Transfer Learning for Human & AI

The*nix (Unix/Linux) operating systems offer many tools for working with text files . A procedure for extracting themost frequently occurring multi-word phrases was devised and then demonstrated on several scientific papers in life sciences . The key lesson learned in this work is that semi-automated methods combining the power of algorithms with the power .…

Regulatory Technology (RegTech) is being used to keep track of constantly changing legislation . Organizations need to ensure that their controls (processes)comply with relevant laws, regulations, and policies . To this end, we introduce regulatory information retrieval (REG-IR), an application of document-to-document information retrieval .…

We prove non-asymptotic polynomial bounds on the convergence of the LangevinMonte Carlo algorithmin . The potential is a convex function which is globally Lipschitz on its domain . It is typically the maximum of a finite number of affine functions on an arbitrary convex set .…

Data-driven computational mechanics replaces phenomenological constitutive functions by performing numerical simulations based on data sets ofrepresentative samples in stress-strain space . The distance of modeling values,e.g. stresses and strains in integration points of a finite elementcalculation, from the data set is minimized with respect to an appropriatemetric, subject to equilibrium and compatibility constraints .…

A semi-supervised approach to acoustic source localization in reverberant environments, based on deep generative modeling, is proposed . Localization inreverberant environments remains an open challenge . VAE-SSL is effectively an end-to-end processing approach whichrelies on minimal preprocessing of RTF-phase features .…

In formal concept analysis, it is well-known that the number of formal concepts can be exponential in the worst case . To analyze the average case, we introduce a probabilistic model for random formal contexts and prove that theaverage number offormal concepts has a superpolynomial asymptotic lower bound .…

This paper considers the problem of planning trajectories for a team of sensors-equipped robots to reduce uncertainty about a dynamical process . Methods based on localsearch provide performance guarantees for optimizing a non-monotone submodularfunction . Our results show up to 60% communication reduction and 80-92%computation reduction on average when coordinating up to 10 robots, while outperforming the coordinate descent based algorithm in achieving a desirable trade-off between sensing and energy expenditure .…

Affect-aware socially assistive robotics (SAR) has shown great potential for augmenting interventions for children with autism spectrum disorders . Current SAR cannot yet perceive the unique and diverse set of atypical cognitive-affective behaviors from children with ASD in an automatic and personalized fashion in long-term (multi-session) real-world interactions .…

We present a streaming problem for which every adversarially-robust streaming algorithm must use polynomial space . There exists a classical (oblivious) streaming algorithm that uses only polylogarithmic space .…

Programmingprotocol-independent packet processors (P4) has emerged as the currently most widespread abstraction, programming language, and concept for data planeprogramming . With SDN, users may providetheir own control plane, that can control network devices through their dataplane APIs . Programmable data planes allow users to define their own data planealgorithms for network devices including appropriate data plane APIs which maybe leveraged by user-defined SDN control .…

Sparsity-Inducing Distribution-basedCompression (SIDCo) is a threshold-based sparsification scheme . SIDCo speeds up training by up to 41:7%, 7:6%, and 1:9% compared to the no-compressionbaseline, Topk, and DGC compressors, respectively . It is faster by imposing lower compression overhead than other compressors with minimal overhead .…

For long videos, human referencesummaries necessary for supervised video summarization techniques are difficult to obtain . We show that thesesummaries are at par with human summaries . We also present a study of differentdesired characteristics of a good summary and demonstrate how it is normal tohave two good summaries with different characteristics .…

An attacker who breaks into an authentication server and steals all of thecryptographic password hashes is able to mount an offline-brute force attack against each user’s password . We introduce a Stackelberg game to model the interaction between a defender (authentication server) and an offline attacker .…

An experiment to test whether the presence of an explanation of expected robot behavior affected asupervisor’s trust in an autonomous robot . Participants who received anexplanation of the robot’s behavior were more likely to focus on their own task at the risk of neglecting their robot supervision task during the first trial .…

Spatial Human Accessibility graph for Planning andEnvironment Analysis (SHAPE) is introduced that brings together the technical challenges of discrete representations of digital models, with human-basedmetrics for evaluating the environment . SHAPE: does not need labeled geometryas input, works with multi-level buildings, captures surface variations (e.g.,slopes…

This article considers Hamiltonian mechanical systems with potential functions admitting jump discontinuities . The focus is on accurate and efficient numerical approximations of their solutions . Various numerical evidence, on theorder of convergence, long time performance, momentum map conservation, and consistency with the computationally-expensive penalty method, are supplied .…

The logical separation of control signaling from data transmission in amobile cellular network has been shown to have significant energy saving potential compared with the legacy systems . However, the energy savings of separationarchitecture remain under 16-17% when compared with legacy systems and thisgain falls to a mere 7% when both architectures are realized under a CloudRAN(CRAN) setting .…

