TY - CONF AB - Game of Life is a simple and elegant model to study dynamical system over networks. The model consists of a graph where every vertex has one of two types, namely, dead or alive. A configuration is a mapping of the vertices to the types. An update rule describes how the type of a vertex is updated given the types of its neighbors. In every round, all vertices are updated synchronously, which leads to a configuration update. While in general, Game of Life allows a broad range of update rules, we focus on two simple families of update rules, namely, underpopulation and overpopulation, that model several interesting dynamics studied in the literature. In both settings, a dead vertex requires at least a desired number of live neighbors to become alive. For underpopulation (resp., overpopulation), a live vertex requires at least (resp. at most) a desired number of live neighbors to remain alive. We study the basic computation problems, e.g., configuration reachability, for these two families of rules. For underpopulation rules, we show that these problems can be solved in polynomial time, whereas for overpopulation rules they are PSPACE-complete. AU - Chatterjee, Krishnendu AU - Ibsen-Jensen, Rasmus AU - Jecker, Ismael R AU - Svoboda, Jakub ID - 8533 SN - 18688969 T2 - 45th International Symposium on Mathematical Foundations of Computer Science TI - Simplified game of life: Algorithms and complexity VL - 170 ER - TY - CONF AB - A regular language L of finite words is composite if there are regular languages L₁,L₂,…,L_t such that L = ⋂_{i = 1}^t L_i and the index (number of states in a minimal DFA) of every language L_i is strictly smaller than the index of L. Otherwise, L is prime. Primality of regular languages was introduced and studied in [O. Kupferman and J. Mosheiff, 2015], where the complexity of deciding the primality of the language of a given DFA was left open, with a doubly-exponential gap between the upper and lower bounds. We study primality for unary regular languages, namely regular languages with a singleton alphabet. A unary language corresponds to a subset of ℕ, making the study of unary prime languages closer to that of primality in number theory. We show that the setting of languages is richer. In particular, while every composite number is the product of two smaller numbers, the number t of languages necessary to decompose a composite unary language induces a strict hierarchy. In addition, a primality witness for a unary language L, namely a word that is not in L but is in all products of languages that contain L and have an index smaller than L’s, may be of exponential length. Still, we are able to characterize compositionality by structural properties of a DFA for L, leading to a LogSpace algorithm for primality checking of unary DFAs. AU - Jecker, Ismael R AU - Kupferman, Orna AU - Mazzocchi, Nicolas ID - 8534 SN - 18688969 T2 - 45th International Symposium on Mathematical Foundations of Computer Science TI - Unary prime languages VL - 170 ER - TY - JOUR AB - We prove some recent experimental observations of Dan Reznik concerning periodic billiard orbits in ellipses. For example, the sum of cosines of the angles of a periodic billiard polygon remains constant in the 1-parameter family of such polygons (that exist due to the Poncelet porism). In our proofs, we use geometric and complex analytic methods. AU - Akopyan, Arseniy AU - Schwartz, Richard AU - Tabachnikov, Serge ID - 8538 JF - European Journal of Mathematics SN - 2199-675X TI - Billiards in ellipses revisited ER - TY - GEN AB - The brain vasculature supplies neurons with glucose and oxygen, but little is known about how vascular plasticity contributes to brain function. Using longitudinal in vivo imaging, we reported that a substantial proportion of blood vessels in the adult brain sporadically occluded and regressed. Their regression proceeded through sequential stages of blood-flow occlusion, endothelial cell collapse, relocation or loss of pericytes, and retraction of glial endfeet. Regressing vessels were found to be widespread in mouse, monkey and human brains. Both brief occlusions of the middle cerebral artery and lipopolysaccharide-mediated inflammation induced an increase of vessel regression. Blockage of leukocyte adhesion to endothelial cells alleviated LPS-induced vessel regression. We further revealed that blood vessel regression caused a reduction of neuronal activity due to a dysfunction in mitochondrial metabolism and glutamate production. Our results elucidate the mechanism of vessel regression and its role in neuronal function in the adult brain. AU - Gao, Xiaofei AU - Li, Jun-Liszt AU - Chen, Xingjun AU - Ci, Bo AU - Chen, Fei