Simultaneous Perturbation Newton Algorithms for Simulation Optimization

We present a new Hessian estimator based on the simultaneous perturbation procedure, that requires three system simulations regardless of the parameter dimension. We then present two Newton-based simulation optimization algorithms that incorporate this Hessian estimator. The two algorithms differ primarily in the manner in which the Hessian estimate is used. Both our algorithms do not compute the inverse Hessian explicitly, thereby saving on computational effort. While our first algorithm directly obtains the product of the inverse Hessian with the gradient of the objective, our second algorithm makes use of the Sherman-Morrison matrix inversion lemma to recursively estimate the inverse Hessian. We provide proofs of convergence for both our algorithms. Next, we consider an interesting application of our algorithms on a problem of road traffic control. Our algorithms are seen to exhibit better performance than two Newton algorithms from a recent prior work.

Effects of refractive index mismatch in optical CT imaging of polymer gel dosimeters

Purpose: Proposing an image reconstruction technique, algebraic reconstruction technique-refraction correction (ART-rc). The proposed method takes care of refractive index mismatches present in gel dosimeter scanner at the boundary, and also corrects for the interior ray refraction. Polymer gel dosimeters with high dose regions have higher refractive index and optical density compared to the background medium, these changes in refractive index at high dose results in interior ray bending. Methods: The inclusion of the effects of refraction is an important step in reconstruction of optical density in gel dosimeters. The proposed ray tracing algorithm models the interior multiple refraction at the inhomogeneities. Jacob's ray tracing algorithm has been modified to calculate the pathlengths of the ray that traverses through the higher dose regions. The algorithm computes the length of the ray in each pixel along its path and is used as the weight matrix. Algebraic reconstruction technique and pixel based reconstruction algorithms are used for solving the reconstruction problem. The proposed method is tested with numerical phantoms for various noise levels. The experimental dosimetric results are also presented. Results: The results show that the proposed scheme ART-rc is able to reconstruct optical density inside the dosimeter better than the results obtained using filtered backprojection and conventional algebraic reconstruction approaches. The quantitative improvement using ART-rc is evaluated using gamma-index. The refraction errors due to regions of different refractive indices are discussed. The effects of modeling of interior refraction in the dose region are presented. Conclusions: The errors propagated due to multiple refraction effects have been modeled and the improvements in reconstruction using proposed model is presented. The refractive index of the dosimeter has a mismatch with the surrounding medium (for dry air or water scanning). The algorithm reconstructs the dose profiles by estimating refractive indices of multiple inhomogeneities having different refractive indices and optical densities embedded in the dosimeter. This is achieved by tracking the path of the ray that traverses through the dosimeter. Extensive simulation studies have been carried out and results are found to be matching that of experimental results. (C) 2015 American Association of Physicists in Medicine.

Super Resolution Using a Single Image Dictionary

To perform super resolution of low resolution images, state-of-the-art methods are based on learning a pair of lowresolution and high-resolution dictionaries from multiple images. These trained dictionaries are used to replace patches in lowresolution image with appropriate matching patches from the high-resolution dictionary. In this paper we propose using a single common image as dictionary, in conjunction with approximate nearest neighbour fields (ANNF) to perform super resolution (SR). By using a common source image, we are able to bypass the learning phase and also able to reduce the dictionary from a collection of hundreds of images to a single image. By adapting recent developments in ANNF computation, to suit super-resolution, we are able to perform much faster and accurate SR than existing techniques. To establish this claim, we compare the proposed algorithm against various state-of-the-art algorithms, and show that we are able to achieve b etter and faster reconstruction without any training.