In wireless sensor networks, simulation practices, system models, and protocols have been published worldwide based on the assumption of randomness . In this proposal, we show how the results are affected by the generalized assumption made on randomness. Insensor node deployment, ambiguity arises due to node error-value ($\epsilon$), and it’s upper bound in the relative position is estimated to understand thedelicacy of diminutives changes.…

Mereology is the study of parts and the relationships that hold between them . We introduce a behavioral approach to mereology, in which systems and their parts are known only by the types of behavior they can exhibit . We give an inter-modal logic that generalizes the usual alethic modalities in the setting of symmetricaccessibility .…

We propose a framework for resilient autonomous navigation in perceptuallychallenging unknown environments . Environments are GPS-denied andperceptually-degraded with variable lighting from dark to lit and obscurants . The proposed approach was deployed on multiple physical systems including skid-steer and tracked robots, a high-speed RC car and legged robots, as a part of Team CoSTAR’s effort to the DARPA Subterranean Challenge, where the team won 2nd and 1st place in the Tunnel and Urban Circuits, respectively .…

Probabilistic forecasting of high dimensional multivariate time series is anotoriously challenging task . Most previous work either makes simple distributionassumptions or abandons modeling cross-series correlations . We introduce a noveltemporal latent auto-encoder method which enables nonlinear factorization ofmultivariary time series, learned end-to-end with a temporal deep learninglatent space forecast model .…

Humanoid robots built with a purely tree typearchitecture tend to be bulky and usually suffer from velocity and force/torquelimitations . This paper presents a novel series-parallel hybrid humanoid calledRH5 which is 2 m tall and weighs only 62.5 kg . The analysis and control of thishumanoid is performed with whole-body trajectory optimization technique based on differential dynamic programming (DDP) We present an improvedcontact stability soft-constrained DDP algorithm which is able to generatephysically consistent walking trajectories for the humanoid that can be tracked via a simple PD position control in a physics simulator .…

A point process on a space is a random bag of elements of that space . A specialcase of this monad morphism gives us Wald’s Lemma, an identity used to calculate the expected value of the sum of a random number of random variables .…

In graph theory, an independent set is a subset of nodes where there are notwo adjacent nodes . Independent set is maximal if no node outside the independent set can join it . In network applications, maximal independent setscan be used as cluster heads in ad hoc and wireless sensor networks .…

Classical liberalism argues that people should be constrained by no-harm principle when they act . We show that rational players constrained by the NHP will produce Pareto efficient outcomes in n-person non-cooperative games . We alsoshow that both rationality and NHP are required for this result .…

We present a new compositional approach to compositional game theory (CGT) based upon Arrows . Arrows, a concept originally from functional programming, and operators to build new Arrows from old . We also model strategiesas graded Arrows and define an operator which builds a new Arrow by taking thecolimit of a graded Arrow .…

Service robots are envisioned to be adaptive to their working environment based on situational knowledge . We use knowledge graphs (KGs) as a common ground for knowledgeexchange . We develop a pattern library for designing KG interfaces for non-expert users .…

Decentralization has been widely acknowledged as a core virtue of blockchains . This paper presents a study of the degree of decentralization in Bitcoin and Ethereum, the two most prominent blockchains, with various decentralization metrics and different granularities within the time dimension .…

Convolutional layers in CNNs implement linear filters which decompose the input into different frequency bands . We show that the small size of CNN kernels make them susceptible to spectral leakage . To address this issue, we propose theuse of larger kernel sizes along with the Hamming window function to alleviateleakage in CNN architectures .…

Denoising and demosaicking are essential yet correlated steps to reconstructa full color image from the raw color filter array (CFA) data . By learning adeep convolutional neural network (CNN), significant progress has been achieved . Most existing CNN-based methods work on a single image whileassuming additive white Gaussian noise .…

Scientific literature contains large volumes of complex, unstructured figuresthat are compound in nature . Separation of these compound figures is critical for informationretrieval from these figures . In this paper, we propose a new strategy forcompound figure separation, which decomposes the compound figures intoconstituent subfigures .…

Multiple-scale TemporalNetwork trained with top-K Contrastive Multiple Instance Learning (MTN-KMIL) The code is available athttps://://://github.com/tianyu0207/MIL . The method outperforms several state-of-the-art methods by a large margin on three benchmark data sets (ShanghaiTech, UCF-Crimeand XD-Violence) The method is a novelsynthesis of a pyramid of dilated convolutions and a self-attention mechanism, with the former capturing the multi-scale short-range temporal dependencies between snippets and the latter capturing long-range time-relationship of snippets .…

graph neural networks (GNNs) have shown powerful ability to handle few-shot classification problem . GNNs aim to classify unseen samples whentrained with limited labeled samples per class . We propose a new relation measure method to explicitly model the task-level relation of one sample to all the others .…

