Transfer Learning for Human & AI

Emotions are usually evoked in humans by images . Recently, extensive researches have been dedicated to understanding the emotions of images . In this chapter, we aim to introduce image emotion analysis (IEA) from a computational perspective . We begin with commonly used emotion representation models frompsychology .…

This paper describes a formal semantics for the Event-B specification language using the theory of institutions . We define an institution for the language, EVT, and prove that it meets validity requirements forsatisfaction preservation and model amalgamation . We also present a series offunctions that show how the constructs of the language can be mapped into our institution .…

Paired/UnpairedInception Discriminative Score (P-IDS/U-IDS) measures the fidelity of inpainted images compared to real images via linearseparability in a feature space . Code is available at https://github.com/zsyzzsoft/co-mod-gan, available at http://://://www.zsyzsoft/coder-gan.org/p-IDS-du-gan . P-IDS is a generic new approach that bridges the gap between image-conditional and recent modulatedunconditional generative architectures via co-modulation .…

Agraph $G$ admits no $K_t-minor but admits a $(\frac32-\varepsilon)t-colouring that is frozen . This disproves three conjectures of Las Vergnas and Meyniel from 1981 .…

Scalable Analytical Analytical Shared Memory Model (SASMM) can predict performance of parallel applications running on a multicore . SASMM uses aprobabilistic and computationally-efficient method to predict the reusedistance profiles of caches in multicores . The model relies on a stochastic, static, static block-level analysis of reuse profiles .…

Users are required to express their queries as a shortlist of words,sentences, or questions . With this short format, a huge amount of information is lost in the process of translating the information need from the actual query size . This loss of information has made the study of queryexpansion technique a strong area of study .…

The concept of cumulativeresidual and past entropy has been proposed in the information theoretic literature . Several properties, including monotonicity and bounds, are obtained for left, right and doubly truncated random variables . One important extension is Kerridgeinaccuracy measure, which are extensions of the corresponding cumulative entropies .…

The Frobenius number $g(S)$ of a set $S$ of non-negative integers with $gcd1$ is the largest integer not expressible as a linear combination of elementsof $S$. In this paper we compute $G_{\bf s} (i)$ for some classical automatic sequences: theevil numbers, the odious numbers, and the lower and upper Wythoff sequences .…

We present progress on local computation algorithms for two coloring of $k$-uniform hypergraphs . We focus on instances that satisfy strengthened assumption of Local Lemma of the form $2^{1-\alpha k} (\Delta+1) e < 1$ We present a procedure that, within similar bounds on running time, solves widerrange of instances by allowing $\alpha$ at most about $0.227$. We discuss how previous works on the subject can be used to obtain an algorithm that worksin polylogarithmic time per query for $0 .139$. …

FPGA-based hardware accelerators have received increasing attention mainly due to their ability to accelerate deep pipelined applications . However, programming such architecture is a challenging task . This paper extends the OpenMP task-based computation offloading model to enable a number of FPGAs to work together as a single Multi-FPGA architecture .…

Low rank tensor approximation is a fundamental tool in modern machinelearning and data science . In this paper, we study the characterization,perturbation analysis, and an efficient sampling strategy for two primarytensor CUR approximations, namely Chidori and Fiber CUR . We also present theoretical error bound of the tensor CUR approximation when (adversarial orGaussian) noise appears .…

Sequential recommendation as an emerging topic has attracted increasing attention due to its important practical significance . Models based on deeplearning and attention mechanism have achieved good performance in sequentialrecommendation . VAE-based models suffer from a common limitation that the representational ability of the obtained approximateposterior distribution is limited, resulting in lower quality of generated samples .…

We associate all small subgraph counting problems with a systematic graphencoding/representation system . The system can serve as a unified foundation for studying andconnecting many important graph problems in theory and practice . We describe relations among graphlets (graph elements) in rigorous mathematicslanguage and from the perspective of graph encoding .…

Reduced basis methods build low-rank approximation spaces for the solutionsets of parameterized PDEs by computing solutions of the given PDE for selected snapshot parameters . Localized reduced basis methods reduce the offline cost of computing these snapshot solutions by insteadconstructing a global space from spatially localized less expensive problems .…

We consider a singularly perturbed reaction diffusion problem as a firstorder two-by-two system . Using piecewise discontinuous polynomials for thefirst component and $H_{div}$-conforming elements for the second component we provide a convergence analysis on layer adapted meshes .…

