@phdthesis{197, abstract = {Modern computer vision systems heavily rely on statistical machine learning models, which typically require large amounts of labeled data to be learned reliably. Moreover, very recently computer vision research widely adopted techniques for representation learning, which further increase the demand for labeled data. However, for many important practical problems there is relatively small amount of labeled data available, so it is problematic to leverage full potential of the representation learning methods. One way to overcome this obstacle is to invest substantial resources into producing large labelled datasets. Unfortunately, this can be prohibitively expensive in practice. In this thesis we focus on the alternative way of tackling the aforementioned issue. We concentrate on methods, which make use of weakly-labeled or even unlabeled data. Specifically, the first half of the thesis is dedicated to the semantic image segmentation task. We develop a technique, which achieves competitive segmentation performance and only requires annotations in a form of global image-level labels instead of dense segmentation masks. Subsequently, we present a new methodology, which further improves segmentation performance by leveraging tiny additional feedback from a human annotator. By using our methods practitioners can greatly reduce the amount of data annotation effort, which is required to learn modern image segmentation models. In the second half of the thesis we focus on methods for learning from unlabeled visual data. We study a family of autoregressive models for modeling structure of natural images and discuss potential applications of these models. Moreover, we conduct in-depth study of one of these applications, where we develop the state-of-the-art model for the probabilistic image colorization task.}, author = {Kolesnikov, Alexander}, issn = {2663-337X}, pages = {113}, publisher = {Institute of Science and Technology Austria}, title = {{Weakly-Supervised Segmentation and Unsupervised Modeling of Natural Images}}, doi = {10.15479/AT:ISTA:th_1021}, year = {2018}, } @article{6774, abstract = {A central problem of algebraic topology is to understand the homotopy groups 𝜋𝑑(𝑋) of a topological space X. For the computational version of the problem, it is well known that there is no algorithm to decide whether the fundamental group 𝜋1(𝑋) of a given finite simplicial complex X is trivial. On the other hand, there are several algorithms that, given a finite simplicial complex X that is simply connected (i.e., with 𝜋1(𝑋) trivial), compute the higher homotopy group 𝜋𝑑(𝑋) for any given 𝑑≥2 . However, these algorithms come with a caveat: They compute the isomorphism type of 𝜋𝑑(𝑋) , 𝑑≥2 as an abstract finitely generated abelian group given by generators and relations, but they work with very implicit representations of the elements of 𝜋𝑑(𝑋) . Converting elements of this abstract group into explicit geometric maps from the d-dimensional sphere 𝑆𝑑 to X has been one of the main unsolved problems in the emerging field of computational homotopy theory. Here we present an algorithm that, given a simply connected space X, computes 𝜋𝑑(𝑋) and represents its elements as simplicial maps from a suitable triangulation of the d-sphere 𝑆𝑑 to X. For fixed d, the algorithm runs in time exponential in size(𝑋) , the number of simplices of X. Moreover, we prove that this is optimal: For every fixed 𝑑≥2 , we construct a family of simply connected spaces X such that for any simplicial map representing a generator of 𝜋𝑑(𝑋) , the size of the triangulation of 𝑆𝑑 on which the map is defined, is exponential in size(𝑋) .}, author = {Filakovský, Marek and Franek, Peter and Wagner, Uli and Zhechev, Stephan Y}, issn = {2367-1734}, journal = {Journal of Applied and Computational Topology}, number = {3-4}, pages = {177--231}, publisher = {Springer}, title = {{Computing simplicial representatives of homotopy group elements}}, doi = {10.1007/s41468-018-0021-5}, volume = {2}, year = {2018}, } @inproceedings{133, abstract = {Synchronous programs are easy to specify because the side effects of an operation are finished by the time the invocation of the operation returns to the caller. Asynchronous programs, on the other hand, are difficult to specify because there are side effects due to pending computation scheduled as a result of the invocation of an operation. They are also difficult to verify because of the large number of possible interleavings of concurrent computation threads. We present synchronization, a new proof rule that simplifies the verification of asynchronous programs by introducing the fiction, for proof purposes, that asynchronous operations complete synchronously. Synchronization summarizes an asynchronous computation as immediate atomic effect. Modular verification is enabled via pending asynchronous calls in atomic summaries, and a complementary proof rule that eliminates pending asynchronous calls when components and their specifications are composed. We evaluate synchronization in the context of a multi-layer refinement verification methodology on a collection of benchmark programs.