TY - CONF
AB - We study multi-label prediction for structured output sets, a problem that occurs, for example, in object detection in images, secondary structure prediction in computational biology, and graph matching with symmetries. Conventional multilabel classification techniques are typically not applicable in this situation, because they require explicit enumeration of the label set, which is infeasible in case of structured outputs. Relying on techniques originally designed for single-label structured prediction, in particular structured support vector machines, results in reduced prediction accuracy, or leads to infeasible optimization problems. In this work we derive a maximum-margin training formulation for multi-label structured prediction that remains computationally tractable while achieving high prediction accuracy. It also shares most beneficial properties with single-label maximum-margin approaches, in particular formulation as a convex optimization problem, efficient working set training, and PAC-Bayesian generalization bounds.
AU - Lampert, Christoph
ID - 3163
TI - Maximum margin multi-label structured prediction
ER -
TY - CONF
AB - Verification of programs with procedures, multi-threaded programs, and higher-order functional programs can be effectively au- tomated using abstraction and refinement schemes that rely on spurious counterexamples for abstraction discovery. The analysis of counterexam- ples can be automated by a series of interpolation queries, or, alterna- tively, as a constraint solving query expressed by a set of recursion free Horn clauses. (A set of interpolation queries can be formulated as a single constraint over Horn clauses with linear dependency structure between the unknown relations.) In this paper we present an algorithm for solving recursion free Horn clauses over a combined theory of linear real/rational arithmetic and uninterpreted functions. Our algorithm performs resolu- tion to deal with the clausal structure and relies on partial solutions to deal with (non-local) instances of functionality axioms.
AU - Gupta, Ashutosh
AU - Popeea, Corneliu
AU - Rybalchenko, Andrey
ED - Yang, Hongseok
ID - 3264
TI - Solving recursion-free Horn clauses over LI+UIF
VL - 7078
ER -
TY - CONF
AB - We present a joint image segmentation and labeling model (JSL) which, given a bag of figure-ground segment hypotheses extracted at multiple image locations and scales, constructs a joint probability distribution over both the compatible image interpretations (tilings or image segmentations) composed from those segments, and over their labeling into categories. The process of drawing samples from the joint distribution can be interpreted as first sampling tilings, modeled as maximal cliques, from a graph connecting spatially non-overlapping segments in the bag [1], followed by sampling labels for those segments, conditioned on the choice of a particular tiling. We learn the segmentation and labeling parameters jointly, based on Maximum Likelihood with a novel Incremental Saddle Point estimation procedure. The partition function over tilings and labelings is increasingly more accurately approximated by including incorrect configurations that a not-yet-competent model rates probable during learning. We show that the proposed methodologymatches the current state of the art in the Stanford dataset [2], as well as in VOC2010, where 41.7% accuracy on the test set is achieved.
AU - Ion, Adrian
AU - Carreira, Joao
AU - Sminchisescu, Cristian
ID - 3266
T2 - NIPS Proceedings
TI - Probabilistic joint image segmentation and labeling
VL - 24
ER -
TY - JOUR
AB - We address the problem of localizing homology classes, namely, finding the cycle representing a given class with the most concise geometric measure. We study the problem with different measures: volume, diameter and radius. For volume, that is, the 1-norm of a cycle, two main results are presented. First, we prove that the problem is NP-hard to approximate within any constant factor. Second, we prove that for homology of dimension two or higher, the problem is NP-hard to approximate even when the Betti number is O(1). The latter result leads to the inapproximability of the problem of computing the nonbounding cycle with the smallest volume and computing cycles representing a homology basis with the minimal total volume. As for the other two measures defined by pairwise geodesic distance, diameter and radius, we show that the localization problem is NP-hard for diameter but is polynomial for radius. Our work is restricted to homology over the ℤ2 field.
AU - Chen, Chao
AU - Freedman, Daniel
ID - 3267
IS - 3
JF - Discrete & Computational Geometry
TI - Hardness results for homology localization
VL - 45
ER -
TY - JOUR
AB - The unintentional scattering of light between neighboring surfaces in complex projection environments increases the brightness and decreases the contrast, disrupting the appearance of the desired imagery. To achieve satisfactory projection results, the inverse problem of global illumination must be solved to cancel this secondary scattering. In this paper, we propose a global illumination cancellation method that minimizes the perceptual difference between the desired imagery and the actual total illumination in the resulting physical environment. Using Gauss-Newton and active set methods, we design a fast solver for the bound constrained nonlinear least squares problem raised by the perceptual error metrics. Our solver is further accelerated with a CUDA implementation and multi-resolution method to achieve 1–2 fps for problems with approximately 3000 variables. We demonstrate the global illumination cancellation algorithm with our multi-projector system. Results show that our method preserves the color fidelity of the desired imagery significantly better than previous methods.
