Transfer Learning for Human & AI

Krylov subspace methods are an essential building block in numerical simulation software . The efficient utilization of modern hardware is a challenge in the development of these methods . Tothis end, we analyze an innovative block Krylovsubspace framework that allowsto balance the computational and data-transfer costs to the hardware .…

Emotion recognition provides valuable information for human-computerinteractions . The large number of input channels involved in emotion recognition are significantly expensive from a memory perspective . The required vector storage is seamlessly reduced by 98% and the frequency of vector requests by at least 1/5 .…

According to one algorithm, PageRank, the worldwide web structure is represented by the Google matrix . Finding such a dominant eigenvector on an ever-growing number of web pages becomes a computationally intensive task incompatible with Moore’s Law . We discuss the feasibility of simulating the PageRank algorithm on large Google matrices using such unconventionalhardware .…

We review state-of-the-art formal methods applied to the verification of machine learning systems . Formal methods can providerigorous correctness guarantees on hardware and software systems . Applying formal methods to verify systems that include machinelearning has only been considered recently .…

Backdoor attack intends to inject hidden backdoor into deep neuralnetworks (DNNs) The prediction of infected models will bemaliciously changed if the hidden backdoor is activated by the attacker-defined trigger . Currently, most existing backdoor attacks adopted the setting of static trigger, $i.e.,$…

Hyperloop system design is a uniquely coupled problem because it involves the design of a complex, high-performance vehicle and its accompanying infrastructure . This work presents a system optimization tool, HOPS, that has been adopted as a central component of the Virgin Hyperloop design process .…

This review is one of the first to focus specifically on gender bias in applications of machine learning . We detail the most widelyused formalizations of fairness in order to address how to make machinelearning models fairer . We discuss the most influential biasmitigation algorithms as applied to domains in which models have a highpropensity for gender discrimination .…

A strong link between complexity theory and literature is possible, i.e. under one proviso, namely that total novels be considered . This paper argues that a total novel ismost conspicuous example of a complex system . Science and literature can be complemented and developed, hand in hand.…

Contact tracing is a very effective way to control the COVID-19-likepandemics. It aims to identify individuals who closely contacted an infectedperson during the incubation period of the virus and notify them to quarantine . However, existing systems suffer from privacy, security, and efficiency issues .…

Synthesizing graceful and life-like behaviors for physically simulated characters has been a fundamental challenge in computer animation . We propose to obviate the need to manually design imitation objectives and mechanisms for motion selection by utilizing a fully automated approach .…

Abraham, Dolev, Geffner, and Halpern proved that, in asynchronous systems, a$(k,t)$-robust equilibrium for $n$ players and a trusted mediator can beimplemented without the mediator as long as $n 4(k+t)$. We prove that this bound istight, in the sense that if $n \le 4(K+T)$ there exist .…

Adaptive Noise Injection (ANI) uses alight-weight DNN on the client-side to inject noise to each input, before sending it to the service provider to perform inference . The method doesnot require prior knowledge of the sensitive attributes and incurs minimalcomputational overheads .…

We introduce and investigate symbolic proof systems for Quantified BooleanFormulas (QBF) operating on Ordered Binary Decision Diagrams (OBDDs) Thesesystems capture QBF solvers that perform symbolic quantifier elimination . We obtain exponential separations from standardclausal proof systems, specifically (long-distance) QU-Resolution and IR-Calc .…

A strong link between complexity theory and literature is possible, i.e. under one proviso, namely that total novels be considered . This paper argues that a total novel ismost conspicuous example of a complex system . Science and literature can be complemented and developed, hand in hand.…

Leadership plays a key role in social animals, including humans, in coordinated activities . Understanding initiation of coordinated activities allows scientists to gain more insight into social speciesbehaviors . In this paper, we formalize FactionInitiator Inference Problem and propose a leadership inference framework .…

Qualitative and quantitative approaches to reasoning about uncertainty can lead to different logical systems for formalizing such reasoning . In the case of reasoning about relative likelihood, a standardqualitative approach using preordered preferential structures yields adramatically different logical system than a quantitative approach usingprobability measures .…

Scene graphs are a powerful structured representation of the underlying content of images, and embeddings derived from them have been shown to be useful in multiple downstream tasks . In this work, we employ a graphconvolutional network to exploit structure in scene graphs and produce imageembeddings useful for semantic image retrieval .…