}, author = {Kragl, Bernhard and Qadeer, Shaz and Henzinger, Thomas A}, issn = {18688969}, location = {Beijing, China}, publisher = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik}, title = {{Synchronizing the asynchronous}}, doi = {10.4230/LIPIcs.CONCUR.2018.21}, volume = {118}, year = {2018}, } @inproceedings{187, abstract = {Given a locally finite X ⊆ ℝd and a radius r ≥ 0, the k-fold cover of X and r consists of all points in ℝd that have k or more points of X within distance r. We consider two filtrations - one in scale obtained by fixing k and increasing r, and the other in depth obtained by fixing r and decreasing k - and we compute the persistence diagrams of both. While standard methods suffice for the filtration in scale, we need novel geometric and topological concepts for the filtration in depth. In particular, we introduce a rhomboid tiling in ℝd+1 whose horizontal integer slices are the order-k Delaunay mosaics of X, and construct a zigzag module from Delaunay mosaics that is isomorphic to the persistence module of the multi-covers. }, author = {Edelsbrunner, Herbert and Osang, Georg F}, location = {Budapest, Hungary}, publisher = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik}, title = {{The multi-cover persistence of Euclidean balls}}, doi = {10.4230/LIPIcs.SoCG.2018.34}, volume = {99}, year = {2018}, } @article{692, abstract = {We consider families of confocal conics and two pencils of Apollonian circles having the same foci. We will show that these families of curves generate trivial 3-webs and find the exact formulas describing them.}, author = {Akopyan, Arseniy}, journal = {Geometriae Dedicata}, number = {1}, pages = {55 -- 64}, publisher = {Springer}, title = {{3-Webs generated by confocal conics and circles}}, doi = {10.1007/s10711-017-0265-6}, volume = {194}, year = {2018}, } @article{77, abstract = {Holes confined in quantum dots have gained considerable interest in the past few years due to their potential as spin qubits. Here we demonstrate two-axis control of a spin 3/2 qubit in natural Ge. The qubit is formed in a hut wire double quantum dot device. The Pauli spin blockade principle allowed us to demonstrate electric dipole spin resonance by applying a radio frequency electric field to one of the electrodes defining the double quantum dot. Coherent hole spin oscillations with Rabi frequencies reaching 140 MHz are demonstrated and dephasing times of 130 ns are measured. The reported results emphasize the potential of Ge as a platform for fast and electrically tunable hole spin qubit devices.}, author = {Watzinger, Hannes and Kukucka, Josip and Vukusic, Lada and Gao, Fei and Wang, Ting and Schäffler, Friedrich and Zhang, Jian and Katsaros, Georgios}, journal = {Nature Communications}, number = {3902 }, publisher = {Nature Publishing Group}, title = {{A germanium hole spin qubit}}, doi = {10.1038/s41467-018-06418-4}, volume = {9}, year = {2018}, } @article{401, abstract = {The actomyosin cytoskeleton, a key stress-producing unit in epithelial cells, oscillates spontaneously in a wide variety of systems. Although much of the signal cascade regulating myosin activity has been characterized, the origin of such oscillatory behavior is still unclear. Here, we show that basal myosin II oscillation in Drosophila ovarian epithelium is not controlled by actomyosin cortical tension, but instead relies on a biochemical oscillator involving ROCK and myosin phosphatase. Key to this oscillation is a diffusive ROCK flow, linking junctional Rho1 to medial actomyosin cortex, and dynamically maintained by a self-activation loop reliant on ROCK kinase activity. In response to the resulting myosin II recruitment, myosin phosphatase is locally enriched and shuts off ROCK and myosin II signals. Coupling Drosophila genetics, live imaging, modeling, and optogenetics, we uncover an intrinsic biochemical oscillator at the core of myosin II regulatory network, shedding light on the spatio-temporal dynamics of force generation.