@article{6944, abstract = {We study the problem of automatically detecting if a given multi-class classifier operates outside of its specifications (out-of-specs), i.e. on input data from a different distribution than what it was trained for. This is an important problem to solve on the road towards creating reliable computer vision systems for real-world applications, because the quality of a classifier’s predictions cannot be guaranteed if it operates out-of-specs. Previously proposed methods for out-of-specs detection make decisions on the level of single inputs. This, however, is insufficient to achieve low false positive rate and high false negative rates at the same time. In this work, we describe a new procedure named KS(conf), based on statistical reasoning. Its main component is a classical Kolmogorov–Smirnov test that is applied to the set of predicted confidence values for batches of samples. Working with batches instead of single samples allows increasing the true positive rate without negatively affecting the false positive rate, thereby overcoming a crucial limitation of single sample tests. We show by extensive experiments using a variety of convolutional network architectures and datasets that KS(conf) reliably detects out-of-specs situations even under conditions where other tests fail. It furthermore has a number of properties that make it an excellent candidate for practical deployment: it is easy to implement, adds almost no overhead to the system, works with any classifier that outputs confidence scores, and requires no a priori knowledge about how the data distribution could change.}, author = {Sun, Rémy and Lampert, Christoph}, issn = {1573-1405}, journal = {International Journal of Computer Vision}, number = {4}, pages = {970--995}, publisher = {Springer Nature}, title = {{KS(conf): A light-weight test if a multiclass classifier operates outside of its specifications}}, doi = {10.1007/s11263-019-01232-x}, volume = {128}, year = {2020}, } @inproceedings{8324, abstract = {The notion of program sensitivity (aka Lipschitz continuity) specifies that changes in the program input result in proportional changes to the program output. For probabilistic programs the notion is naturally extended to expected sensitivity. A previous approach develops a relational program logic framework for proving expected sensitivity of probabilistic while loops, where the number of iterations is fixed and bounded. In this work, we consider probabilistic while loops where the number of iterations is not fixed, but randomized and depends on the initial input values. We present a sound approach for proving expected sensitivity of such programs. Our sound approach is martingale-based and can be automated through existing martingale-synthesis algorithms. Furthermore, our approach is compositional for sequential composition of while loops under a mild side condition. We demonstrate the effectiveness of our approach on several classical examples from Gambler's Ruin, stochastic hybrid systems and stochastic gradient descent. We also present experimental results showing that our automated approach can handle various probabilistic programs in the literature.}, author = {Wang, Peixin and Fu, Hongfei and Chatterjee, Krishnendu and Deng, Yuxin and Xu, Ming}, booktitle = {Proceedings of the ACM on Programming Languages}, issn = {2475-1421}, number = {POPL}, publisher = {ACM}, title = {{Proving expected sensitivity of probabilistic programs with randomized variable-dependent termination time}}, doi = {10.1145/3371093}, volume = {4}, year = {2020}, } @article{6184, abstract = {We prove edge universality for a general class of correlated real symmetric or complex Hermitian Wigner matrices with arbitrary expectation. Our theorem also applies to internal edges of the self-consistent density of states. In particular, we establish a strong form of band rigidity which excludes mismatches between location and label of eigenvalues close to internal edges in these general models.}, author = {Alt, Johannes and Erdös, László and Krüger, Torben H and Schröder, Dominik J}, issn = {0091-1798}, journal = {Annals of Probability}, number = {2}, pages = {963--1001}, publisher = {Institute of Mathematical Statistics}, title = {{Correlated random matrices: Band rigidity and edge universality}}, doi = {10.1214/19-AOP1379}, volume = {48}, year = {2020}, } @article{15037, abstract = {Protein abundance and localization at the plasma membrane (PM) shapes plant development and mediates adaptation to changing environmental conditions. It is regulated by ubiquitination, a post-translational modification crucial for the proper sorting of endocytosed PM proteins to the vacuole for subsequent degradation. To understand the significance and the variety of roles played by this reversible modification, the function of ubiquitin receptors, which translate the ubiquitin signature into a cellular response, needs to be elucidated. In this study, we show that TOL (TOM1-like) proteins function in plants as multivalent ubiquitin receptors, governing ubiquitinated cargo delivery to the vacuole via the conserved Endosomal Sorting Complex Required for Transport (ESCRT) pathway. TOL2 and TOL6 interact with components of the ESCRT machinery and bind to K63-linked ubiquitin via two tandemly arranged conserved ubiquitin-binding domains. Mutation of these domains results not only in a loss of ubiquitin binding but also altered localization, abolishing TOL6 ubiquitin receptor activity. Function and localization of TOL6 is itself regulated by ubiquitination, whereby TOL6 ubiquitination potentially modulates degradation of PM-localized cargoes, assisting in the fine-tuning of the delicate interplay between protein recycling and downregulation. Taken together, our findings demonstrate the function and regulation of a ubiquitin receptor that mediates vacuolar degradation of PM proteins in higher plants.}, author = {Moulinier-Anzola, Jeanette and Schwihla, Maximilian and De-Araújo, Lucinda and Artner, Christina and Jörg, Lisa and Konstantinova, Nataliia and Luschnig, Christian and Korbei, Barbara}, issn = {1674-2052}, journal = {Molecular Plant}, keywords = {Plant Science, Molecular Biology}, number = {5}, pages = {717--731}, publisher = {Elsevier}, title = {{TOLs function as ubiquitin receptors in the early steps of the ESCRT pathway in higher plants}}, doi = {10.1016/j.molp.2020.02.012}, volume = {13}, year = {2020}, } @article{15036, abstract = {The assembly of a septin filament requires that homologous monomers must distinguish between one another in establishing appropriate interfaces with their neighbors. To understand this phenomenon at the molecular level, we present the first four crystal structures of heterodimeric septin complexes. We describe in detail the two distinct types of G-interface present within the octameric particles, which must polymerize to form filaments. These are formed between SEPT2 and SEPT6 and between SEPT7 and SEPT3, and their description permits an understanding of the structural basis for the selectivity necessary for correct filament assembly. By replacing SEPT6 by SEPT8 or SEPT11, it is possible to rationalize Kinoshita's postulate, which predicts the exchangeability of septins from within a subgroup. Switches I and II, which in classical small GTPases provide a mechanism for nucleotide-dependent conformational change, have been repurposed in septins to play a fundamental role in molecular recognition. Specifically, it is switch I which holds the key to discriminating between the two different G-interfaces. Moreover, residues which are characteristic for a given subgroup play subtle, but pivotal, roles in guaranteeing that the correct interfaces are formed.}, author = {Rosa, Higor Vinícius Dias and Leonardo, Diego Antonio and Brognara, Gabriel and Brandão-Neto, José and D'Muniz Pereira, Humberto and Araújo, Ana Paula Ulian and Garratt, Richard Charles}, issn = {0022-2836}, journal = {Journal of Molecular Biology}, keywords = {Molecular Biology, Structural Biology}, number = {21}, pages = {5784--5801}, publisher = {Elsevier}, title = {{Molecular recognition at septin interfaces: The switches hold the key}}, doi = {10.1016/j.jmb.2020.09.001}, volume = {432}, year = {2020}, }