New computing paradigms, driven by requirements of different classes of problems and applications, emerge rapidly . We discuss how the development simulators can be cost-effective in accelerating design of realsystems . We highlight strategic role played by different types of publications, conferences, and educational programs .…

New WebHID API allows developers to write Web drivers in pureJavaScript that talk to Human Interface Device (HID) devices via the HID Protocol . One example of a driver has been realized in the projectJoy-Con-WebHID, which allows for fun pastimes like playing the Google Chromebrowser’s offline dinosaur game by jumping .…

Zeus trains an agent that learns to adaptively modify the input video segments to an action classification network . Zeus is capable of answering the query at a user-specified target accuracy using a query optimizer that trains the agent based on an accuracy-aware reward function .…

We develop a theory of vector spaces spanned by orbit-finite sets . We give a decision procedure for equivalence of weighted registerautomata . The algorithm runs in exponential time, and inpolynomial time for a fixed number of registers . We candecide, with the same complexity, language equivalence for unambiguous registerautomicomata, which improves previous results in three ways: (a) we allow for ordercomparisons on atoms, and not just equality; (b) the complexity isexponentially better .…

Point-of-Interest (POI) recommendation is an important task in location-based networks . We propose to jointly model variousspatial-temporal factors by context-aware non-successive modeling . In the long-term module, we split user’s all historical check-ins into seven sequences, then we combine them by attention .…

Vision-based Continuous Sign Language Recognition (CSLR) aims to recognizeunsegmented gestures from image sequences . To solve this problem, we propose a Visual Alignment Constraint (VAC) to enhance the feature extractor with more alignment supervision . The proposed VAC achieves competitive performance on two challenging CSLR datasets and experimental results show its effectiveness .…

Much has changed since the inaugural ICGSE conference in 2006 . Tools have improved and awareness of cultural differences is widespread . But the pervasive and profound impact of software inthe world — especially for societal scale systems such as social media –forces new and deeply challenging responsibilities on both developers andademics .…

This research represents an attempt to ignite the growth of a crowd sourced simulation ecosystem of reusable subcomponents for Agent-Based Models . Due to the inherent complexity of simulations, developing this ecosystem will be moredifficult than other knowledge sharing ecosystems, such as machine learning libraries .…

CodeTrans is an encoder-decoder transformer model for tasks in the software engineering domain . CodeTrans outperforms the state-of-the-art models on all the tasks . These breakthroughs point out apromising direction for process source code and crack software engineering tasks . To expeditefuture works in thesoftware engineering domain, we have published ourpre-trained models of CodeTrans.…

There are several isomorphic constructions for the irreducible polynomialrepresentations of the general linear group in characteristic zero . The twomost well-known versions are called Schur modules and Weyl modules . Steven Samused a Weyl module implementation in 2009 for his Macaulay2 package PieriMaps .…

The protocol essentially pairs each row in the dataset with a fake row and perform simple randomization on all rows . Experiments on real-world language confirm that the protocol’s error is small . We also bound the impact corrupt users have on the estimates and show that it compares favorably to prior state-of-the-art models of the shuffler .…

RFquack is an open-source tool and library firmware that combinesthe flexibility of a software-based approach with the determinism and performance of embedded RF frontends . It has an IPython shell and 9 firmware modules for: spectrum scanning, automatic carrier detection and bitrate estimation, in-flight packet filtering andmanipulation, MouseJack, and RollJam (as examples).…

Mobile edge computing (MEC) has been envisioned as a promising paradigm to handle the massive volume of data generated from ubiquitous mobile devices forenabling intelligent services with the help of artificial intelligence (AI) The integration of FL and blockchain leads to a new paradigm, called FLchain, which potentially transforms intelligent MEC networks intodecentralized, secure, and privacy-enhancing systems .…

The objective of this paper is to provide a unifying view of these identities . In the first part of the paper, a general derivative identity for theconditional mean is derived . The new identity is used to generalize many of the known identities and derive some newones .…

Malware has become a widely used means in cyber attacks in recent decades . Static Malware analysis is effective for known malwares but it does not work for zero day malware which leads to the need of dynamic malware detection .…

The problem of repairing linear codes and, in particular, Reed Solomon (RS)codes has attracted a lot of attention in recent years . In this problem, a failed codesymbol (node) needs to be repaired by downloading as little information as possible from a subset of the remaining nodes .…

The rapid development of artificial intelligence, especially deep learning technology, has advanced autonomous driving systems . However, the safety and security of deep learning-based autonomousdriving are severely challenged by these attacks . This survey provides a thorough analysis of different attacksthat may jeopardize ADSs, as well as the corresponding state-of-the-art defensemechanisms .…

