cell Image Classi cation: A Comparative Analysis Praful Agrawal, Mayank Vatsa, Richa Singh Indraprastha Institute of Information Technology, Delhi, India Abstract. HEp-2 cell image classi cation is an important and relatively unexplored area of research. This paper presents an experimental analysis.
Masters Student Selected for Golden Key Leadership Summit. BY:. Click here for isiZulu version. In recognition of his academic achievements and community service, UKZN Master’s student and Golden Key member, Mr Bayabonga Zulu, was one of 200 students from around the world selected to attend the Golden Key Leadership Summit in Niagara Falls, Canada.
This is part of the speaker's PhD thesis and is an extension of joint work with A.Adem, O.A. Camarena and G.W. Semenoff.. The vast majority of current work on this problem (HMC, SGLD, variational) is based on mimicking the field of optimization, in particular gradient based methods, and as a consequence focusses on Riemann integrals. This severely limits the applicability of these methods.