Transfer Learning for Human & AI

There is a growing research body of Human-in-the-loop (HITL)NLP frameworks that continuously integrate human feedback to improve the model itself . HITL NLP research is nascent but multifarious — solving various NLPproblems, collecting diverse feedback from different people, and applyingdifferent methods to learn from collected feedback .…

We study a finite-horizon two-person zero-sum risk-sensitive stochastic game . We establish the existence of a solution to the Shapley equation (SE) by an approximation technique . We also develop a value iteration-type algorithm for approaching to the value of the game .…

Since the introduction of the first Bitcoin blockchain in 2008, differentdecentralized blockchain systems have emerged with public and private accessibility . It has been widelyacknowledged that no single blockchain network will fit all use cases . Inrealization of this multi-blockchain paradigm, initiatives in buildingInteroperability-Facilitating Platforms (IFPs) that aim at bridging differentblockchains have come to the fore .…

When data matrices are symmetric, a short-term recurrence can be used to generate an associated orthonormalbasis . In the past this procedure was abandoned because it requires twice thenumber of linear system solves per iteration than with the classical long-termmethod .…

Secure spatial modulation(SM) is becoming a hot research area . Its basic idea is to exploit both theindex of activated transmit antenna and amplitude phase modulation signal to carry messages, improve security, and save energy . To further enhance the secrecy rate (SR) performance, adeep-neural-network (DNN) PA strategy is proposed.…

The no free lunch theorem states that no model is better suited to every problem . This paper addresses this issue by proposing the use of meta-learning to infer population-based black-box optimizers that canautomatically adapt to specific classes of problems .…

This paper investigates a device-to-device (D2D) cooperative computingsystem . An user can offload part of its computation task to nearby idle users with the aid of an intelligent reflecting surface (IRS) We propose tominimize the total computing delay via jointly optimizing the computation taskassignment, transmit power, bandwidth allocation, and phase beamforming of theIRS .…

There are patterns and techniques which software engineers use to ensure their systems gracefully degrade . But tuning and configuration is hard to get right and it is expensive to explore possible changes to components andtechniques in complex systems . To fill these gaps, we propose Quartermaster to model and simulate systems and fault-tolerant techniques .…

The drone-based last-mile delivery is an emerging technology designed toautomate the delivery process by utilizing drones loaded on a truck totransport parcels to customers . A novel greedy algorithm is proposed to solve the problem that incorporates the real-world operational cost of AVs, traveling distances, and varying load capacities of .…

We consider the problem of characterizing an inner bound to the capacity region of a $3-$user classical-quantum interference channel ($3-$CQIC) Thebest known coding scheme for communicating over CQICs is based on unstructuredrandom codes and employs the techniques of message splitting and superpositioncoding .…

In economics, matching theory is coined for pairing two agents in a specific market to reach a stable or optimal state . In computer science, all branches of matching problems have emerged, such as the question-answer matching in informationretrieval, user-item matching in a recommender system, and entity-relation matching in the knowledge graph .…

This work considers the problem of computing the CANDECOMP/PARAFAC (CP)decomposition of large tensors . One popular way is to translate the problem into a sequence of overdetermined least squares subproblems with Khatri-Raoproduct (KRP) structure . In this work, for tensor with different levels ofimportance in each fiber, we present a mini-batch stochastic gradient descent algorithm with .importance…

Versatile video coding (VVC) is the next generation video coding standard developed by the joint video experts team (JVET) and released in July 2020 . VVCintroduces several new coding tools providing a significant coding gain over the HEVC standard . A new algorithm is proposed to encrypt informat-compliant and constant bitrate the transform coefficients (TCs) together with other syntax elements at the level of the entropy encoder .…

A recent study showed that 70% of researchers fail to reproduce their peers’s experiments and more than half fail to replicate their own experiments . We report the development of a programming interfaceto enhance findability and accessibility of research data .…

The vastness of the design space created by the combination of a large number of computational mechanisms, including machine learning, is an obstacle to creating an artificial general intelligence (AGI) The whole-brain architecture approachdivides the brain-inspired AGI development process into the task of designing the brain reference architecture and developing each component using the BRA .…

A supervised learning approach for the solution of large-scale nonlinearstabilization problems is presented . A stabilizing feedback law is trained from a dataset generated from State-dependent Riccati Equation solves . The training phase is enriched by the use gradient information in the loss function, which is weighted through the use of hyperparameters .…