The paper follows an operadic approach to provide a bialgebraic description of substitution for Lie-Butcher series . We show how the well-known description for substitution in Butcher’s $B$-series can beobtained from the pre-Lie operad . We then apply the same construction to the post-Lieoperad to arrive at a bialgebra $\mathcal{Q}$ .…

The hull of a linear code over finite fields is the intersection of the codeand its dual, which was introduced by Assmus and Key . In this paper, we develop a method to construct linear codes with trivial hull ( LCD codes) and one-dimensional hull .…

We construct semi-discrete-in-time schemes for the incompressible Navier-Stokesequations with periodic boundary conditions . The schemes are of implicit-explicit type based on a scalar auxiliary variable (SAV) approach . It is shown that numerical solutions of these schemesare uniformly bounded without any restriction on time step size .…

Disasters are constant threats to humankind, and beyond losses in lives, they cause many implicit yet profound societal issues such as wealth disparity and digital divide . People from marginalized groups (e.g.,minority, low income, and poor education) are more vulnerable to communicationoutages .…

We show that the problem of code construction for multiple access channel(MAC) resolvability can be reduced to the simpler problem of building source codes . We propose a combination of multiple source codes, used in a black box manner, and leverages randomness recycling via distributed hashing and block-Markov coding .…

We analyse the impact of the selective pressure for the global optimisationcapabilities of steady-state EAs . For the standard bimodal benchmark function\twomax we rigorously prove that using uniform parent selection leads toexponential runtimes with high probability to locate both optima for the standard ($\mu$+1)~EA and $\mu $1/RLS with any polynomial population sizes .…

Reliability network theory was recently used inthe context of polar coding and more generally in connection with decreasingmonomial codes . In this article, we investigate how the concept of averagereliability is applied for polar codes designed for the binary erasure channel .…

ParaMax is able to achieve anear ML-BER performance up to 160×160 and 80×80 Large MIMO for low-ordermodulations such as BPSK and QPSK, respectively, only requiring less than ten sof processing elements . In 12×24 MIMo with 16-QAM at SNR 16dB, in 12 x24 MimO with 16 QAM, in .…

CNN-based algorithm predicts camera pose (3D translation and 3D rotation) from a single image . Pose and uncertainty are learned together with a single loss function . We propose a new fully convolutional architecture, namedCoordiNet, designed to embed some of the scene geometry .…

We present an algorithm for multi-objective optimization of computationallyexpensive problems . The proposed algorithm is based on solving a set ofsurrogate problems defined by models of the real one . Thecompetitiveness of our approach is demonstrated by an experimental comparison with one state-of-the-art surrogate-assisted evolutionary algorithm .…

PoseFormer is a purely transformer-based approach for 3D human pose estimation in videos without convolutionalarchitectures involved . Transformer architectures have become the model of choice in natural language processing . PoseFormer achieves state-of-the-art performance on two benchmark datasets: Human3.6M and MPI-INF-3DHP .…

The structureand existence of deterministic and stationary coding policies that are optimalfor the infinite horizon average cost (distortion) problem is established . Inaddition to existence results, for finite blocklength (horizon) $T$ theperformance of an optimal coding policy is shown to approach the infinite timehorizon optimum at a rate $O(\frac{1}{T)$.…

In the literatur there exist approximation methods for McKean-Vlasovstochastic differential equations which have a computational effort of order$3$. In this article we introduce full-history recursion multilevel Picard(MLP) approximations .…

Service robots are entering more and more in our daily life . It is necessary for the robot to know how to grasp andmanipulate various objects in different situations . Most state-of-the-art grasping approaches addressed fourdegrees of-freedom (DoF) object grasping, where the robot is forced to grasp objects from above based on grasp synthesis of a given top-down scene .…

Nano-sized air robots have the potential to revolutionize the Internet-of-Things ecosystem, acting as autonomous,unobtrusive, and ubiquitous smart sensors . We show how a PULP-based processor, aboard a nano-UAV, is sufficient for the real-time execution (up to 135 frame/s) of our novel DNN, calledPULP-Frontnet .…

We introduce the ‘pdf-related distributions’ These are defined interms of transformation of continuous random variables through theirown probability density functions . We investigate their main characteristics, with reference to the general form of the distribution, the quantiles, and some notions of reliability theory .…

Self-adaptation approaches usually rely on closed-loop controllers that avoid human intervention from adaptation . There are situations where humaninvolvement in the adaptation process is beneficial or even necessary . We investigate alternative human-in-the-loop strategies by using a combination of digitaltwins and virtual reality (VR) interfaces .…