Rainbow matchings in strongly edge-colored graphs

A rainbow matching of an edge-colored graph G is a matching in which no two edges have the same color. There have been several studies regarding the maximum size of a rainbow matching in a properly edge-colored graph G in terms of its minimum degree 3(G). Wang (2011) asked whether there exists a function f such that a properly edge-colored graph G with at least f (delta(G)) vertices is guaranteed to contain a rainbow matching of size delta(G). This was answered in the affirmative later: the best currently known function Lo and Tan (2014) is f(k) = 4k - 4, for k >= 4 and f (k) = 4k - 3, for k <= 3. Afterwards, the research was focused on finding lower bounds for the size of maximum rainbow matchings in properly edge-colored graphs with fewer than 4 delta(G) - 4 vertices. Strong edge-coloring of a graph G is a restriction of proper edge-coloring where every color class is required to be an induced matching, instead of just being a matching. In this paper, we give lower bounds for the size of a maximum rainbow matching in a strongly edge-colored graph Gin terms of delta(G). We show that for a strongly edge-colored graph G, if |V(G)| >= 2 |3 delta(G)/4|, then G has a rainbow matching of size |3 delta(G)/4|, and if |V(G)| < 2 |3 delta(G)/4|, then G has a rainbow matching of size |V(G)|/2] In addition, we prove that if G is a strongly edge-colored graph that is triangle-free, then it contains a rainbow matching of size at least delta(G). (C) 2015 Elsevier B.V. All rights reserved.

PERCOLATION ON THE INFORMATION-THEORETICALLY SECURE SIGNAL TO INTERFERENCE RATIO GRAPH

We consider a continuum percolation model consisting of two types of nodes, namely legitimate and eavesdropper nodes, distributed according to independent Poisson point processes in R-2 of intensities lambda and lambda(E), respectively. A directed edge from one legitimate node A to another legitimate node B exists provided that the strength of the signal transmitted from node A that is received at node B is higher than that received at any eavesdropper node. The strength of the signal received at a node from a legitimate node depends not only on the distance between these nodes, but also on the location of the other legitimate nodes and an interference suppression parameter gamma. The graph is said to percolate when there exists an infinitely connected component. We show that for any finite intensity lambda(E) of eavesdropper nodes, there exists a critical intensity lambda(c) < infinity such that for all lambda > lambda(c) the graph percolates for sufficiently small values of the interference parameter. Furthermore, for the subcritical regime, we show that there exists a lambda(0) such that for all lambda < lambda(0) <= lambda(c) a suitable graph defined over eavesdropper node connections percolates that precludes percolation in the graphs formed by the legitimate nodes.

Timing Recovery Algorithms and Architectures for 2-D Magnetic Recording Systems

We investigate the problem of timing recovery for 2-D magnetic recording (TDMR) channels. We develop a timing error model for TDMR channel considering the phase and frequency offsets with noise. We propose a 2-D data-aided phase-locked loop (PLL) architecture for tracking variations in the position and movement of the read head in the down-track and cross-track directions and analyze the convergence of the algorithm under non-separable timing errors. We further develop a 2-D interpolation-based timing recovery scheme that works in conjunction with the 2-D PLL. We quantify the efficiency of our proposed algorithms by simulations over a 2-D magnetic recording channel with timing errors.

On the Parameterized Complexity of Finding Separators with Non-Hereditary Properties

We study the problem of finding small s-t separators that induce graphs having certain properties. It is known that finding a minimum clique s-t separator is polynomial-time solvable (Tarjan in Discrete Math. 55:221-232, 1985), while for example the problems of finding a minimum s-t separator that induces a connected graph or forms an independent set are fixed-parameter tractable when parameterized by the size of the separator (Marx et al. in ACM Trans. Algorithms 9(4): 30, 2013). Motivated by these results, we study properties that generalize cliques, independent sets, and connected graphs, and determine the complexity of finding separators satisfying these properties. We investigate these problems also on bounded-degree graphs. Our results are as follows: Finding a minimum c-connected s-t separator is FPT for c=2 and W1]-hard for any ca parts per thousand yen3. Finding a minimum s-t separator with diameter at most d is W1]-hard for any da parts per thousand yen2. Finding a minimum r-regular s-t separator is W1]-hard for any ra parts per thousand yen1. For any decidable graph property, finding a minimum s-t separator with this property is FPT parameterized jointly by the size of the separator and the maximum degree. Finding a connected s-t separator of minimum size does not have a polynomial kernel, even when restricted to graphs of maximum degree at most 3, unless .

