### FR +E N

, 2013

"... Randomized matching heuristics with quality guarantees on shared memory parallel computers Fanny Dufossé, Kamer Kaya, Bora Uçar ha l-0 ..."

Abstract
- Add to MetaCart

Randomized matching heuristics with quality guarantees on shared memory parallel computers Fanny Dufossé, Kamer Kaya, Bora Uçar ha l-0

### Open Access

"... Robust flash denoising/deblurring by iterative guided filtering ..."

(Show Context)
### Multiple Kernel Learning Using Nearest Neighbor Classifiers

"... We study the problem of multiple kernel learning (MKL) in a classifica-tion setting. We first examine the kernel alignment metric and show that maximizing the alignment of a kernel with the target kernel Y Y T cor-responds to a constrained minimization of the margin loss of a weighted Nearest-Neighb ..."

Abstract
- Add to MetaCart

(Show Context)
We study the problem of multiple kernel learning (MKL) in a classifica-tion setting. We first examine the kernel alignment metric and show that maximizing the alignment of a kernel with the target kernel Y Y T cor-responds to a constrained minimization of the margin loss of a weighted Nearest-Neighbor (NN) classifier. Current MKL methods (both single and two-stage) use the Support Vector Machine classifier and the hinge loss. We expand the framework to include the NN classifier and the margin loss, in addition to the hinge loss for classification. This results in multiple com-binations of classifier and loss functions for multiple kernel learning. We make a thorough empirical study of the combinations. The NN classifier is particularly suitable to perform MKL on large datasets, with training a speedup of O (n2) over MKL algorithms that use SVMs. 1

### AND THE COMMITTEE ON GRADUATE STUDIES

, 2010

"... Re-distributed by Stanford University under license with the author. This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License. ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate ..."

Abstract
- Add to MetaCart

(Show Context)
Re-distributed by Stanford University under license with the author. This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License. ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate

### FR +E N

, 2013

"... Bipartite matching heuristics with quality guarantees on shared memory parallel computers Fanny Dufossé, Kamer Kaya, Bora Uçar ha l-0 ..."

Abstract
- Add to MetaCart

Bipartite matching heuristics with quality guarantees on shared memory parallel computers Fanny Dufossé, Kamer Kaya, Bora Uçar ha l-0

### 1 DNA Meets the SVD

"... This paper introduces an important area of computational cell biology where complex, publicly available genomic data is being examined by linear algebra methods, with the aim of revealing biological and medical insights. Section 1: What’s New? Since the time of Gregor Mendel, biologists have been at ..."

Abstract
- Add to MetaCart

This paper introduces an important area of computational cell biology where complex, publicly available genomic data is being examined by linear algebra methods, with the aim of revealing biological and medical insights. Section 1: What’s New? Since the time of Gregor Mendel, biologists have been attempting to understand how genes determine biological properties. Differences in genes largely explain biological diversity. But in spite of this all humans are recognisably the same due to our control systems that respond to driving forces such as feeding, stress, infection, age, sex and environment. These controls operate at all possible levels, many of which can now be studied using high-throughput technology. Microarrays observe the transfer of information from deoxyribonucleic acid (DNA), containing around 30,000 genes, to messenger ribonucleic acid (mRNA). In this way the state of all these genes can be recorded for individual samples. In terms of the functioning of the cell, genes are important because the mRNA that they create goes on to produce proteins, and proteins are the catalysts of all cells ’ activities. Maybe 20,000 mRNA signals are

### Nonlocal Image Editing

"... Abstract — In this paper, we introduce a new image editing tool based on the spectrum of a global filter computed from image affinities. Recently, it has been shown that the global filter derived from a fully connected graph representing the image can be approximated using the Nyström extension. Thi ..."

Abstract
- Add to MetaCart

(Show Context)
Abstract — In this paper, we introduce a new image editing tool based on the spectrum of a global filter computed from image affinities. Recently, it has been shown that the global filter derived from a fully connected graph representing the image can be approximated using the Nyström extension. This filter is computed by approximating the leading eigenvectors of the filter. These orthonormal eigenfunctions are highly expressive of the coarse and fine details in the underlying image, where each eigenvector can be interpreted as one scale of a data-dependent multiscale image decomposition. In this filtering scheme, each eigenvalue can boost or suppress the corresponding signal com-ponent in each scale. Our analysis shows that the mapping of the eigenvalues by an appropriate polynomial function endows the filter with a number of important capabilities, such as edge-aware sharpening, denoising, tone manipulation, and abstraction, to name a few. Furthermore, the edits can be easily propagated across the image. Index Terms — Image editing, non-local filters, Nyström extension.

### Ranking via Sinkhorn Propagation

"... Abstract: It is of increasing importance to develop learning meth-ods for ranking. In contrast to many learning objectives, however, the ranking problem presents difficulties due to the fact that the space of permutations is not smooth. In this paper, we examine the class of rank-linear objective fu ..."

Abstract
- Add to MetaCart

(Show Context)
Abstract: It is of increasing importance to develop learning meth-ods for ranking. In contrast to many learning objectives, however, the ranking problem presents difficulties due to the fact that the space of permutations is not smooth. In this paper, we examine the class of rank-linear objective functions, which includes popular metrics such as precision and discounted cumulative gain. In particular, we observe that expectations of these gains are completely characterized by the marginals of the corresponding distribution over permutation matrices. Thus, the expectations of rank-linear objectives can always be described through locations in the Birkhoff polytope, i.e., doubly-stochastic ma-trices (DSMs). We propose a technique for learning DSM-based ranking functions using an iterative projection operator known as Sinkhorn normalization. Gradients of this operator can be computed via back-propagation, resulting in an algorithm we call Sinkhorn propagation, or SinkProp. This approach can be combined with a wide range of gradient-based approaches to rank learning. We demonstrate the utility of SinkProp on several information retrieval data sets. 1.