Decoupling optimization methods perform motion planning in the S-L and S-T domains respectively . These methods require a large freeconfiguration space to plan the lane change trajectory . In dense dynamicscenes, it is easy to cause the failure of trajectory planning and be cut in by others .…

Deep operator networks (DeepONets) are receiving increased attention because they can approximate nonlinear operators between infinite-dimensional Banach spaces . Despite their early promise, they typically require large training data-sets consisting of paired input-output observations which may be expensive to obtain .…

Deep neural networks (DNNs) are observed to be successful in patternclassification . However, high classification performances of DNNs are related to their large training sets . Most studies have employed a transformation such as commonspatial patterns (CSP) before the classification process .…

This paper presents an algorithmic framework for the distributed on-linesource seeking, termed as ‘DoSS’, with a multi-robot system in an unknowndynamical environment . The algorithm, building on a novel concept called dummyconfidence upper bound (D-UCB), integrates both estimation of the unknownenvironment and task planning for the multiple robots simultaneously .…

In this paper, we propose a continuous iterative algorithm for theanti-Cheeger cut problem . It does not need rounding at all and has advantages that allsubproblems have explicit analytic solutions . It can also be easily combined with the maxcut iterations for breaking out of local optima and improving the solution quality .…

Process mining acts as a valuable tool to analyse the behaviour of anorganisation by offering techniques to discover, monitor and enhance real processes . The key to process mining is to discovery understandable processmodels . However, real-life logs can be complex with redundant activities, which share similar behaviour but have different syntax .…

Mainly based on neural networks, existing approaches largely model multimodal interactions in an implicit and hard-to-understand manner . We address this limitation with inspirations from quantum theory . In our quantum-inspired framework, the wordinteraction within a single modality and the interaction across modalities areformulated with superposition and entanglement respectively at differentstages .…

Fusion-FlowNet is a sensor fusion framework forenergy-efficient optical flow estimation using both frame- and event-based sensors . The method generalizes well across distinct environments and challenging lighting conditions . It demonstrates state-of-the-art optical flow prediction on the Multi-Vehicle Stereo EventCamera (MVSEC) dataset .…

Recent dataprotection legislation tries to tackle these issues, but it is inadequate . Individuals lack oversight over systems that process their data . Wethink the conflict between data protection and increasingly data-based systemsshould be solved differently . This enables their data sovereignty, allowing individuals to benefit from sensible data usage while addressingpotentially harmful consequences of data mis usage .…

In this paper we introduce a novel family of semantics called weakly completesemantics . The new semantics are conflict-free, non-admissible and employ a weaker notion of admissibility . The semantics are also an example of ambiguity blocking Dunganian semantics and the first semantics totackle the problem of self-defeating attacking arguments .…

This paper presents a novel quasi-centralized approach for collision-freepath planning of multi-robot systems (MRS) in obstacle-ridden environments . Anew formation potential fields (FPF) concept is proposed around a virtual agent, located at the center of the formation . The path of the virtual agent is centrallyplanned and the robots at the minima of the FPF are forced to move along withthe virtual agent .…

In this paper, we investigate dynamic resource scheduling for downlink multi-channel non-orthogonalmultiple access (MC-NOMA) systems over time-varying fading channels . We address the weighted average sum rate maximization problem with quality-of-service (QoS) constraints . We also develop a heuristic joint subchannel assignment and power allocation (Joint-SAPA) algorithm with very low computational complexity that solves each subproblem .…

The reliable fusion of depth maps from multiple viewpoints has become animportant problem in many 3D reconstruction pipelines . The performance of object pose estimation relies heavily on the quality of depth data . Industrial grade depth cameras often fail to sense depth orgenerate unreliable measurements from a single viewpoint .…

The importance of real-time notification has been growing for social services and Intelligent Transporting System (ITS) We present a system that allows a topic to be managed on multiple edge servers so that messages are processed near the publishers, even when publishers are spread over a widearea .…

The two methods are distinct and not compatible . The former prunessolutions, whereas the latter prunes areas of the search that do not have solutions, like standard conflict clauses . Both approaches require what we call \emph{full symmetry, namely apropositionally-consistent mapping between the literals .…

In this paper, we analyze the landscape of the true loss of a ReLU neuralnetwork with one hidden layer . We provide a complete classification of the critical points in the case where the target function is affine . In particular,we prove that local minima and saddle points have to be of a special form and there are no local maxima .…