Transfer Learning for Human & AI

Unusually large prize pools in lotteries like Mega Millions and Powerball attract additional bettors . We propose a way to increase the expected value of lottery tickets by minimizing collisions . Our approach involves partitioning the ticket space among different vendors andpairing them off to ensure no collisions among pairs .…

The present paper concerns filtered de la Vall\’ee Poussin (VP) interpolation . This approximation model isinteresting for applications because it combines the advantages of the Lagrange polynomial approximation with the ones of filtered approximation . We analyze the simultaneous approximationprovided by the derivatives of the VP interpolation polynomials .…

We analyze the well-posedness of an anisotropic, nonlocal diffusion equation . We also present an application of the advection-diffusion equation to anomalous transport of solutes . We present a class of fractional-order operators and presentrigorous estimates for the solution of the corresponding anisotrop anomalousdiffusion equations .…

An essential feature of the subdiffusion equations with the $alpha$-ordertime fractional derivative is the weak singularity at the initial time . Theweak regularity of the solution is usually characterized by a regularityparameter . We present a refined fractional-type Gr\”onwall inequality and a rigorous analysis for thetruncation errors .…

Millimeter-wave communication with high throughput and high reliability is poised to be a gamechanger for V2X and VR applications . mmReliable achieves close to 100% reliabilityproviding 1.5 times better throughput than traditional single-beam systems . It creates custom beam patterns with multiple lobes andoptimizes their angle, phase, and amplitude to maximize the signal strength at the receiver .…

A large class of ergodic control algorithms relies on spectral analysis, which suffers from the curse of dimensionality . The proposed solution is efficient both computationally and storage-wise, hencemaking it suitable for its online implementation in robotic systems . The approach is applied to a peg-in-hole insertion task using a 7-axis Franka EmikaPanda robot, where the task can be achieved without the use of force/torque sensors.…

We study the problem of distributed multi-robot coverage over an unknown,non-uniform sensory field . We propose an adaptive coverage algorithm called DeterministicSequencing of Learning and Coverage (DSLC) that schedules learning and coverageepochs such that its emphasis gradually shifts from exploration to exploitation .…

Craniomaxillofacial reconstruction with patient-specific customizedcraniofacial implants (CCIs) is most commonly performed for large-sizedskeletal defects . Because the exact size of skull resection may not be known prior to the surgery, in the single-stage cranioplasty, a large CCI isprefabricated and resized intraoperatively with a manual-cutting process .…

ITS is based on self-organizing wireless networks, known as vehicular ad-hoc networks (VANETs) Mobile vehicles in VANET might play the role of stationary sensors in infrastructure-based networks . They can detect, gather and disseminate real-time data about traffic, driving conditions and potential hazards onroads.…

A trace ratio optimization problem over the Stiefel manifold is investigated from theory and numerical computations . A new frameworkand its instantiated concrete models are proposed and demonstrated on realworld data sets . Numerical results show that the efficiency of the proposed methods and effectiveness of the new multi-view subspace learningmodels are demonstrated .…

Modular training refers to strictly local training without both the forward and the backward pass . End-to-endbackpropagation (E2EBP) is the de facto standard for training deeparchitectures . These alternatives can match or surpass the performance of E2 EBP on challenging datasets such as ImageNet .…

Machine learning models of materials$^{1-5$ accelerate discovery compared toab initio methods . Deep Adaptive Regressive Regressive Weighted Intelligent Network (DARWIN) predicts K$_2$CuX$_3$ (X = Cl, Br) as a promising materialsfamily, based on its electronegativity difference . We synthesized and found these materials to be stable, direct bandgap UV emitters.…

We provide evidence that commonly held intuitions when designing quantum circuits can be misleading . Our findings assumeethat qubits are and will remain a very scarce resource . Our method uses multipleways of decomposing Toffoli gates into Clifford+T gates . We illustrate ourmethod on addition and multiplication circuits using ripple-carry .…

In response to COVID-19, a vast number of visualizations have been created to communicate information to the public . Information exposure in a public health crisis can impact people’s attitudes towards and responses to the crisis andrisks, and ultimately the trajectory of a pandemic .…

Artificial Intelligence-based symptom checkers (AISCs) have sprung up in everyday healthcare practice . Users assess the medical authority of AISCs using various factors includingautomated decisions and interaction design patterns of AISC apps, associations with established medical authorities like hospitals, and comparisons with other technologies .…

Scoring rules aggregate individual rankings by assigning some points to each position in each ranking such that the total sum of points provides the overall ranking of the alternatives . The current scheme seems to be a reasonable compromise in optimising the abovegoals .…