Bugs that persist into releases of video games can have negative impacts on developers and users . We examine 12,122 bug fixes from 723 updates for 30 popular games on the Steam platform . We find that information bugs appear most frequently in updates, while Crash bugs recur the most frequently and are often treated as more severe than other bug types .…

Data errors are a common challenge in machine learning (ML) projects and generally cause significant performance degradation in ML-enabledsoftware systems . Data validation process demands significant engineering resources for tool development and maintenance . We alsopropose a data validation framework (DVF) for systematizing the adoption of adata validation process .…

Microgrid (MG) energy management is an important part of MG operation . Each electrical entity is modeled as an agent which has a neural network topredict its own Q-values . The correlated Q-equilibrium is used to coordinate the operation among agents .…

EVEREST is a novel H2020 project started on October 1st, 2020 that aims to develop a holistic environment for co-design of HPDA applications on heterogeneous, distributed, and secure platforms . EVEREST focuses on programmability issues through a data-driven design approach, theuse of hardware-accelerated AI, and an efficient runtime monitoring withvirtualization support .…

Cyber-attacks continue to grow, both in terms of volume and sophistication . This is aided by an increase in available computational power, expanding attacksurfaces, and advancements in the human understanding of how to make attacksundetectable . Machine learning is utilised to defend against these attacks .…

The rapid dissemination and adoption of smart speakers has enabled substantial opportunities to improve human health . Smart speakers carry several unique advantages that have the potential to catalyze newfields of health research, particularly in out-of-hospital environments . Therecent rise and ubiquity of these smart computing systems hold significantpotential for enhancing chronic disease management, enabling passiveidentification of unwitnessed medical emergencies, detecting subtle changes in behavior and cognition, limiting isolation, and potentially allowing widespread, passive, remote monitoring of respiratory diseases that impact the public health .…

Consensus maximisation (MaxCon) aims to find the largest subset of data that fits the model . This is particularly important where there are large number of outliers (grossor pseudo) in the observed data . Results forboth synthetic and real visual data experiments show that the MBF based algorithm is capable of generating a near optimal solution relatively quickly .…

The external information complexity of afunction $f(x,y)$ is the minimum amount of information a two-party protocolcomputing $f$ must reveal to an outside observer about the input . We obtain new separation results for the two party external informationcomplexity of boolean functions .…

Analysis of the fetal heart rate during pregnancy is essential for monitoring the proper development of the fetus . The challenge lies in the extraction ofthe fetal ECG from the mother’s ECG during pregnancy . This approach has theadvantage of being a reliable and non-invasive technique .…

We propose a simple approach that efficiently embeds mention-entity-entitypairs in dense space through a BERT-based bi-encoder . We achieve a new state-of-the-art 84.28% accuracy on top-50 candidates on the Zeshel dataset, compared to the previous 82.06% on the top-64 of (Wu et al.,…

We introduce task of changing the narrative point of view, wherecharacters are assigned a narrative perspective that is different from the one originally used by the writer . We introduce a benchmark dataset containing a wide range oftypes of narratives annotated with changes in point of point of .…

An increasing share of captured images and videos are transmitted for storage and remote analysis by computer vision algorithms, rather than to be viewed by humans . We introduce ‘Connectors’ that are inserted betweenthe decoder and the task algorithms to enable a direct transformation of the compressed content, which was previously optimized for a specific task, tomultiple other machine tasks .…

Simplicial complexes form an important class of topological spaces that are frequently used to in many applications areas such as computer-aided design, computer graphics, and simulation . The representation learning on graphs, whichare just 1-d simplicial complexes, has witnessed a great attention and success in the past few years .…

We devise corruptedcompositional grammars that probe for limitations in the compositionalinductive bias of frequently used models . We propose a hierarchicalmodel which might show an inductive bias towards relocatable atomic groups oftokens, thus potentially encouraging the emergence of words . We experiment withprobing for the .…

In this paper, we study the controllability of an Expressive TTS system trained on a dataset for a continuous control . The dataset is the Blizzard 2013dataset based on audiobooks read by a female speaker containing a greatvariability in styles and expressiveness .…