}, author = {Qin, Xiang and Hannezo, Edouard B and Mangeat, Thomas and Liu, Chang and Majumder, Pralay and Liu, Jjiaying and Choesmel Cadamuro, Valerie and Mcdonald, Jocelyn and Liu, Yinyao and Yi, Bin and Wang, Xiaobo}, journal = {Nature Communications}, number = {1}, publisher = {Nature Publishing Group}, title = {{A biochemical network controlling basal myosin oscillation}}, doi = {10.1038/s41467-018-03574-5}, volume = {9}, year = {2018}, } @article{318, abstract = {The insect’s fat body combines metabolic and immunological functions. In this issue of Developmental Cell, Franz et al. (2018) show that in Drosophila, cells of the fat body are not static, but can actively “swim” toward sites of epithelial injury, where they physically clog the wound and locally secrete antimicrobial peptides.}, author = {Casano, Alessandra M and Sixt, Michael K}, journal = {Developmental Cell}, number = {4}, pages = {405 -- 406}, publisher = {Cell Press}, title = {{A fat lot of good for wound healing}}, doi = {10.1016/j.devcel.2018.02.009}, volume = {44}, year = {2018}, } @article{410, abstract = {Lesion verification and quantification is traditionally done via histological examination of sectioned brains, a time-consuming process that relies heavily on manual estimation. Such methods are particularly problematic in posterior cortical regions (e.g. visual cortex), where sectioning leads to significant damage and distortion of tissue. Even more challenging is the post hoc localization of micro-electrodes, which relies on the same techniques, suffers from similar drawbacks and requires even higher precision. Here, we propose a new, simple method for quantitative lesion characterization and electrode localization that is less labor-intensive and yields more detailed results than conventional methods. We leverage staining techniques standard in electron microscopy with the use of commodity micro-CT imaging. We stain whole rat and zebra finch brains in osmium tetroxide, embed these in resin and scan entire brains in a micro-CT machine. The scans result in 3D reconstructions of the brains with section thickness dependent on sample size (12–15 and 5–6 microns for rat and zebra finch respectively) that can be segmented manually or automatically. Because the method captures the entire intact brain volume, comparisons within and across studies are more tractable, and the extent of lesions and electrodes may be studied with higher accuracy than with current methods.}, author = {Masís, Javier and Mankus, David and Wolff, Steffen and Guitchounts, Grigori and Jösch, Maximilian A and Cox, David}, journal = {Scientific Reports}, number = {1}, publisher = {Nature Publishing Group}, title = {{A micro-CT-based method for quantitative brain lesion characterization and electrode localization}}, doi = {10.1038/s41598-018-23247-z}, volume = {8}, year = {2018}, } @article{277, abstract = {Arabidopsis and human ARM protein interact with telomerase. Deregulated mRNA levels of DNA repair and ribosomal protein genes in an Arabidopsis arm mutant suggest non-telomeric ARM function. The human homolog ARMC6 interacts with hTRF2. Abstract: Telomerase maintains telomeres and has proposed non-telomeric functions. We previously identified interaction of the C-terminal domain of Arabidopsis telomerase reverse transcriptase (AtTERT) with an armadillo/β-catenin-like repeat (ARM) containing protein. Here we explore protein–protein interactions of the ARM protein, AtTERT domains, POT1a, TRF-like family and SMH family proteins, and the chromatin remodeling protein CHR19 using bimolecular fluorescence complementation (BiFC), yeast two-hybrid (Y2H) analysis, and co-immunoprecipitation. The ARM protein interacts with both the N- and C-terminal domains of AtTERT in different cellular compartments. ARM interacts with CHR19 and TRF-like I family proteins that also bind AtTERT directly or through interaction with POT1a. The putative human ARM homolog co-precipitates telomerase activity and interacts with hTRF2 protein in vitro. Analysis of Arabidopsis arm mutants shows no obvious changes in telomere length or telomerase activity, suggesting that ARM is not essential for telomere maintenance. The observed interactions with telomerase and Myb-like domain proteins (TRF-like family I) may therefore reflect possible non-telomeric functions. Transcript levels of several DNA repair and ribosomal genes are affected in arm mutants, and ARM, likely in association with other proteins, suppressed expression of XRCC3 and RPSAA promoter constructs in luciferase reporter assays. In conclusion, ARM can participate in non-telomeric functions of telomerase, and can also perform its own telomerase-independent functions.}, author = {Dokládal, Ladislav and Benková, Eva and Honys, David and Dupláková, Nikoleta and Lee, Lan and Gelvin, Stanton and Sýkorová, Eva}, journal = {Plant Molecular Biology}, number = {5}, pages = {407 -- 420}, publisher = {Springer}, title = {{An armadillo-domain protein participates in a telomerase interaction network}}, doi = {10.1007/s11103-018-0747-4}, volume = {97}, year = {2018}, }