Recent advances in probabilistic modelling have led to a large number of ‘likelihood-free’ algorithms which do not require numerical evaluation of likelihoods . We provide a benchmark with inference tasks and performance metrics . We found that the choice of performance metric is critical, that even state-of-the-art algorithms have substantial room for improvement .…

Identifying defective items in larger sets is a main problem with manyapplications in real life situations . We investigate old and new identifiability conditions contributing this problem both from atheoretical and applied perspective . What is the precise tradeoff between number of nodes and number of paths such that at most $k$ nodes can beidentified unambiguously ?…

Existing air quality forecasting methods cannot effectively model thediffusion processes of air pollutants between cities and monitoring stations, which may suddenly deteriorate the air quality of a region . In this paper, we propose HighAir, i.e., a hierarchical graph neural network-based air quality forecast method, which adopts an encoder-decoder architecture and considers complex air quality influencing factors, e.g.,…

Type4Py is a deep similaritylearning-based type inference model for Python . It uses a neuralnetwork model that learns to discriminate between types of the same kind and dissimilar types in a high-dimensional space . Nearest neighbor search suggests likely type signatures of given Python functions .…

Representation learning (RL) methods learn objects’ latent embeddings where information is preserved by distances . Since distances are invariant to certainlinear transformations, one may obtain different embeddeddings while preserving the same information . Such misalignment worsens the performance of dynamic network inference tasks .…

An important problem in decision theory concerns the aggregation of individual rankings/ratings into a collective evaluation . We illustrate a new aggregation method in the context of the 2007 MSOM’s student paper competition . This approach is potentially useful in managerialdecision making problems by a committee selecting projects from a large set or capital budgeting involving multiple priorities .…

Presentation attack detection (PAD) has received significant attention from research communities and has become a major field of research . Instead of completely extracting hand-crafted texture features or relying only on deep neural networks, we address theproblem via fusing both wide and deep features in a unified neuralarchitecture .…

DCKP has numerous applications and is however computationally challenging (NP-hard) Algorithm combines memetic framework with threshold search to find high quality solutions . Algorithm is highly competitive compared to the state-of-the-art methods . The code ofour algorithm will be made publicly available and the code of our algorithm will also be made public .…

The gold standard to assess respiration during sleep is polysomnography; atechnique that is burdensome, expensive (both in analysis time and measurementcosts), and difficult to repeat . Automation of respiratory analysis can improve test efficiency and enable accessible implementation opportunities worldwide .…

Binary reverse engineering is a challenging task because it often requires reasoning using both domain-specific knowledge and logical inference . We argue that deductivedivedivedatabases serve as a natural abstraction for interfacing between visual analysis tools and high-performance logicalinference engines that compute facts about binaries .…

The remarkable ability of deep neural networks to perfectly fit training datawhen optimized by gradient-based algorithms is yet to be fully explained theoretically . Explanations by recent theoretical works rely on the networks to be wider by orders of magnitude than the ones used in practice .…

In the present work, we propose a consistent and conservative model formultiphase and multicomponent incompressible flows . Each phase has a background fluid called the pure phase, each pair of phases is immiscible, and components are dissolvable in some specific phases .…

Falx is a synthesis-powered visualization tool that allows users to specify visualizations without needing to worry about data layout . Falxautomatically infers the visualization specification and transforms the data to match the design . In a study with 33 data analysts on four visualization tasks, we found that users can effectively adopt Falxto create visualizations they otherwise cannot implement .…

A question answering (QA) system is a type of conversational AI that generates natural language answers to questions posed by human users . QAsystems often form the backbone of interactive dialogue systems, and have been studied extensively for a wide variety of tasks ranging from restaurant recommendations to medical diagnostics .…

Logging, which means recording sequences of events, can be used for analyzing system failures and malfunctions, and identifying the causes and locations of such issues . Modern systems may produce such a large amount of events monitoring every individual event is not feasible .…

Social networks of ‘Restaurants’ and ‘Indoor Recreation’ are more likely to be tight-knit communities of pre-existing friendshipswhereas ‘Places of Worship’ and “Community Amenities’ page categories are morelikely to bridge new friendship ties . Certain types of Facebook pages demonstrate a large degree of clustering suggesting community preference or potential complementarities to concentration .…

The current design of aerodynamic shapes, like airfoils, involvescomputationally intensive simulations to explore the possible design space . In this work, we propose adata-driven shape encoding and generating method, which automatically learnsrepresentations . The representations are then used in the optimization ofsynthesized airfoil shapes based on their aerodynamic performance .…

Randomized Controlled Trials (RCTs) are widely used to improve business and policy decisions in various sectors such as healthcare, education, criminology, and marketing . In many scenarios, multiple parties hold different parts of the data for conducting and analyzing RCTs .…

