Block Full Rank Linearizations of Rational Matrices

Block full rank pencils introduced in [Dopico et al., Local linearizations ofrational matrices] allow us to obtain localinformation about zeros that are not poles of rational matrices . This new family of linearizations is important as it generalizes and includes the structures appearing in most of the .…

Deep Feature Augmentation for Occluded Image Classification

The proposed approach achieves 11.21% and 9.14%average increases in classification accuracy for the ResNet50 networks fine-tuned on the occlusion-exclusive and occlusions-inclusive training sets . The experiments on various datasets and networkstructures show that the deep feature augmentation significantly improves theclassification accuracy of occluded images without a noticeable influence on the performance of clean images .…

Stability and Transparency in Series Elastic Actuation A Two Port Analysis

Series Elastic Actuation (SEA) enables high fidelity and robust force control, improving the safety of physical human-robot interaction (pHRI) Safety is an imperative design criterion that limits interaction performance . In the literature,coupled stability of one-port models of SEA has been studied for various controllers while rendering certain basic environments, and the necessary andsufficient conditions for such passive terminations have been derived .…

Learning Sequences of Manipulation Primitives for Robotic Assembly

Manipulation Primitives have enough complexity to keep the search tree shallow . They are generic enough to generalize across a wide range of assembly tasks . Direct sim2real transfer (without retraining in real) achieves excellent success rates on challenging assembly tasks, such as round peginsertion with 100 micron clearance or square peg insertion with large holeposition/orientation estimation errors .…

The effect of quadrature rules on finite element solutions of Maxwell variational problems Consistency estimates on meshes with straight and curved elements

We study the effects of numerical quadrature rules on error convergence rates on Maxwell-type variational problems . We detail sufficient conditions with respect to mesh refinement and precision for the quadratures rules so as to guarantee convergence rates . On curved domains, we isolatethe error contribution to numerical quirality rules.…

Site to Site Internet Traffic Control

Queues allow network operators to control traffic: where queues build, they can enforce scheduling and shaping policies . Bundler uses a novel inner control loop between asendbox (in the sender’s site) and a receivebox . Enforcing this sending rate ensures that bottleneck queues that would have built up from the bundle’s packets now shift from the bottleneck to the sendbox .…

Self Concordant Analysis of Generalized Linear Bandits with Forgetting

Generalized Linear Bandits (GLB) offer a solid theoretical framework to address them . We propose a novel confidence-based algorithm for the maximum-likehoodestimator with forgetting and analyze its perfomance in abruptly changing environments . We focus on self-concordant GLB (which include logistic and Poisson regression) withforgetting achieved either by the use of a sliding window or exponentialweights .…

Distributed Machine Learning for Computational Engineering using MPI

We propose a framework for training neural networks that are coupled withpartial differential equations (PDEs) in a parallel computing environment . Our parallel computing model views datacommunication as a node in the computational graph for numerical simulation . The advantage of our model is that data communication and computing are cleanlyseparated and thus provide better flexibility, modularity, and testability .…

Identification of Matrix Joint Block Diagonalization

The joint block diagonalization problem (BJBDP) plays an important role in independent subspace analysis (ISA) Numerical simulations validate our theoretical results . We propose a “bi-block diagonalization” method to solve BJBDP, and establish sufficient conditions under which the method is able to accomplish the task .…

DRF A Framework for High Accuracy Autonomous Driving Vehicle Modeling

An accurate vehicle dynamic model is the key to bridge the gap betweensimulation and real road test in autonomous driving . In this paper, we present a Dynamic model-Residual correction model Framework (DRF) for vehicle dynamicmodeling . Compared to classic rule-based and learning-based vehicle dynamic models, DRFaccomplishes as high as 74.12% to 85.02% of absolute trajectory error drop among all DRF variations .…

Semi supervised Autoencoding Projective Dependency Parsing

We describe two end-to-end autoencoding models for semi-supervisedgraph-based dependency parsing . Both models consist of two parts: an encoder enhanced by deep neuralnetworks (DNN) that can utilize the contextual information to encode the input into latent variables, with exactinference . Both LAP and GAP admit a unified structure of loss functions for labeled and unlabeled data with shared parameters .…

