Transfer Learning for Human & AI

FedSVD protects raw data through a singular valueinvariance mask, which can be further removed from the SVD results . The reconstruction error is around 0.000001% of the raw data,validating the lossless property of FedSvd . The scalability of the algorithm is nearly the same as the standalone SVD algorithm, says the authors .…

To model and forecast flight delays accurately, it is crucial to harness various vehicle trajectory and contextual sensor data on airport tarmac areas . These heterogeneous sensor data, if modelled correctly, can be used to generate a situational awareness map .…

We describe a piecewise collocation method for computing periodic solutions of renewal equations . Then, we rigorously prove its convergence under the framework proposed in [S. Maset, Numer. Math.,133 (2016), pp. 525–555] We show some numerical experiments on models frompopulations dynamics which confirm the order of convergence obtained theoretically .…

The State of AI Ethics aims to help anyone, from machine learning experts to humanrights activists and policymakers, quickly digest and understand the latest developments in the field . Report distills the research and reporting surrounding various domains related to the ethics of AI, with aparticular focus on four key themes: Ethical AI, Fairness & Justice, Humans &Tech, and Privacy .…

Current deep learning techniques for style transfer would not be optimal for design support since their “one-shot” transfer does not fit exploratory design processes . To overcome this gap, we propose parametric transcription, which transcribes an end-to-end style transfer effect into parameters available in existing content editing tools .…

We discuss a pointwise numerical differentiation formula on multivariatescattered data . The formula is based on the coefficients of local polynomial interpolation atDiscrete Leja Points . Error bounds and sensitivity estimates to perturbations are provided . Several numerical tests are presented to show the accuracy of the approximation of partial derivatives of any order compatible with thefunction regularity .…

The numerical solution of spectral fractional diffusion problems in the form${\mathcal A}^\alpha u = f$ is studied . The BURA method has almost optimal computational complexity, assuming that anoptimal PCG iterative solution method is applied to the involved auxiliarylinear systems .…

The most commonly used publication metrics for individual researchers are the total number of publications, and Hirsch’s $h$-index . Most other author-level measures in the literature share this cumulative property . By contrast, we aim to studynon-cumulative measures that answer the question “in terms of citation impact,what have you done lately?”…

Dark patterns have emerged as a set of methods to exploit cognitive biases . These patterns can have consequences that might range fromconvenience to global disasters . We present a case of a drug company and anelectronic medical record vendor who colluded to modify the medical record’s interface to induce clinicians to increase the prescription of extended-releaseopioids .…

Poor medication adherence presents serious economic and health problems including compromised treatment effectiveness, medical complications, and billions of dollars in wasted medicine or procedures . There is an urgent need to leverage light, smart, and minimally obtrusive technology such assmartwatches to develop user tools to improve medication use and adherence .…

Skin contraction is an important biophysical process that takes place duringand after the recovery of deep tissue injury . This process is mainly caused by fibroblasts (skin cells) thatexert pulling forces on the surrounding extracellular matrix (ECM) Modelling is done in multiple scales: agent-based modelling on the microscale and continuum-based .…

The methodology is based on Ranganathan’s original faceted approach,contextualized to the problem of classifying substance concepts . The keynovelty is that the hierarchy is built exploiting the visual properties ofsubstance concepts, while the linguistically defined properties ofclassification concepts are only used to describe substance concepts.…

Two operads are said to belong to the same Wilf class if they have the samegenerating series . If an operad has afinite Groebner basis, then the monomial basis of the operad forms anunambiguous context-free language . We discuss the deterministicgrammar which defines the language .…

In this paper, we use a model of large random network game to explore the conditions under which the Nash equilibrium (NE) of the game is affected by a perturbation . We use a phase transition phenomenon observed in finite rankdeformations of large .…

The FRaGenLP algorithm forgenerating random linear programming problems of large dimension $n$ on clustercomputing systems . The algorithm uses two likeness metrics toprevent the addition of a new random inequality that is similar to one already present in the constraint system .…

High-dimensional expanders generalize the notion of expander graphs to higher-dimensional simplicial complexes . Noelementary combinatorial construction with near-optimal expansion is known . Our construction achieves a spectral gap of $Omega ($Omega) for randomwalks on the $k$-dimensional faces, which is only quadratically worse than the optimal bound of $Theta (Theta) The construction is based on a high-dimensional variant of a tensor product .…

A word w is called a synchronizing (recurrent, reset) word of a deterministicfinite automaton (DFA) if w brings all states of the automaton to some state . Cernyconjectured in 1964 that every n-state synchronizing DFA possesses asynchronizing word of length at most (n -1)2.…

