Transfer Learning for Human & AI

Creating haptic interfaces capable of rendering the rich sensation needed fordexterous manipulation is crucial for the advancement of human-in-the-looptelerobotic systems (HiLTS) One limiting factor has been the absence of detailed knowledge of the effect of operator limb dynamics and hapticexploration dynamics on haptic perception .…

Scheduled sampling is an effective method to alleviate the exposure biasproblem of neural machine translation . It simulates the inference scene by replacing ground-truth target input tokens with predicted ones during training . Despite its success, its critical schedule strategies are merelybased on training steps, ignoring the real-time model competence .…

The micro-macro (mM) decomposition approach is considered for the numericalsolution of the Vlasov–Poisson–Lenard–Bernstein (VPLB) system . In the mM approach, the kineticdistribution function is decomposed as $f=\mathcal{E}[\boldsymbol{\rho}_{f]]$ and $g$, the microscopicdistribution, is defined such that $g$ is defined . We aim to design numerical methods for the .…

In this work we show that high dimensional expansion implies locally testablecode . We define a notion that we callhigh-dimensional-expanding-system (HDE-system) We show that a code that can be modelled over HDE-system islocally testable . This implies that high-dimensional expansion phenomenon solely implies local testability of codes .…

A common approach to improve software quality is to use programming guidelines to avoid common kinds of errors . In this paper, we consider theproblem of enforcing guidelines for Featherweight Java (FJ) We formalize guidelines as sets of finite or infinite execution traces and develop aregion-based type and effect system for FJ that can enforce such guidelines .…

Automatic metrics are commonly used as the exclusive tool for declaring thesuperiority of one machine translation system’s quality over another . We investigate which metrics have the highest accuracy to makesystem-level quality rankings for pairs of systems . We show that the sole use of BLEUnegatively affected the past development of improved models.…

The objective is to schedule a destination visit sequence for a traveler of unit speed to minimizethe Minkowski $p$-norm of the resulting vector of visit/service times . For $p =\infty$ the problem becomes a path variant of the TSP . We also study the all-norm-TSP problem [Golovin et al.…

Quantum computing promises to solve difficult optimization problems more efficiently than classical computers, but requires fault-tolerant quantum computers with millions of qubits . Hybrid algorithms combining classical and quantum computers are used to overcome errors introduced by today’s quantum computers .…

Blockchains based on the Nakamoto consensus protocol have shown promise in several applications, including cryptocurrencies . However, these blockchains have inherent scalability limits caused by the protocol’s consensusproperties . This paper proposes a novel method, Ironclad,that improves blockchain consistency by assigning different weights to randomlyselected blocks .…

Transformer-based encoder-decoder models produce a fused token-wiserepresentation after every encoder layer . We investigate the effects of allowing the encoder to preserve and explore alternative hypotheses, combined at the end of the encoding process . We design and examine a $\textit{Multi-stream Transformer}$ architecture and find that splitting theTransformer encoder into multiple encoder streams and allowing the model to merge multiple representational hypotheses improves performance .…

Abstract reasoning, i.e., inferring complicated patterns from givenobservations, is a central building block of artificial general intelligence . We propose logic-guided generation (LoGe), a novelgenerative DNN framework that reduces abstract reasoning as an optimization problem in propositional logic . LoGe is composed of three steps: extractpropositional variables from images, reason the answer variables with a logiclayer, and reconstruct the answer image from the variables .…

Under assumptions about the data-generative process, the causal graph can often be identified up to anequivalence class . Proposing new realistic assumptions to circumscribe suchequivalence classes is an active field of research . In this work, we propose anew set of assumptions that constrain possible causal relationships based on the nature of the variables .…

Gaussian Bayesian networks are widely used to model causal interactions among continuous variables . In this work, we study the problem of learning a fixed-structureGaussianBayesian network up to a bounded error in total variation distance . We show that the commonly used node-wise least squares regression (LeastSquares) has a near-optimal sample complexity .…

We have a limited understanding of how we integrate haptic information inreal-time from our upper limbs to perform complex bimanual tasks . In order to understand how information from both limbs is used to create a unified percept, it is important to study both the limbs separately first .…

This paper presents a framework for modeling failure in quasi-brittlegeomaterials under different loading conditions . A micromechanics-based model is proposed in which the field variables are linked to physical mechanisms atthe microcrack level . Damage is related to the growth of microcracks, while plasticity is related .…