Modified Greedy Pursuits for Improving Sparse Recovery

Compressive Sensing (CS) theory combines the signal sampling and compression for sparse signals resulting in reduction in sampling rate. In recent years, many recovery algorithms have been proposed to reconstruct the signal efficiently. Subspace Pursuit and Compressive Sampling Matching Pursuit are some of the popular greedy methods. Also, Fusion of Algorithms for Compressed Sensing is a recently proposed method where several CS reconstruction algorithms participate and the final estimate of the underlying sparse signal is determined by fusing the estimates obtained from the participating algorithms. All these methods involve solving a least squares problem which may be ill-conditioned, especially in the low dimension measurement regime. In this paper, we propose a step prior to least squares to ensure the well-conditioning of the least squares problem. Using Monte Carlo simulations, we show that in low dimension measurement scenario, this modification improves the reconstruction capability of the algorithm in clean as well as noisy measurement cases.

Synthetic image super resolution using FeatureMatch

In this paper, we propose a super resolution (SR) method for synthetic images using FeatureMatch. Existing state-of-the-art super resolution methods are learning based methods, where a pair of low-resolution and high-resolution dictionary pair are trained, and this trained pair is used to replace patches in low-resolution image with appropriate matching patches from the high-resolution dictionary. In this paper, we show that by using Approximate Nearest Neighbour Fields (ANNF), and a common source image, we can by-pass the learning phase, and use a single image for dictionary. Thus, reducing the dictionary from a collection obtained from hundreds of training images, to a single image. We show that by modifying the latest developments in ANNF computation, to suit super resolution, we can perform much faster and more accurate SR than existing techniques. To establish this claim we will compare our algorithm against various state-of-the-art algorithms, and show that we are able to achieve better and faster reconstruction without any training phase.

Parameterized Algorithms and Kernels for 3-Hitting Set with Parity Constraints

The 3-Hitting Set problem involves a family of subsets F of size at most three over an universe U. The goal is to find a subset of U of the smallest possible size that intersects every set in F. The version of the problem with parity constraints asks for a subset S of size at most k that, in addition to being a hitting set, also satisfies certain parity constraints on the sizes of the intersections of S with each set in the family F. In particular, an odd (even) set is a hitting set that hits every set at either one or three (two) elements, and a perfect code is a hitting set that intersects every set at exactly one element. These questions are of fundamental interest in many contexts for general set systems. Just as for Hitting Set, we find these questions to be interesting for the case of families consisting of sets of size at most three. In this work, we initiate an algorithmic study of these problems in this special case, focusing on a parameterized analysis. We show, for each problem, efficient fixed-parameter tractable algorithms using search trees that are tailor-made to the constraints in question, and also polynomial kernels using sunflower-like arguments in a manner that accounts for equivalence under the additional parity constraints.

Query Click and Text Similarity Graph for Query Suggestions

Query suggestion is an important feature of the search engine with the explosive and diverse growth of web contents. Different kind of suggestions like query, image, movies, music and book etc. are used every day. Various types of data sources are used for the suggestions. If we model the data into various kinds of graphs then we can build a general method for any suggestions. In this paper, we have proposed a general method for query suggestion by combining two graphs: (1) query click graph which captures the relationship between queries frequently clicked on common URLs and (2) query text similarity graph which finds the similarity between two queries using Jaccard similarity. The proposed method provides literally as well as semantically relevant queries for users' need. Simulation results show that the proposed algorithm outperforms heat diffusion method by providing more number of relevant queries. It can be used for recommendation tasks like query, image, and product suggestion.

Kinematics and dynamics of Jansen leg mechanism: A bond graph approach

The Jansen mechanism is a one degree-of-freedom, planar, 12-link, leg mechanism that can be used in mobile robotic applications and in gait analysis. This paper presents the kinematics and dynamics of the Jansen leg mechanism. The forward kinematics, accomplished using circle intersection method, determines the trajectories of various points on the mechanism in the chassis (stationary link) reference frame. From the foot point trajectory, the step length is shown to vary linearly while step height varies non-linearly with change in crank radius. A dynamic model for the Jansen leg mechanism is proposed using bond graph approach with modulated multiport transformers. For given ground reaction force pattern and crank angular speed, this model helps determine the motor torque profile as well as the link and joint stresses. The model can therefore be used to rate the actuator torque and in design of the hardware and controller for such a system. The kinematics of the mechanism can also be obtained from this dynamic model. The proposed model is thus a useful tool for analysis and design of systems based on the Jansen leg mechanism. (C) 2015 Elsevier B.V. All rights reserved.