AU - Lu, Nannan AU - Shen, Bo AU - Zheng, Lijun AU - Jia, Jie-Min AU - Yi, Yating AU - Zhang, Shiwen AU - Shi, Ying-Chao AU - Shi, Kaibin AU - Propson, Nicholas E AU - Huang, Yubin AU - Poinsatte, Katherine AU - Zhang, Zhaohuan AU - Yue, Yuanlei AU - Bosco, Dale B AU - Lu, Ying-mei AU - Yang, Shi-bing AU - Adams, Ralf H. AU - Lindner, Volkhard AU - Huang, Fen AU - Wu, Long-Jun AU - Zheng, Hui AU - Han, Feng AU - Hippenmeyer, Simon AU - Stowe, Ann M. AU - Peng, Bo AU - Margeta, Marta AU - Wang, Xiaoqun AU - Liu, Qiang AU - Körbelin, Jakob AU - Trepel, Martin AU - Lu, Hui AU - Zhou, Bo O. AU - Zhao, Hu AU - Su, Wenzhi AU - Bachoo, Robert M. AU - Ge, Woo-ping ID - 8616 T2 - bioRxiv TI - Reduction of neuronal activity mediated by blood-vessel regression in the brain ER - TY - GEN AB - A look at international activities on Open Science reveals a broad spectrum from individual institutional policies to national action plans. The present Recommendations for a National Open Science Strategy in Austria are based on these international initiatives and present practical considerations for their coordinated implementation with regard to strategic developments in research, technology and innovation (RTI) in Austria until 2030. They are addressed to all relevant actors in the RTI system, in particular to Research Performing Organisations, Research Funding Organisations, Research Policy, memory institutions such as Libraries and Researchers. The recommendation paper was developed from 2018 to 2020 by the OANA working group "Open Science Strategy" and published for the first time in spring 2020 for a public consultation. The now available final version of the recommendation document, which contains feedback and comments from the consultation, is intended to provide an impetus for further discussion and implementation of Open Science in Austria and serves as a contribution and basis for a potential national Open Science Strategy in Austria. The document builds on the diverse expertise of the authors (academia, administration, library and archive, information technology, science policy, funding system, etc.) and reflects their personal experiences and opinions. AU - Mayer, Katja AU - Rieck, Katharina AU - Reichmann, Stefan AU - Danowski, Patrick AU - Graschopf, Anton AU - König, Thomas AU - Kraker, Peter AU - Lehner, Patrick AU - Reckling, Falk AU - Ross-Hellauer, Tony AU - Spichtinger, Daniel AU - Tzatzanis, Michalis AU - Schürz, Stefanie ID - 8695 TI - Empfehlungen für eine nationale Open Science Strategie in Österreich / Recommendations for a National Open Science Strategy in Austria ER - TY - JOUR AB - As part of the Austrian Transition to Open Access (AT2OA) project, subproject TP1-B is working on designing a monitoring solution for the output of Open Access publications in Austria. This report on a potential Open Access monitoring approach in Austria is one of the results of these efforts and can serve as a basis for discussion on an international level. AU - Danowski, Patrick AU - Ferus, Andreas AU - Hikl, Anna-Laetitia AU - McNeill, Gerda AU - Miniberger, Clemens AU - Reding, Steve AU - Zarka, Tobias AU - Zojer, Michael ID - 8706 IS - 2 JF - Mitteilungen der Vereinigung Österreichischer Bibliothekarinnen und Bibliothekare TI - „Recommendation“ for the further procedure for open access monitoring. Deliverable of the AT2OA subproject TP1-B VL - 73 ER - TY - JOUR AB - Mosaic analysis with double markers (MADM) technology enables concomitant fluorescent cell labeling and induction of uniparental chromosome disomy (UPD) with single-cell resolution. In UPD, imprinted genes are either overexpressed 2-fold or are not expressed. Here, the MADM platform is utilized to probe imprinting phenotypes at the transcriptional level. This protocol highlights major steps for the generation and isolation of projection neurons and astrocytes with MADM-induced UPD from mouse cerebral cortex for downstream single-cell and low-input sample RNA-sequencing experiments. For complete details on the use and execution of this protocol, please refer to Laukoter et al. (2020b). AU - Laukoter, Susanne AU - Amberg, Nicole AU - Pauler, Florian AU - Hippenmeyer, Simon ID - 8978 IS - 3 JF - STAR Protocols SN - 2666-1667 TI - Generation and isolation of single cells from mouse brain with mosaic analysis with double markers-induced uniparental chromosome disomy VL - 1 ER - TY - CONF AB - We introduce LRT-NG, a set of techniques and an associated toolset that computes a reachtube (an over-approximation of the set of reachable states over a given time horizon) of a nonlinear dynamical system. LRT-NG significantly advances the state-of-the-art Langrangian Reachability and its associated