In the field of mathematical physics, there exist many physically interestingnonlinear dispersive equations with peakon solutions . Peakon solutions are solitary waves with discontinuous first-order derivative at the wave peak . In this paper, we apply the multi-layer physics-informed neural networks (PINNs) deep learning to study the data-driven peakon and periodic peakon solution of somewell-known nonlinear dispersion equations with initial-boundary valueconditions such as the Camassa-Holm (CH) equation, Degasperis-Procesi equation, modified CH equation with cubic nonlinearity, mCH-Novikov equation, andetc.…

Explanations in human-human interactions are socially-situated . Explainable AI(XAI) approaches have been predominantly algorithm-centered . We introduce and explore Social Transparency (ST), a sociotechnically informed perspective thatorporates the socio-organizational context into explaining AI-mediateddecision-making . We suggestedconstitutive design elements of ST and developed a conceptual framework tounpack ST’s effect and implications at the technical, decision-making, and organizational level .…

This paper approaches the integrated lot sizing and scheduling problem(ILSSP) In which non-identical machines work in parallel with non-triangularsequence-dependent setup costs and times, setup carry-over and capacity limit . The proposed matheuristicssignificantly outperformed CPLEX solver with a fixed CPU time limit in most of the tested instances .…

Sound event detection (SED) is a hot topic in consumer and smart city applications . The paper explores the combination of extreme quantization to asmall-footprint binary neural network (BNN) with the highly energy-efficient,RISC-V-based (8+1)-core GAP8 microcontroller . This BNN reaches a 77.9% accuracy, just 7% lower than the full-precision version .…

We investigate how to solve the cross-corpus news recommendation for unseen users in the future . We propose atransfer learning model (dubbed TrNews) for news recommendation to transferthe knowledge from a source corpus to a target corpus . We show through experiments on real-world datasets that TrNews is better than various baselines in terms of four metrics .…

Augmented Reality is the fusion of virtual components and our realsurroundings . The simultaneous visibility of generated and natural objectsoften requires users to direct their selective attention to a specific target . In this study, we investigated whether thistarget is real or virtual by using machine learning techniques to classifyelectroencephalographic (EEG) data collected in Augmented reality scenarios .…

Following onset of the COVID-19 pandemic and subsequent lockdowns, software engineers’ daily life was disrupted and abruptly forced into remoteworking from home . This change deeply impacted typical working routines,affecting both well-being and productivity . Several tech companies allow their employees to work from home indefinitely if they wish to do so .…

We study a heterogeneous wireless sensor network (WSN) where N heterogeneousaccess points gather data from densely deployed sensors and transmittheir sensed information to M heterogeneous fusion centers (FCs) via multi-hopwireless communication . We propose an iterative algorithm to deploy nodes in static and mobile WSNs .…

In e-commerce industry, user behavior sequence data has been widely used in many business units such as search and merchandising to improve their products . In this paper, we propose a Financial Service scenarioDeep learning based Behavior data representation method for Clustering(FinDeepBehaviorCluster) to detect fraudulent transactions .…

AEOLUS is a security overlay that proactively embeds a dynamic acoustic nonce at the time of user interaction . It detects the presence of the embedded nonce inthe recorded speech to ensure freshness of the input speech . AEOLus yields 0.5% FRR at0% FAR for speech re-use prevention upto a distance of 4 meters in threereal-world environments with different background noise levels .…

Street lighting system has always been the traditional manual system ofilluminating the streets in Bangladesh . A dedicated person is posted only to control the street lights of a zone, who roams around the zonal area to switch on and switch off the lights two times a day .…

When watching non-VR games on platforms such as Twitch, spectators just see what the player sees, as the physical interaction is mostly unimportant for the overall impression . In VR, the immersivefull-body interaction is a crucial part of the player experience .…

The task for speaker verification (SV) is to decide an utterance is spoken by a target or imposter speaker . In most SV studies, a log-likelihood ratio(L_LLR) score is estimated based on a generative probability model on speakerfeatures . However, thegenerative model usually focuses on feature distributions and does not have thediscriminative feature selection ability, which is easy to be distracted bynuisance features .…

Edge computing is emerging to enable AIoT applications. Edge computing enables generating insights and making decisions at or near the data source, reducing the amount of data sent to thecloud or a central repository . We propose a framework forfacilitating machine learning at the edge for AIOT applications, to enablecontinuous delivery, deployment, and monitoring of machine learning models atthe edge (Edge MLOps) The contribution is an architecture that includesservices, tools, and methods for delivering fleet analytics at scale.…

Deep learning has shown its power in many applications, including objectdetection in images, natural-language understanding, and speech recognition . But most deep learning models within Android apps can easily be obtained via mature reverse engineering, while themodels’ exposure may invite adversarial attacks .…