We have a limited understanding of how we integrate haptic information inreal-time from our upper limbs to perform complex bimanual tasks . In order to understand how information from both limbs is used to create a unified percept, it is important to study both the limbs separately first .…

We present a novel computational paradigm for process design in manufacturing . It incorporates simulation responses to optimize manufacturing processes in high-dimensional temporal and spatial design spaces . This research opens new avenues forhigh-dimensional manufacturing design using solid mechanics simulation toolssuch as finite element methods .…

The widely used domain invariant feature learning (IFL) methods relies on aligning the marginal concept shift w.r.t. $p(x|y)$ and the marginal label shift . In this work, we propose a domain generalization (DG) approach to learn on several labeled source domains and transfer knowledge to a target domain that is inaccessible in training .…

Survey presents an overview of eXplainableArtificial Intelligence (XAI) used in deep learning-based medical image analysis . Paper concludes with an outlook of future opportunities for XAI in medical imaging analysis . The call for explainability of XAI methods grows, especially in high-stakes decision-making areas such asmedical image analysis, says the survey .…

A key challenge of affective computing research is discovering ways to transfer affect models that are built in the laboratory to real world settings . The existing gap between in vitro and in vivo applications is mainly caused by limitations related to affect sensing including intrusiveness, hardware malfunctions, availability of sensors, privacy and security .…

“Bad” data has a direct impact on 88% of companies, with the average company losing 12% of its revenue due to it . Duplicates – multiple but different representations of the same real-world entities – are among the main reasons for poor data quality .…

MFGNet aims to boost the message communication between visible and thermaldata by adaptively adjusting the convolutional kernels for various input images . To address issues caused by heavy occlusion, fast motion, and out-of-view, we propose to conduct a joint local and global search byexploiting a new direction-aware target-driven attention mechanism .…

Qanaat is a scalablemulti-enterprise permissioned blockchain system that guaranteesconfidentiality . The experimental results reveal the efficiency of QanaAT in processing multi-shard and multi-enterprisetransactions . QanaAt consists of multiple enterprises where each enterprisepartitions its data into multiple shards and replicates a data shard on a cluster of nodes to provide fault tolerance .…

A deep neuralnetwork estimates the attenuation of the separated background signal such that the sound quality, quantified using the Artifact-related Perceptual Score, meets an adjustable target . Our experiments show that the proposed method is able to control the trade-off with an accuracy that is adequate for real-world dialogue enhancement applications .…

Measuring meaning is a central problem in cultural sociology and word embeddings may offer powerful new tools to do so . But like any tool, they buildon and exert theoretical assumptions . In certain ways, word embedding methods are vulnerable to the same, enduring critiques of these premises .…

PLanCompS project has developed an initial collection of funcons for translation of functional and imperative languages . Thebehaviour of each funcon is defined, once and for all, using a modular variantof structural operational semantics . The definitions are available online .…

The S.rensen–Dice Coefficient has recently seen rising popularity as aloss function (also known as Dice loss) due to its robustness in tasks where the number of negative samples significantly exceeds that of positive samples . Conventional training of polyphonic sound event detection systemswith binary cross-entropy loss often results in suboptimal detectionperformance as the training is often overwhelmed by updates from negativesamples .…

The current dominance of deep neural networks in natural language processing is based on contextual embeddings such as ELMo, BERT, and BERT derivatives . Most existing work focuses on English; in contrast, we present here the firstmultilingual empirical comparison of two ELMo and several monolingual and multilingual BERT models using 14 tasks in nine languages .…

Autonomous vehicles require large Deep-Learning (DL) models and powerful hardware platforms to operate reliably in real-time . SAGE: a methodology for selectively offloading the key energy-consuming modules of DL architectures to the cloud to optimize edge energy usage while meeting latency constraints .…

Most of the models available in the literaturedepend on strong structural assumptions and they often suffer from the curse ofdimensionality . We propose an end-to-end trainable model built on the CNN andConvolutional LSTM (ConvLSTM) The proposed model focuses on local structures andspatiotemporal correlations .…

This paper specifies an extensive form as a 5-ary relation (i.e. set ofquintuples) which satisfies certain abstract axioms . Each quintuple is understood to list a player, a situation (e.g. information set), a decisionnode, a decision node, an action, and a successor node .…