AU - Sheng, Yu
AU - Cutler, Barbara
AU - Chen, Chao
AU - Nasman, Joshua
ID - 3269
IS - 4
JF - Computer Graphics Forum
TI - Perceptual global illumination cancellation in complex projection environments
VL - 30
ER -
TY - CONF
AB - The persistence diagram of a filtered simplicial com- plex is usually computed by reducing the boundary matrix of the complex. We introduce a simple op- timization technique: by processing the simplices of the complex in decreasing dimension, we can “kill” columns (i.e., set them to zero) without reducing them. This technique completely avoids reduction on roughly half of the columns. We demonstrate that this idea significantly improves the running time of the reduction algorithm in practice. We also give an output-sensitive complexity analysis for the new al- gorithm which yields to sub-cubic asymptotic bounds under certain assumptions.
AU - Chen, Chao
AU - Kerber, Michael
ID - 3270
TI - Persistent homology computation with a twist
ER -
TY - CHAP
AB - In this paper we present an efficient framework for computation of persis- tent homology of cubical data in arbitrary dimensions. An existing algorithm using simplicial complexes is adapted to the setting of cubical complexes. The proposed approach enables efficient application of persistent homology in domains where the data is naturally given in a cubical form. By avoiding triangulation of the data, we significantly reduce the size of the complex. We also present a data-structure de- signed to compactly store and quickly manipulate cubical complexes. By means of numerical experiments, we show high speed and memory efficiency of our ap- proach. We compare our framework to other available implementations, showing its superiority. Finally, we report performance on selected 3D and 4D data-sets.
AU - Wagner, Hubert
AU - Chen, Chao
AU - Vuçini, Erald
ED - Peikert, Ronald
ED - Hauser, Helwig
ED - Carr, Hamish
ED - Fuchs, Raphael
ID - 3271
T2 - Topological Methods in Data Analysis and Visualization II
TI - Efficient computation of persistent homology for cubical data
ER -
TY - THES
AU - Maître, Jean-Léon
ID - 3273
TI - Mechanics of adhesion and de‐adhesion in zebrafish germ layer progenitors
ER -
TY - JOUR
AB - Diffusing membrane constituents are constantly exposed to a variety of forces that influence their stochastic path. Single molecule experiments allow for resolving trajectories at extremely high spatial and temporal accuracy, thereby offering insights into en route interactions of the tracer. In this review we discuss approaches to derive information about the underlying processes, based on single molecule tracking experiments. In particular, we focus on a new versatile way to analyze single molecule diffusion in the absence of a full analytical treatment. The method is based on comprehensive comparison of an experimental data set against the hypothetical outcome of multiple experiments performed on the computer. Since Monte Carlo simulations can be easily and rapidly performed even on state-of-the-art PCs, our method provides a simple way for testing various - even complicated - diffusion models. We describe the new method in detail, and show the applicability on two specific examples: firstly, kinetic rate constants can be derived for the transient interaction of mobile membrane proteins; secondly, residence time and corral size can be extracted for confined diffusion.
AU - Ruprecht, Verena
AU - Axmann, Markus
AU - Wieser, Stefan
AU - Schuetz, Gerhard
ID - 3287
IS - 8
JF - Current Protein & Peptide Science
TI - What can we learn from single molecule trajectories?
VL - 12
ER -
TY - JOUR
AB - The zonula adherens (ZA) of epithelial cells is a site of cell-cell adhesion where cellular forces are exerted and resisted. Increasing evidence indicates that E-cadherin adhesion molecules at the ZA serve to sense force applied on the junctions and coordinate cytoskeletal responses to those forces. Efforts to understand the role that cadherins play in mechanotransduction have been limited by the lack of assays to measure the impact of forces on the ZA. In this study we used 4D imaging of GFP-tagged E-cadherin to analyse the movement of the ZA. Junctions in confluent epithelial monolayers displayed prominent movements oriented orthogonal (perpendicular) to the ZA itself. Two components were identified in these movements: a relatively slow unidirectional (translational) component that could be readily fitted by least-squares regression analysis, upon which were superimposed more rapid oscillatory movements. Myosin IIB was a dominant factor responsible for driving the unilateral translational movements. In contrast, frequency spectrum analysis revealed that depletion of Myosin IIA increased the power of the oscillatory movements. This implies that Myosin IIA may serve to dampen oscillatory movements of the ZA. This extends our recent analysis of Myosin II at the ZA to demonstrate that Myosin IIA and Myosin IIB make distinct contributions to junctional movement at the ZA.