Block convolution is a hardware-friendly, simple, yet efficient convolution operation that can avoid the off-chip transfer of intermediate feature maps atrun-time . The fundamental idea is to eliminate thedependency of feature map tiles in the spatial dimension when spatial tiling is used, which is realized by splitting a feature map into independent blocks sothat convolution can be performed separately on individual blocks .…

For the General Factor problem we are given an undirected graph $G$ and foreach a finite set $B_v$ of non-negative integers . Cornu\’ejols (1988) showed that the decision version of General Factor is poly-timesolvable . Dudycz and Paluch (2018) extended this result for the minimizationand maximization versions .…

Most available image denoising methods are supervised where the pairs of noisy/clean pages are required . However, this assumption is rarely met in real settings . Here, we propose a unified end-to-end unsupervised deep learning model, for the firsttime, that can effectively remove multiple types of noise, including salt \&pepper noise, blurred and/or faded text, as well as watermarks .…

Hypersequent GV (HGV) is a modular and extensible corecalculus for functional programming with session types that enjoy deadlockfreedom, confluence, and strong normalisation . HGV exploits hyper-environments, which are collections of type environments, to ensure that structuralcongruence is type preserving . As a consequence we obtain a tight operationalcorrespondence between HGV and HCP, a hypersequent-based process-calculusinterpretation of classical linear logic .…

Numerical approximation of the Boltzmann equation presents a challenging problem due to its high-dimensional, nonlinear, and nonlocal collision operator . The Petrov-Galerkin spectral method uses Chebyshev functions to achieve desired convergence and conservation properties . We demonstrate the superior accuracy of the proposed method in comparison to the Fourier spectral method with the help of the then-uniform fast Fourier transform (NuffT) The basis functions (bothtest and trial functions) are carefully chosen to be carefully chosen .…

Casson fluid flow model tightly coupled with variable coefficients ADR ($VADR$) equation . Casson viscositycoefficient is taken to be dependent upon solute mass concentration . This paper presents the stability and convergence analyses of the stabilized finiteelement solution . The proposed expressions of the stabilization parameters help in obtaining optimal order of convergences .…

A high-order quadrature algorithm is presented for computing integrals overcurved surfaces and volumes whose geometry is implicitly defined by levelsets of multivariate polynomials . Complexgeometry is automatically handled by the algorithm, including, e.g., multi-component domains, tunnels, and junctions arising from multiplepolynomial level sets, as well as self-intersections, cusps, and other kinds of singularities .…

I find that several models for information sharing in social networks can beinterpreted as age-dependent multi-type branching processes . This allows to characterize criticality in(real and random) social networks . For random networks, I develop amoment-closure method that handles the high-dimensionality of these models: By modifying the timing of sharing with followers, all users can be represented by a single representative, while leaving the total progeny unchanged .…

In this paper, the authors derive several contiguous relations for terminating multivariatehypergeometric series . With these contiguous relations one can prove severalrecursion formulas of those series . This theoretical result allows to computeintegrals over products of Jacobi polynomials in a very efficient recursiveway .…

The ability to detect and segment moving objects in a scene is essential forbuilding consistent maps, making future state predictions, avoiding collisions,and planning . We propose a novel approach that pushes the current state of the art in LiDAR-only moving object segmentation forward toprovide relevant information for autonomous robots and other vehicles .…

In 2019, The Centers for Medicare and Medicaid Services (CMS) launched an Artificial Intelligence (AI) Health Outcomes Challenge seeking solutions topredict risk in value-based care for incorporation into CMS Innovation Centerpayment and service delivery models . We propose a comprehensive modeling solution centered on a deeplearning framework for this data .…

Fused deposition modeling (FDM) using mobile robots instead of thegantry-based 3D printer enables additive manufacturing at a larger scale with higher speed . This introduces challenges including accurate localization, control of the printhead, and design of a stable mobile manipulator with lowvibrations and proper degrees of freedom .…

Theory of planned behavior (TPB), technology acceptance model (TAM),and service quality of transit are conceptualized to assess the learningbehavioral intention (LBI) of commuters in Greater Kuala Lumpur . An onlinesurvey was conducted to understand the LBI of 117 travelers who use the technology to engage in the informal learning process during daily commuting .…

A widely established set of unsupervised node embedding methods can beinterpreted as consisting of two distinctive steps: definition of asimilarity matrix based on graph of interest followed by explicit orimplicit factorization of such matrix . We propose a matrixfactorization method based on a loss function that generalizes that of theskip-gram model with negative sampling to arbitrary similarity matrices .…

A recent trend of research in artificial intelligence, called solutiondiversity, has focused on the development of notions of optimality that may be more appropriate in settings where subjectivity is essential . When combined with techniques from parameterized complexity theory, theparadigm of diversity of solutions offers a powerful algorithmic framework to address problems of practical relevance .…