tool LRT. From a theoretical perspective, LRT-NG is superior to LRT in three ways. First, it uses for the first time an analytically computed metric for the propagated ball which is proven to minimize the ball’s volume. We emphasize that the metric computation is the centerpiece of all bloating-based techniques. Secondly, it computes the next reachset as the intersection of two balls: one based on the Cartesian metric and the other on the new metric. While the two metrics were previously considered opposing approaches, their joint use considerably tightens the reachtubes. Thirdly, it avoids the "wrapping effect" associated with the validated integration of the center of the reachset, by optimally absorbing the interval approximation in the radius of the next ball. From a tool-development perspective, LRT-NG is superior to LRT in two ways. First, it is a standalone tool that no longer relies on CAPD. This required the implementation of the Lohner method and a Runge-Kutta time-propagation method. Secondly, it has an improved interface, allowing the input model and initial conditions to be provided as external input files. Our experiments on a comprehensive set of benchmarks, including two Neural ODEs, demonstrates LRT-NG’s superior performance compared to LRT, CAPD, and Flow*. AU - Gruenbacher, Sophie AU - Cyranka, Jacek AU - Lechner, Mathias AU - Islam, Md Ariful AU - Smolka, Scott A. AU - Grosu, Radu ID - 9103 SN - 07431546 T2 - Proceedings of the 59th IEEE Conference on Decision and Control TI - Lagrangian reachtubes: The next generation VL - 2020 ER - TY - CONF AB - Recent works have shown that gradient descent can find a global minimum for over-parameterized neural networks where the widths of all the hidden layers scale polynomially with N (N being the number of training samples). In this paper, we prove that, for deep networks, a single layer of width N following the input layer suffices to ensure a similar guarantee. In particular, all the remaining layers are allowed to have constant widths, and form a pyramidal topology. We show an application of our result to the widely used LeCun’s initialization and obtain an over-parameterization requirement for the single wide layer of order N2. AU - Nguyen, Quynh AU - Mondelli, Marco ID - 9221 T2 - 34th Conference on Neural Information Processing Systems TI - Global convergence of deep networks with one wide layer followed by pyramidal topology VL - 33 ER - TY - CONF AB - Optimizing convolutional neural networks for fast inference has recently become an extremely active area of research. One of the go-to solutions in this context is weight pruning, which aims to reduce computational and memory footprint by removing large subsets of the connections in a neural network. Surprisingly, much less attention has been given to exploiting sparsity in the activation maps, which tend to be naturally sparse in many settings thanks to the structure of rectified linear (ReLU) activation functions. In this paper, we present an in-depth analysis of methods for maximizing the sparsity of the activations in a trained neural network, and show that, when coupled with an efficient sparse-input convolution algorithm, we can leverage this sparsity for significant performance gains. To induce highly sparse activation maps without accuracy loss, we introduce a new regularization technique, coupled with a new threshold-based sparsification method based on a parameterized activation function called Forced-Activation-Threshold Rectified Linear Unit (FATReLU). We examine the impact of our methods on popular image classification models, showing that most architectures can adapt to significantly sparser activation maps without any accuracy loss. Our second contribution is showing that these these compression gains can be translated into inference speedups: we provide a new algorithm to enable fast convolution operations over networks with sparse activations, and show that it can enable significant speedups for end-to-end inference on a range of popular models on the large-scale ImageNet image classification task on modern Intel CPUs, with little or no retraining cost. AU - Kurtz, Mark AU - Kopinsky, Justin AU - Gelashvili, Rati AU - Matveev, Alexander AU - Carr, John AU - Goin, Michael AU - Leiserson, William AU - Moore, Sage AU - Nell, Bill AU - Shavit, Nir AU - Alistarh, Dan-Adrian ID - 9415 SN - 2640-3498 T2 - 37th International Conference on Machine Learning, ICML 2020 TI - Inducing and exploiting activation sparsity for fast neural network inference VL - 119 ER - TY - CONF AB - The family of feedback alignment (FA) algorithms aims to provide a more biologically motivated alternative to backpropagation (BP), by substituting the computations that are unrealistic to be implemented in physical brains. While FA algorithms have been shown to work well in practice, there is a lack of rigorous theory proofing their learning capabilities. Here we introduce the first feedback alignment algorithm with provable learning guarantees. In contrast to existing work, we do not require any assumption about the size or depth of the network except that it has a single output neuron, i.e., such as for binary classification tasks. We show that our FA algorithm can deliver its theoretical promises in practice, surpassing the learning performance of existing FA methods and matching backpropagation in binary classification tasks. Finally, we demonstrate the limits of our FA variant when the number of output neurons grows beyond a certain quantity. AU - Lechner, Mathias ID - 10672 T2 - 8th International Conference on Learning Representations TI - Learning representations for binary-classification without backpropagation ER - TY - CONF AB - A natural approach to generative modeling of videos is to represent them as a composition of moving objects. Recent works model a set of 2D sprites over a slowly-varying background, but without considering the underlying 3D scene that gives rise to them. We instead propose to model a video as the view seen while moving through a scene with multiple 3D objects and a 3D background. Our model is trained from monocular videos without any supervision, yet learns to generate coherent 3D scenes containing several moving objects. We conduct detailed experiments on two datasets, going beyond the visual complexity supported by state-of-the-art generative approaches. We evaluate our method on depth-prediction and 3D object detection---tasks which cannot be addressed by those earlier works---and show it out-performs them even on 2D instance segmentation and tracking. AU - Henderson, Paul M AU - Lampert, Christoph ID - 8188 SN - 9781713829546 T2 - 34th Conference on Neural Information Processing Systems TI - Unsupervised object-centric video generation and decomposition in 3D VL - 33 ER - TY - BOOK AB - This booklet is a collection of abstracts presented at the AHPC conference. ED - Schlögl, Alois ED - Kiss, Janos ED - Elefante, Stefano ID - 7474 SN - 978-3-99078-004-6 TI - Austrian High-Performance-Computing meeting (AHPC2020) ER - TY - CONF AB - Quantization converts neural networks into low-bit fixed-point computations which can be carried out by efficient integer-only hardware, and is standard practice for the deployment of neural networks on real-time embedded devices. However, like their real-numbered counterpart, quantized networks are not immune to malicious misclassification caused by adversarial attacks. We investigate how quantization affects a network’s robustness to adversarial attacks, which is a formal verification question. We show that neither robustness nor non-robustness are monotonic with changing the number of bits for the representation and, also, neither are preserved by quantization from a real-numbered network. For this reason, we introduce a verification method for quantized neural networks which, using SMT solving over bit-vectors, accounts for their exact, bit-precise semantics. We built a tool and analyzed the effect of quantization on a classifier for the MNIST dataset. We demonstrate that, compared to our method, existing methods for the analysis of real-numbered networks often derive false conclusions about their quantizations, both when determining robustness and when detecting attacks, and that existing methods for quantized networks often miss attacks. Furthermore, we applied our method beyond robustness, showing how the number of bits in quantization enlarges the gender bias of a predictor for students’ grades. AU - Giacobbe, Mirco AU - Henzinger, Thomas A AU - Lechner, Mathias ID - 7808 SN - 03029743 T2 - International Conference on Tools and Algorithms for the Construction and Analysis of Systems TI - How many bits does it take to quantize your neural network? VL - 12079 ER - TY - CONF AB - Isomanifolds are the generalization of isosurfaces to arbitrary dimension and codimension, i.e. manifolds defined as the zero set of some multivariate vector-valued smooth function f: ℝ^d → ℝ^(d-n). A natural (and efficient) way to approximate an isomanifold is to consider its Piecewise-Linear (PL) approximation based on a triangulation 𝒯 of the ambient space ℝ^d. In this paper, we give conditions under which the PL-approximation of an isomanifold is topologically equivalent to the isomanifold. The conditions are easy to satisfy in the sense that they can always be met by taking a sufficiently fine triangulation 