Given the boundary representation of a solid model, HexDom creates a hex-dominant mesh by using a semi-automated polycube-based mesh generation method . The resulting hexahedral dominant mesh includeshexahedra, tetrahedra and triangular prisms . We also apply our software to a number of other complex models totest their robustness .…

Transformer architectures have brought about fundamental changes to the computational linguistic field . Success implies drastic changes incross-modal tasks with language and vision . We discuss its current limitations and speculate upon some of the prospects that we find imminent .…

This paper provides the first analysis of the differentially privatecomputation of three centrality measures on arbitrary weighted graphs . We do so by finding lower bounds on the amounts of nois that a randomised algorithm needs to add in order to make the output of each measure differently private .…

In this paper, we investigate the state estimation problem over multiple Markovian packet drop channels . We propose a locally optimal stationary estimator that minimizes the error variance in the steady state . We provide a sufficient condition under which the MS detectability can be decoupled .…

In this paper, we introduce the notion of periodic safety, which requiresthat the system trajectories periodically visit a subset of a forward-invariantsafe set . We design a model predictive controlpolicy for high-level planning with a bound on the rate of change for thereference trajectory to guarantee that periodic safety is achieved .…

Reinforcement learning (RL) provides a model-free approach to designing anoptimal controller for nonlinear dynamical systems . However, the learning process requires a considerable number of trial-and-error experiments using the poorly controlled system, and accumulates wear and tear on the plant .…

Researchers need access to Online Social Network (OSN) activity on real-world events . OSN data has particular implications for social network analysis . Findings lead to a set of guidelines for researchers planning to collect online data streams to infer social networks, say researchers .…

In reinforcement learning (RL) algorithms, exploratory control inputs are used during learning to acquire knowledge for decision making and control, while the true dynamics of a controlled object is unknown . In this paper, we propose an automatic exploration process adjustment method for safe RL incontinuous state and action spaces utilizing a linear nominal model of thecontrolled object .…

Post-fault dynamics of short-term voltage stability (SVS) presentspatial-temporal characteristics, but existing data-driven methods for online SVS assessment fail to incorporate such characteristics into their models effectively . The proposed STGCN utilizes graph convolution to integrate network topologyinformation into the learning model to exploit spatial information .…

This paper presents conditions for ensuring forward invariance of safe setsunder sampled-data system dynamics with piecewise-constant controllers and fixed time-steps . We show that the proposed conditions are lessconservative than those in earlier studies . The advantages of the new approaches are demonstrated via simulations on an obstacle-avoidance problemfor a unicycle agent and on a spacecraft attitude reorientation problem, inwhich the new approach achieve the mission objectives for cases where the existing methods fail .…

This paper studies the extremum seeking control (ESC) problem for a class of nonlinear systems . We focus on a family of constraints allowing to reformulate the original nonlinear system in the so-called input-output normal form . The effectiveness of the proposed ESC scheme is shown through extensive numerical simulations .…

Polyhedral Lyapunov functions can approximate any norm arbitrarily well . They are used to study stability of linear time varying systems without being conservative . However the computational cost associated with using them grows unbounded as the size of their representation increases .…

This paper reports a comprehensive study on the applicability of ultra-scaledferroelectric FinFETs with 6 nm thick hafnium zirconium oxide layer forneuromorphic computing . Process variation, flicker noise, and device agingcharacterization have been performed . A pre-trained neuralnetwork with 97.5% inference accuracy on the MNIST dataset has been adopted asthe baseline .…

This paper presents a novel hapticshared autonomy teleoperation paradigm that uses haptic feedback to inform the user about the inner state of a shared autonomy paradigm, while stillguaranteeing safety . This differs from haptic shared control, which uses the user’s feedback but gives the human operator full control over the robot’s actions .…

This paper deals with linear algebraic equations where the global coefficientmatrix and constant vector are given respectively . We propose two sets of exponentiallystable continuous-time distributed algorithms that do not require the individual agent matrices to be invertible . The first algorithm works for time-varying weight-balanced directednetworks, and the second algorithm is based on estimatingnon-distributed terms in the centralized algorithm using dynamic averageconsensus .…

This paper proposes a Lyapunov regularized RL approach for optimal frequencycontrol for transient stability in lossy networks . Transient stability of power systems is becoming increasingly important because of the growing integration of renewable resources . The study shows that introducing theLyapunova regularization enables the controller to be stabilizing and achievesmaller losses .…