AU - Smutny, Michael
AU - Wu, Selwin
AU - Gomez, Guillermo
AU - Mangold, Sabine
AU - Yap, Alpha
AU - Hamilton, Nicholas
ID - 3288
IS - 7
JF - PLoS One
TI - Multicomponent analysis of junctional movements regulated by Myosin II isoforms at the epithelial zonula adherens
VL - 6
ER -
TY - JOUR
AB - Analysis of genomic data requires an efficient way to calculate likelihoods across very large numbers of loci. We describe a general method for finding the distribution of genealogies: we allow migration between demes, splitting of demes [as in the isolation-with-migration (IM) model], and recombination between linked loci. These processes are described by a set of linear recursions for the generating function of branch lengths. Under the infinite-sites model, the probability of any configuration of mutations can be found by differentiating this generating function. Such calculations are feasible for small numbers of sampled genomes: as an example, we show how the generating function can be derived explicitly for three genes under the two-deme IM model. This derivation is done automatically, using Mathematica. Given data from a large number of unlinked and nonrecombining blocks of sequence, these results can be used to find maximum-likelihood estimates of model parameters by tabulating the probabilities of all relevant mutational configurations and then multiplying across loci. The feasibility of the method is demonstrated by applying it to simulated data and to a data set previously analyzed by Wang and Hey (2010) consisting of 26,141 loci sampled from Drosophila simulans and D. melanogaster. Our results suggest that such likelihood calculations are scalable to genomic data as long as the numbers of sampled individuals and mutations per sequence block are small.
AU - Lohse, Konrad
AU - Harrison, Richard
AU - Barton, Nicholas H
ID - 3290
IS - 3
JF - Genetics
TI - A general method for calculating likelihoods under the coalescent process
VL - 189
ER -
TY - CONF
AB - Animating detailed liquid surfaces has always been a challenge for computer graphics researchers and visual effects artists. Over the past few years, researchers in this field have focused on mesh-based surface tracking to synthesize extremely detailed liquid surfaces as efficiently as possible. This course provides a solid understanding of the steps required to create a fluid simulator with a mesh-based liquid surface.
The course begins with an overview of several existing liquid-surface-tracking techniques and the pros and cons of each method. Then it explains how to embed a triangle mesh into a finite-difference-based fluid simulator and describes several methods for allowing the liquid surface to merge together or break apart. The final section showcases the benefits and further applications of a mesh-based liquid surface, highlighting state-of-the-art methods for tracking colors and textures, maintaining liquid volume, preserving small surface features, and simulating realistic surface-tension waves.
AU - Wojtan, Christopher J
AU - Müller Fischer, Matthias
AU - Brochu, Tyson
ID - 3297
TI - Liquid simulation with mesh-based surface tracking
ER -
TY - CONF
AB - We present a new algorithm for enforcing incompressibility for Smoothed Particle Hydrodynamics (SPH) by preserving uniform density across the domain. We propose a hybrid method that uses a Poisson solve on a coarse grid to enforce a divergence free velocity ﬁeld, followed by a local density correction of the particles. This avoids typical grid artifacts and maintains the Lagrangian nature of SPH by directly transferring pressures onto particles. Our method can be easily integrated with existing SPH techniques such as the incompressible PCISPH method as well as weakly compressible SPH by adding an additional force term. We show that this hybrid method accelerates convergence towards uniform density and permits a signiﬁcantly larger time step compared to earlier approaches while producing similar results. We demonstrate our approach in a variety of scenarios with signiﬁcant pressure gradients such as splashing liquids.
AU - Raveendran, Karthik
AU - Wojtan, Christopher J
AU - Turk, Greg
ED - Spencer, Stephen
ID - 3298
TI - Hybrid smoothed particle hydrodynamics
ER -
TY - CONF
AB - We introduce propagation models, a formalism designed to support general and efficient data structures for the transient analysis of biochemical reaction networks. We give two use cases for propagation abstract data types: the uniformization method and numerical integration. We also sketch an implementation of a propagation abstract data type, which uses abstraction to approximate states.
AU - Henzinger, Thomas A
AU - Mateescu, Maria
ID - 3299
TI - Propagation models for computing biochemical reaction networks
ER -
TY - CONF
AB - The chemical master equation is a differential equation describing the time evolution of the probability distribution over the possible “states” of a biochemical system. The solution of this equation is of interest within the systems biology field ever since the importance of the molec- ular noise has been acknowledged. Unfortunately, most of the systems do not have analytical solutions, and numerical solutions suffer from the course of dimensionality and therefore need to be approximated. Here, we introduce the concept of tail approximation, which retrieves an approximation of the probabilities in the tail of a distribution from the total probability of the tail and its conditional expectation. This approximation method can then be used to numerically compute the solution of the chemical master equation on a subset of the state space, thus fighting the explosion of the state space, for which this problem is renowned.