The computation of pointlikes can beinterpreted as the algebraic counterpart of the covering problem . This leadsto the notion of conelikes for the corresponding algebraic framework . We apply this framework to the Trotter-Weil hierarchy and both the full and the half levels of the $\text{FO}^2$ quantifier alternation hierarchy .…

In recent years, developing AI for robotics has raised much attention . The interaction of vision and language of robots is particularly difficult . Giving robots an understanding of visual semantics and languagesemantics will improve inference ability . In this paper, we propose a novelmethod-VSGM (Visual Semantic Graph Memory), which uses the semantic graph to obtain better visual image features, improve the robot’s visual understandingability.…

This study integrates the educators digital competency (DC), as an individualcharacteristic construct of the task-technology fit (TTF) theory, to examine abetter fit between Moodle using and teaching task . The Task-Technology Fit was also found as an influential construct, which positively and significantly affected both Moodles utilization and teachers task performance .…

Support for lower precision computation is becoming more common inaccelerator hardware due to lower power usage, reduced data movement and increased computational performance . However, computational science and engineering (CSE) problems require double precision accuracy in several domains . We seek the best methods for incorporating multiple precisions into theGMRES linear solver; these include iterative refinement and parallelizablepreconditioners .…

We consider two-variable first-order logic and its quantifieralternation hierarchies over both finite and infinite words . Deciding the presence or absence of such a pattern in agiven automaton is in $\mathbf{NL$ . This leads to algorithms for deciding the levels of the $\text{FO}^2$ quantifier alternationhierarchies .…

In [Lehrenfeld, Olshanskii. propose a new discretization method for PDEs on moving domains in thesetting of unfitted finite element methods, which is provably higher-orderaccurate in space and time . The keyingredient in the method, the transformation of the underlying mesh, is becomestime-dependent which gives rise to some technical issues .…

Analyzing structural properties of social networks has many applications . However, these applications are not supported by federated social networks that allow users to store their social links locally on their end devices . The private hierarchical clustertrees enable a service provider to query the community structure around a user at various granularities without the users having to share their raw contact information .…

As AI systems are increasingly involved in decision making, it also becomesimportant that they elicit appropriate levels of trust from their users . We identify that a research gap exists regarding the role of personal values in trust in AI .…

All known algorithms need at least $Omega(m)$ communication for coloring and MIS . We study the communication cost (or message complexity) of these problems . We ask: Can we solve problems such as coloringand MIS using sublinear, i.e., $o(m), communication, and if so under whatconditions?…

We present an algorithm for producing a seamless animated loop from a single image . The algorithm detects periodic structures, such as the windows of abuilding or the steps of a staircase, and generates a non-trivial displacement vector field that maps each segment of the structure onto a neighboring segment .…

Unsupervised outlier detection, which predicts if a test sample is an outlier, is an importantbut challenging task . Recently, methods based on the two-stage frameworkachieve state-of-the-art performance on this task . The framework leverages self-supervised representation learning algorithms to train a feature extractoron inlier data .…

Fake news on Social Media undermines democratic institutions and processes . We propose interventions that focus on individual user empowerment, and social media structural change that is prophylactic (pre exposure), rather than therapeutic (post exposure) We propose three interventions i) psychologicalinoculation, ii) fostering digital and media literacy and i) imposition of user transaction costs .…

Unsupervised deep learning solutions for textline segmentation are beginning to gain popularity . The benefit of our approach is zero manual labelling effort . We use an unsupervised learning method that embeds document image patches to a compact Euclidean space where distancescorrespond to a coarse text line pattern similarity .…

Animals are diverse in shape, but building a deformable shape model for a newspecies is not always possible due to the lack of 3D data . We present a method to capture new species using an articulated template and images of that species .…

Recent progress in network-based audio event classification has shown the benefit of pre-training models on visual data such as ImageNet . While this process allows knowledge transfer across different domains, training a model on large-scale visual datasets is time consuming .…

Montreal AI Ethics Institute’s The State of AI Ethics captures the most relevant developments in AI Ethics since October 2020 . Report includes exclusive content written by world-class AI Ethics experts from universities, research institutes,consulting firms, and governments . Unique to this report is “The Abuse andMisogynoir Playbook,” written by Dr.…

CoTexT is a transformer-based architecture encoder-decoder model that learns the representative context between naturallanguage (NL) and programming language (PL) through multi-task learning . It achieves state-of-the-art on all downstream tasks such as code summarizing/documentation, codegeneration, defect detection, code debugging, etc. We evaluate multi-tasks on different generation and classification tasks on CodeXGLUE and it achieves state of the art on alldownstream tasks .…