Deep Convolutional Neural Networks (DCNNs) and their variants have beenwidely used in large scale face recognition(FR) We propose a simple yet efficient training mechanism called MultiFace . The proposed mechanism canaccelerate 2-3 times with the softmax loss and 1.2-1.5 times with Arcface orCosface, while achieving state-of-the-art performances in several benchmarkdatasets .…

We develop a generative network-based ZSLapproach equipped with the proposed Cross Knowledge Learning (CKL) scheme and Taxonomy Regularization (TR) Our approach is superior to the state-of-the-art methods in terms of ZSL image classification and retrieval, we say . We also show that our approach can be superior to those of the state of the art methods .…

In recent years, there has been a shift in facial behavior analysis from the laboratory-controlled conditions to the challenging in-the-wild conditions due to superior performance of deep learning based approaches for many realworld applications . Labeling process of huge training data demands lot of human support with strong domainexpertise for facial expressions or action units, which is difficult to obtainin real-time environments .…

In stable coronary artery disease, reduction in mortality and/ormyocardial infarction with revascularization over medical therapy has not beenreliably achieved . We aim to develop an automatic algorithm by deep learning to extractcoronary arteries from ICAs . After segmentation, an arterial stenosis detection algorithm was developed to extract vascular centerlines and calculate arterial diameter to evaluate stenotic level .…

Using disease masks, we propose a spatial attention method using disease masks thatdescribe the areas where diseases mainly occur . We then apply the spatialattention to find the precise disease area by highlighting the highestprobability of disease occurrence . We show that the proposed method results in superior localization performances compared to state-of-the-art methods .…

This paper studies the dynamics of opinion formation and polarization insocial media . We investigate whether the stance of users with respect to contentious subjects is influenced by online discussions that they are exposed to, and by the interactions with users supporting different stances .…

Visual surveillance aims to stably detect a foreground object using acontinuous image acquired from a fixed camera . Recent deep learning methods based on supervised learning show superior performance compared to classicalbackground subtraction algorithms . However, there is still a room for improvement in static foreground, dynamic background, hard shadow, illumination changes, camouflage, etc.…

EmbedSeg is an end-to-end trainable deep learning method based on the work by Neven et al. It uses the medoid instead of embedding each pixel to thecentroid of any given instance . The pipeline has a small enough memory footprint to be used on virtually all CUDA enabled laptop hardware .…

We consider nearest neighbor weighted random walks on the $d$-dimensional box$[n]^d$ that are governed by some function $g[0,1] The walks cover an abundance of interesting special cases, e.g., the mean-field Potts model, posteriorcollapsed Gibbs sampling for Latent Dirichlet allocation and certain Bayesianposteriors for models in nuclear physics .…

Convolutional neural networks (CNN) have been widely used for boosting the performance of many machine intelligence tasks . The AdderNet can practically achieve 16% enhancement inspeed, 67.6%-71.4% decrease in logic resource utilization and 47.85%-77.9% reduction in power consumption compared to CNN under the same circuitarchitecture .…

Supervision by Registration and Triangulation (SRT) is anunsupervised approach that utilizes unlabeled multi-view video to improve the accuracy and precision of landmark detectors . It can be leveraged to augment existing training data during detectortraining. End-to-end training is made possible by differentiable registrationand 3D triangulation modules.…

Performance testing is part of the development process at MongoDB, and integrated into our continuous integration system . We believe that we have created and exploited a virtuous cycle:performance test improvements drive impact, which drives more use . Overall, MongoDB is gettingfaster and we avoid shipping major performance regressions to our customers because of this infrastructure .…

First-of-its-kind model that recommends 3D movies based on viewer’s subjective ratings accounting correlation between viewer’s visualdiscomfort and stereoscopic-artifact perception . The proposed model is trainedand tested on benchmark Nama3ds1-cospad1 and LFOVIAS3DPh2 S3D video qualityassessment datasets . The experiments revealed that resulting .…

Current unsupervised anomaly detection and localisation systems are commonlyformulated as one-class classifiers that depend on an effective estimation of the distribution of normal images and robust criteria to identify anomalies . However, the current systems tendsto be unstable for classes that are under-represented in the training set, and the anomaly identification criteria commonly explored in thefield does not work well for multi-scale structural and non-structuralanomalies .…

The proposedload balancer reduces the execution response time in big data applicationsperformed on clouds . Scheduling, in general, is an NP-hard problem . We recommend two mathematicaloptimization models to perform dynamic resource allocation to virtual machines and task scheduling . We evaluate the performance of proposed algorithms in terms of response time, turnaround time, throughput metrics, andrequest distribution with some of the existing algorithms that show significantimprovements .…

Internet of Things (IoT) devices are becoming increasingly popular and areinfluencing many application domains such as healthcare and transportation . Existing NN trainingsolutions in this domain either ignore privacy considerations or assume that the privacy requirements are homogeneous across all users .…