𝒯. This contrasts with previous results on the triangulation of manifolds where, in arbitrary dimensions, delicate perturbations are needed to guarantee topological correctness, which leads to strong limitations in practice. We further give a bound on the Fréchet distance between the original isomanifold and its PL-approximation. Finally we show analogous results for the PL-approximation of an isomanifold with boundary. AU - Boissonnat, Jean-Daniel AU - Wintraecken, Mathijs ID - 7952 SN - 1868-8969 T2 - 36th International Symposium on Computational Geometry TI - The topological correctness of PL-approximations of isomanifolds VL - 164 ER - TY - CONF AB - Given a finite point set P in general position in the plane, a full triangulation is a maximal straight-line embedded plane graph on P. A partial triangulation on P is a full triangulation of some subset P' of P containing all extreme points in P. A bistellar flip on a partial triangulation either flips an edge, removes a non-extreme point of degree 3, or adds a point in P ⧵ P' as vertex of degree 3. The bistellar flip graph has all partial triangulations as vertices, and a pair of partial triangulations is adjacent if they can be obtained from one another by a bistellar flip. The goal of this paper is to investigate the structure of this graph, with emphasis on its connectivity. For sets P of n points in general position, we show that the bistellar flip graph is (n-3)-connected, thereby answering, for sets in general position, an open questions raised in a book (by De Loera, Rambau, and Santos) and a survey (by Lee and Santos) on triangulations. This matches the situation for the subfamily of regular triangulations (i.e., partial triangulations obtained by lifting the points and projecting the lower convex hull), where (n-3)-connectivity has been known since the late 1980s through the secondary polytope (Gelfand, Kapranov, Zelevinsky) and Balinski’s Theorem. Our methods also yield the following results (see the full version [Wagner and Welzl, 2020]): (i) The bistellar flip graph can be covered by graphs of polytopes of dimension n-3 (products of secondary polytopes). (ii) A partial triangulation is regular, if it has distance n-3 in the Hasse diagram of the partial order of partial subdivisions from the trivial subdivision. (iii) All partial triangulations are regular iff the trivial subdivision has height n-3 in the partial order of partial subdivisions. (iv) There are arbitrarily large sets P with non-regular partial triangulations, while every proper subset has only regular triangulations, i.e., there are no small certificates for the existence of non-regular partial triangulations (answering a question by F. Santos in the unexpected direction). AU - Wagner, Uli AU - Welzl, Emo ID - 7990 SN - 18688969 T2 - 36th International Symposium on Computational Geometry TI - Connectivity of triangulation flip graphs in the plane (Part II: Bistellar flips) VL - 164 ER - TY - CONF AB - In a straight-line embedded triangulation of a point set P in the plane, removing an inner edge and—provided the resulting quadrilateral is convex—adding the other diagonal is called an edge flip. The (edge) flip graph has all triangulations as vertices, and a pair of triangulations is adjacent if they can be obtained from each other by an edge flip. The goal of this paper is to contribute to a better understanding of the flip graph, with an emphasis on its connectivity. For sets in general position, it is known that every triangulation allows at least edge flips (a tight bound) which gives the minimum degree of any flip graph for n points. We show that for every point set P in general position, the flip graph is at least -vertex connected. Somewhat more strongly, we show that the vertex connectivity equals the minimum degree occurring in the flip graph, i.e. the minimum number of flippable edges in any triangulation of P, provided P is large enough. Finally, we exhibit some of the geometry of the flip graph by showing that the flip graph can be covered by 1-skeletons of polytopes of dimension (products of associahedra). A corresponding result ((n – 3)-vertex connectedness) can be shown for the bistellar flip graph of partial triangulations, i.e. the set of all triangulations of subsets of P which contain all extreme points of P. This will be treated separately in a second part. AU - Wagner, Uli AU - Welzl, Emo ID - 7807 SN - 9781611975994 T2 - Proceedings of the Annual ACM-SIAM Symposium on Discrete Algorithms TI - Connectivity of triangulation flip graphs in the plane (Part I: Edge flips) VL - 2020-January ER - TY - GEN AB - The Mytilus complex of marine mussel species forms a mosaic of hybrid zones, found