AU - Henzinger, Thomas A
AU - Mateescu, Maria
ID - 3301
TI - Tail approximation for the chemical master equation
ER -
TY - CONF
AB - Cloud computing aims to give users virtually unlimited pay-per-use computing resources without the burden of managing the underlying infrastructure. We present a new job execution environment Flextic that exploits scal- able static scheduling techniques to provide the user with a flexible pricing model, such as a tradeoff between dif- ferent degrees of execution speed and execution price, and at the same time, reduce scheduling overhead for the cloud provider. We have evaluated a prototype of Flextic on Amazon EC2 and compared it against Hadoop. For various data parallel jobs from machine learning, im- age processing, and gene sequencing that we considered, Flextic has low scheduling overhead and reduces job du- ration by up to 15% compared to Hadoop, a dynamic cloud scheduler.
AU - Henzinger, Thomas A
AU - Singh, Anmol
AU - Singh, Vasu
AU - Wies, Thomas
AU - Zufferey, Damien
ID - 3302
TI - Static scheduling in clouds
ER -
TY - CHAP
AB - Alpha shapes have been conceived in 1981 as an attempt to define the shape of a finite set of point in the plane. Since then, connections to diverse areas in the sciences and engineering have developed, including to pattern recognition, digital shape sampling and processing, and structural molecular biology. This survey begins with a historical account and discusses geometric, algorithmic, topological, and combinatorial aspects of alpha shapes in this sequence.
AU - Herbert Edelsbrunner
ID - 3311
T2 - Tessellations in the Sciences
TI - Alpha shapes - a survey
ER -
TY - GEN
AB - We study the 3D reconstruction of plant roots from multiple 2D images. To meet the challenge caused by the delicate nature of thin branches, we make three innovations to cope with the sensitivity to image quality and calibration. First, we model the background as a harmonic function to improve the segmentation of the root in each 2D image. Second, we develop the concept of the regularized visual hull which reduces the effect of jittering and refraction by ensuring consistency with one 2D image. Third, we guarantee connectedness through adjustments to the 3D reconstruction that minimize global error. Our software is part of a biological phenotype/genotype study of agricultural root systems. It has been tested on more than 40 plant roots and results are promising in terms of reconstruction quality and efficiency.
AU - Zheng, Ying
AU - Gu, Steve
AU - Edelsbrunner, Herbert
AU - Tomasi, Carlo
AU - Benfey, Philip
ID - 3312
T2 - Proceedings of the IEEE International Conference on Computer Vision
TI - Detailed reconstruction of 3D plant root shape
ER -
TY - CONF
AB - Interpreting an image as a function on a compact sub- set of the Euclidean plane, we get its scale-space by diffu- sion, spreading the image over the entire plane. This gener- ates a 1-parameter family of functions alternatively defined as convolutions with a progressively wider Gaussian ker- nel. We prove that the corresponding 1-parameter family of persistence diagrams have norms that go rapidly to zero as time goes to infinity. This result rationalizes experimental observations about scale-space. We hope this will lead to targeted improvements of related computer vision methods.
AU - Chen, Chao
AU - Edelsbrunner, Herbert
ID - 3313
T2 - Proceedings of the IEEE International Conference on Computer Vision
TI - Diffusion runs low on persistence fast
ER -
TY - JOUR
AB - We consider two-player games played in real time on game structures with clocks where the objectives of players are described using parity conditions. The games are concurrent in that at each turn, both players independently propose a time delay and an action, and the action with the shorter delay is chosen. To prevent a player from winning by blocking time, we restrict each player to play strategies that ensure that the player cannot be responsible for causing a zeno run. First, we present an efficient reduction of these games to turn-based (i.e., not concurrent) finite-state (i.e., untimed) parity games. Our reduction improves the best known complexity for solving timed parity games. Moreover, the rich class of algorithms for classical parity games can now be applied to timed parity games. The states of the resulting game are based on clock regions of the original game, and the state space of the finite game is linear in the size of the region graph. Second, we consider two restricted classes of strategies for the player that represents the controller in a real-time synthesis problem, namely, limit-robust and bounded-robust winning strategies. Using a limit-robust winning strategy, the controller cannot choose an exact real-valued time delay but must allow for some nonzero jitter in each of its actions. If there is a given lower bound on the jitter, then the strategy is bounded-robust winning. We show that exact strategies are more powerful than limit-robust strategies, which are more powerful than bounded-robust winning strategies for any bound. For both kinds of robust strategies, we present efficient reductions to standard timed automaton games. These reductions provide algorithms for the synthesis of robust real-time controllers.
AU - Chatterjee, Krishnendu
AU - Henzinger, Thomas A
AU - Prabhu, Vinayak
ID - 3315
IS - 4
JF - Logical Methods in Computer Science
TI - Timed parity games: Complexity and robustness
VL - 7
ER -