PI-Net is a privacy-preserving mechanism that achieves image obfuscation with PI guarantee . PI-net achieves significantly better privacy utility trade-off through public imagedata, says study . We propose a formal privacy notion for images, with the consideration of the perceptual similarity, we proposeperceptual indistinguishability (PI) We also propose PI-NET, a privacy protection mechanism that protects images from public data, as well as the privacy-utility trade-offs associated with the privacy of the image .…

Jekyll is a neural styletransfer framework that takes as input a biomedical image of a patient andtranslates it to a new image that indicates an attacker-chosen disease condition . The potential for fraudulent claims based on such generated ‘fake’medical images is significant, and we demonstrate successful attacks on bothX-rays and retinal fundus image modalities.…

Ground segmentation of point clouds remains challenging because of the sparseand unordered data structure . This paper proposes the GSECnet – GroundSegmentation network for Edge Computing . It is designed to be deployable on a low-poweredge computing unit . Remarkably, our framework achieves the inference runtime of 135.2Hz on a desktop platform.…

We study the fundamental problem of frequency estimation under both privacyand communication constraints . We adopt the model of multiparty differential privacy(MDP), which is more general than local differential privacy (LDP) and centralized) differential privacy . Our protocols achieve optimality (up tologarithmic factors) permissible by the more stringent of the two constraints .…

In this paper, we propose a novel wireless architecture mounted on a high-altitude aerial platform . By installing RIS on the aerial platform, rich line-of-sight and full-area coverage can be achieved, thereby, overcoming the limitations of the conventional terrestrial RIS .…

This paper introduces a dual-critic reinforcement learning (RL) framework to address the problem of frame-level bit allocation in HEVC/H265 . The objectiveis to minimize the distortion of a group of pictures (GOP) under a rateconstraint . Previous RL-based methods tackle such a constrained optimizationproblem by maximizing a single reward function .…

Most existing RGB-D SOD models have no cross-modal interactions or have unidirectional interactions from depth to RGB in their encoder stages . We propose to conduct progressivebi-directional interactions as early in the encoder stage . We also design an effective light-weight group decoder to achieve accurate final saliencyprediction .…

Understanding the context of complex and cluttered scenes is a challenging problem for semantic segmentation . It is difficult to model the context without prior and additional supervision because the scene’s factors, such as the scale, shape, and appearance of objects, vary considerably in the scene .…

The characteristics of high loss rate, resource constraint, being eager forgood security haven’t been fully considered in the existing key establishmentprotocols of wireless sensor networks . Existing protocols are not practicable enoughdue to their overlong agreement packets and single round key establishment .…

Traditionally, for most machine learning settings, gaining some degree ofexplainability that tries to give users more insights into how and why thenetwork arrives at its predictions, restricts the underlying model and hinders performance to a certain degree . For example, decision trees are thought of asbeing more explainable than deep neural networks but they lack performance on visual tasks .…

In this paper, we study the statistical characterization and modeling of distributed multi-reconfigurable intelligent surface (RIS)-aided wireless systems . In the ERA scheme, all RISs participate in assisting thecommunication of a pair of transceivers, whereas in the ORA scheme, only themost appropriate RIS participates and the remaining RISs are utilized for other purposes .…

A neural network tends to firstunder-fit the minor classes by classifying most of their data into the major classes in early training epochs . We argue that such an under-fitting phase over-emphasizes the competition between major and minor classes, hindersthe neural network from learning the discriminative knowledge that can be begeneralized to test data .…

Fast channel estimation in millimeter-wave (mmWave) systems is a fundamentalenabler of high-gain beamforming, which boosts coverage and capacity . The proposed approach, referred to as transformed spatial domain channelestimation (TSDCE) algorithm, exhibits robustness to additive white Gaussiannoise by combining low-rank approximations and sample autocorrelation functions .…

Reshaping accurate and realistic 3D human bodies poses a fundamental challenge forperson identification, online shopping and virtual reality . Existing approaches often suffer from complex measurement by rangecameras or high-end scanners, which either involve heavy expense cost or resultin low quality .…

In Proof-of-Work (PoW) blockchains, the average waiting time to generate ablock is inversely proportional to the computing power of the miner . A group of individual miners can form amining pool to aggregate their computing power to solve the puzzle together and share the reward contained in the block .…