across temperate regions of the globe. This allows us to study "replicated" instances of secondary contact between closely-related species. Previous work on this complex has shown that local introgression is both widespread and highly heterogeneous, and has identified SNPs that are outliers of differentiation between lineages. Here, we developed an ancestry-informative panel of such SNPs. We then compared their frequencies in newly-sampled populations, including samples from within the hybrid zones, and parental populations at different distances from the contact. Results show that close to the hybrid zones, some outlier loci are near to fixation for the heterospecific allele, suggesting enhanced local introgression, or the local sweep of a shared ancestral allele. Conversely, genomic cline analyses, treating local parental populations as the reference, reveal a globally high concordance among loci, albeit with a few signals of asymmetric introgression. Enhanced local introgression at specific loci is consistent with the early transfer of adaptive variants after contact, possibly including asymmetric bi-stable variants (Dobzhansky-Muller incompatibilities), or haplotypes loaded with fewer deleterious mutations. Having escaped one barrier, however, these variants can be trapped or delayed at the next barrier, confining the introgression locally. These results shed light on the decay of species barriers during phases of contact. AU - Simon, Alexis AU - Fraisse, Christelle AU - El Ayari, Tahani AU - Liautard-Haag, Cathy AU - Strelkov, Petr AU - Welch, John AU - Bierne, Nicolas ID - 13073 TI - How do species barriers decay? concordance and local introgression in mosaic hybrid zones of mussels ER - TY - GEN AB - Domestication is a human-induced selection process that imprints the genomes of domesticated populations over a short evolutionary time scale, and that occurs in a given demographic context. Reconstructing historical gene flow, effective population size changes and their timing is therefore of fundamental interest to understand how plant demography and human selection jointly shape genomic divergence during domestication. Yet, the comparison under a single statistical framework of independent domestication histories across different crop species has been little evaluated so far. Thus, it is unclear whether domestication leads to convergent demographic changes that similarly affect crop genomes. To address this question, we used existing and new transcriptome data on three crop species of Solanaceae (eggplant, pepper and tomato), together with their close wild relatives. We fitted twelve demographic models of increasing complexity on the unfolded joint allele frequency spectrum for each wild/crop pair, and we found evidence for both shared and species-specific demographic processes between species. A convergent history of domestication with gene-flow was inferred for all three species, along with evidence of strong reduction in the effective population size during the cultivation stage of tomato and pepper. The absence of any reduction in size of the crop in eggplant stands out from the classical view of the domestication process; as does the existence of a “protracted period” of management before cultivation. Our results also suggest divergent management strategies of modern cultivars among species as their current demography substantially differs. Finally, the timing of domestication is species-specific and supported by the few historical records available. AU - Arnoux, Stephanie AU - Fraisse, Christelle AU - Sauvage, Christopher ID - 13065 TI - VCF files of synonymous SNPs related to: Genomic inference of complex domestication histories in three Solanaceae species ER - TY - CONF AB - This work analyzes the latency of the simplified successive cancellation (SSC) decoding scheme for polar codes proposed by Alamdar-Yazdi and Kschischang. It is shown that, unlike conventional successive cancellation decoding, where latency is linear in the block length, the latency of SSC decoding is sublinear. More specifically, the latency of SSC decoding is O(N 1−1/µ ), where N is the block length and µ is the scaling exponent of the channel, which captures the speed of convergence of the rate to capacity. Numerical results demonstrate the tightness of the bound and show that most of the latency reduction arises from the parallel decoding of subcodes of rate 0 and 1. AU - Mondelli, Marco AU - Hashemi, Seyyed Ali AU - Cioffi, John AU - Goldsmith, Andrea ID - 8536 SN - 21578095 T2 - IEEE International Symposium on Information Theory - Proceedings TI - Simplified successive cancellation decoding of polar codes has sublinear latency VL - 2020-June ER -