A coupled discriminative dictionary and transformation learning approach with applications to cross domain matching

Cross domain and cross-modal matching has many applications in the field of computer vision and pattern recognition. A few examples are heterogeneous face recognition, cross view action recognition, etc. This is a very challenging task since the data in two domains can differ significantly. In this work, we propose a coupled dictionary and transformation learning approach that models the relationship between the data in both domains. The approach learns a pair of transformation matrices that map the data in the two domains in such a manner that they share common sparse representations with respect to their own dictionaries in the transformed space. The dictionaries for the two domains are learnt in a coupled manner with an additional discriminative term to ensure improved recognition performance. The dictionaries and the transformation matrices are jointly updated in an iterative manner. The applicability of the proposed approach is illustrated by evaluating its performance on different challenging tasks: face recognition across pose, illumination and resolution, heterogeneous face recognition and cross view action recognition. Extensive experiments on five datasets namely, CMU-PIE, Multi-PIE, ChokePoint, HFB and IXMAS datasets and comparisons with several state-of-the-art approaches show the effectiveness of the proposed approach. (C) 2015 Elsevier B.V. All rights reserved.

Fully 3D refraction correction dosimetry system

The irradiation of selective regions in a polymer gel dosimeter results in an increase in optical density and refractive index (RI) at those regions. An optical tomography-based dosimeter depends on rayline path through the dosimeter to estimate and reconstruct the dose distribution. The refraction of light passing through a dose region results in artefacts in the reconstructed images. These refraction errors are dependant on the scanning geometry and collection optics. We developed a fully 3D image reconstruction algorithm, algebraic reconstruction technique-refraction correction (ART-rc) that corrects for the refractive index mismatches present in a gel dosimeter scanner not only at the boundary, but also for any rayline refraction due to multiple dose regions inside the dosimeter. In this study, simulation and experimental studies have been carried out to reconstruct a 3D dose volume using 2D CCD measurements taken for various views. The study also focuses on the effectiveness of using different refractive-index matching media surrounding the gel dosimeter. Since the optical density is assumed to be low for a dosimeter, the filtered backprojection is routinely used for reconstruction. We carry out the reconstructions using conventional algebraic reconstruction (ART) and refractive index corrected ART (ART-rc) algorithms. The reconstructions based on FDK algorithm for cone-beam tomography has also been carried out for comparison. Line scanners and point detectors, are used to obtain reconstructions plane by plane. The rays passing through dose region with a RI mismatch does not reach the detector in the same plane depending on the angle of incidence and RI. In the fully 3D scanning setup using 2D array detectors, light rays that undergo refraction are still collected and hence can still be accounted for in the reconstruction algorithm. It is found that, for the central region of the dosimeter, the usable radius using ART-rc algorithm with water as RI matched medium is 71.8%, an increase of 6.4% compared to that achieved using conventional ART algorithm. Smaller diameter dosimeters are scanned with dry air scanning by using a wide-angle lens that collects refracted light. The images reconstructed using cone beam geometry is seen to deteriorate in some planes as those regions are not scanned. Refraction correction is important and needs to be taken in to consideration to achieve quantitatively accurate dose reconstructions. Refraction modeling is crucial in array based scanners as it is not possible to identify refracted rays in the sinogram space.

A coupled discriminative dictionary and transformation learning approach with applications to cross domain matching

Cross domain and cross-modal matching has many applications in the field of computer vision and pattern recognition. A few examples are heterogeneous face recognition, cross view action recognition, etc. This is a very challenging task since the data in two domains can differ significantly. In this work, we propose a coupled dictionary and transformation learning approach that models the relationship between the data in both domains. The approach learns a pair of transformation matrices that map the data in the two domains in such a manner that they share common sparse representations with respect to their own dictionaries in the transformed space. The dictionaries for the two domains are learnt in a coupled manner with an additional discriminative term to ensure improved recognition performance. The dictionaries and the transformation matrices are jointly updated in an iterative manner. The applicability of the proposed approach is illustrated by evaluating its performance on different challenging tasks: face recognition across pose, illumination and resolution, heterogeneous face recognition and cross view action recognition. Extensive experiments on five datasets namely, CMU-PIE, Multi-PIE, ChokePoint, HFB and IXMAS datasets and comparisons with several state-of-the-art approaches show the effectiveness of the proposed approach. (C) 2015 Elsevier B.V. All rights reserved.