What s All the FUSS About Free Universal Sound Separation Data

The Free Universal Sound Separation (FUSS) dataset is a new corpus for experiments in separating mixtures of an unknown number of sounds . The dataset consists of 23 hours of single-sourceaudio data drawn from 357 classes . To simulate reverberation, an acoustic room simulator is used to generate impulse responses of box shaped rooms with frequency-dependentreflective walls .…

Unified greedy approximability beyond submodular maximization

We propose a new class of objective functions of cardinality constrainedmaximization problems for which the greedy algorithm guarantees a constantapproximation . We show a tight bound of $\frac{\alpha$-$\alpha$-augmentablefunctions . For weighted rankfunctions of independence systems, our tight bound becomes $1/q$ forindependence systems of rank quotient at least $q$.…

At most 4 47 n stable matchings

We improve the upper bound for the maximum possible number of stablematchings among $n$ men and $n# women . We develop a novel formulation of a probabilistictechnique that is easy to apply and may be of independent interest in counting other combinatorial objects .…

CVC Contrastive Learning for Non parallel Voice Conversion

CVC only requires one-way GAN training when it comes tonon-parallel one-to-one voice conversion, while improving speech quality and reducing training time . CVC further demonstrates performance improvements in many-out-speaker conversion, enabling the conversion from unseen speakers . We propose CVC, acontrastive learning-based adversarial model for voice conversion .…

Characterizing and Utilizing the Interplay Between Core and Truss Decompositions

Finding dense regions in a graph is an important problem in network analysis . Core decomposition and truss decomposition address this problem from different perspectives . Core-TRUSSDD is an algorithm to identify discrepancies between core andtruss decompositions . Ouralgorithm provides an efficient solution to retrieve the outliers in thenetworks, which correspond to the two anomalous behaviors .…

Search based Kinodynamic Motion Planning for Omnidirectional Quadruped Robots

Autonomous navigation has played an increasingly significant role inquadruped robots system . But existing works on path planning used traditional search-based or sample-based methods which did not consider thekinodynamic characteristics of quadruped robots . The superior performance of our work is demonstrated through simulated comparisons and by using our quadruped robotJueying Mini in our experiments .…

Differential Dynamic Programming with Nonlinear Safety Constraints Under System Uncertainties

Safe operation of systems such as robots requires them to plan and executetrajectories subject to safety constraints . When those systems are subject touncertainties in their dynamics, ensuring that the constraints are not violated is challenging . In this paper, we propose a safe trajectory optimization andcontrol approach (Safe-CDDP) for systems under additive uncertainties and non-linear safety constraints based on constrained differential dynamic dynamicprogramming .…

Focusing Phenomena in Linear Discrete Inverse Problems in Acoustics

The development is given in the context of sound-field reproduction where the source strengths are the inverse solution needed to recreate aprescribed pressure field at discrete locations . The behaviour of the system isfundamentally tied to the amount of acoustic crosstalk at each control point as a result of the focusing operation inherent to the pseudoinverse .…

Budget Sharing for Multi Analyst Differential Privacy

Differentially private mechanisms are designed to answer a single set of queries and optimize the total accuracy across the entire set . Multiple stakeholders that need to be satisfied by the datarelease often have competing goals and priorities . We present novel DPalgorithms that provably satisfy all our desiderata and empirically demonstratethat they incur low error on realistic tasks .…

Blockchain for Decentralization of Internet Prospects Trends and Challenges

This paper focuses on using using using Blockchain to provide a robust and secure decentralizedcomputing system . The paper conducts a literature review on Bitcoin-based methods capable for the decentralization of the future Internet . We have identified three consensus algorithmsbeing PoP, Paxos, and PoAH to be more adequate for reaching consensus inBlockchain-enabled Internet architecture .…

Exponential Polynomial Time Integrators

In this paper we extend the polynomial time integration framework to includeexponential integration for both partitioned and unpartitioned initial value problems . We then demonstrate its utility by constructing a new class ofparallel exponential block methods based on the Legendre points .…

Grasping in the Dark Compliant Grasping using Shadow Dexterous Hand and BioTac Tactile Sensor

The task of imitating human hand in robotic end-effectors, especially in scenarios where visual input is limited or absent, is an extremely challenging one . In this paper we present an adaptive, compliantgrasping strategy using only tactile feedback . The proposed algorithm can grasp objects of varying shapes, sizes and weights without having a priori knowledge of the objects .…