To examine reproducibility of COVID-19 research, we create a dataset of pre-prints posted to arXiv, bioRxiv, medRxv, and SocArXiv . We extract the text from these pre-prints and parse them looking for keywords signalling theavailability of the data and code underpinning the pre-printed .…

In this paper, an experimental testbed and associated psychophysical paradigmare presented for understanding how people discriminate torsional stiffness . Participants explored virtual virtualtorsion springs by rotating their forearms . The discrimination results will inform futureinvestigation into understanding how stiffness percepts vary .…

Let $H$ be the cartesian product of a family of finite abeliangroups indexed by a finite set . A given poset (i.e., partially orderedset) gives rise to a posetmetric on a given set . We prove that if$P} is Fourier-reflexive, then its dualpartition $Lambda$ coincides with the partition of the dual poset of $\mathbf{H$ .…

We describe cases where real recommender systems were modified in the service of various human values such as diversity, fairness, well-being, time wellspent, and factual accuracy . From this we identify the current practice ofvalues engineering: the creation of classifiers from human-created data with value-based labels .…

We analyze the Fourier growth of various well-studied classes of “structured”$\mathbb{F}_2$-polynomials . This study is motivated by applications inpseudorandomness, in particular recent results and conjectures due to[CHHL19,CHLT19,CGLSS20] We show that any symmetric degree-$d$ $p$ has $L_1$ Fourier weight at level $k$ and this is tight for any constant $k$.…

DeltaCharger is a novel charging robot with an Inverted Delta structure for 3D positioning of electrodes to achieve robust and safe transferring energy . The embedded high-fidelity tactile sensors allow toestimate the angular, vertical and horizontal misalignments between electrodeson the charger mechanism and electrodes on the target robot using pressure data .…

We are looking at families of functions or measures on the torus (indimension one and two) which are specified by a finite number of parameters$N$. The task, for a given family, is to look at a small number of Fouriercoefficients of the object, at a set of locations that is predetermined and maydepend only on $N$ and determine the object .…

Set similarity search is a problem of central interest to a wide variety of applications such as data cleaning and web search . Past approaches on setsimilarity search utilize either heavy indexing structures, incurring largesearch costs or indexes that produce large candidate sets .…

C++ has been adopted as a major HPC language, displacing C to a large extent, and Fortran to some degree as well . MPIs syntax and semantics are defined and extended with C and F91 interfaces that align with the capabilities and limitations of C89 and F77 .…

The widely used domain invariant feature learning (IFL) methods relies on aligning the marginal concept shift w.r.t. $p(x|y)$ and the marginal label shift . In this work, we propose a domain generalization (DG) approach to learn on several labeled source domains and transfer knowledge to a target domain that is inaccessible in training .…

Zero-resource named entity recognition (NER) severely suffers from datascarcity in a specific domain or language . Most studies on zero-resource NERtransfer knowledge from various data by fine-tuning on different auxiliary tasks . In this paper, we tackle the problem by transferring knowledge from three aspects, i.e.,…

In order to apply the recent successes of automated plant phenotyping and machine learning on a large scale, efficient and general algorithms must be bedesigned to intelligently split crop fields into small, yet actionable,portions that can then be processed by more complex algorithms .…

The Light-Up puzzle, also known as the AKARI puzzle, has never been solved using modern artificial intelligence (AI) methods . This project is an effort to apply new AI techniques for solving the Light-up puzzle faster and more computationallyefficient . The algorithms explored for producing optimal solutions include hillclimbing, simulated annealing, feed-forward neural network (FNN), and CNN, and an evolutionary theory algorithm .…

Due to the linearity of quantum operations, it is not straightforward to implement nonlinear transformations on a quantum computer, making some tasks like a neural network hard to be achieved . In this work, wedefine a task called nonlinear transformation of complex amplitudes and provide an algorithm to achieve this task .…

To examine reproducibility of COVID-19 research, we create a dataset of pre-prints posted to arXiv, bioRxiv, medRxv, and SocArXiv . We extract the text from these pre-prints and parse them looking for keywords signalling theavailability of the data and code underpinning the pre-printed .…

We present a parallelized geometric multigrid (GMG) method, based on thecell-based Vanka smoother, for higher order space-time finite element methods . The GMG solver is applied as apreconditioner for GMRES iterations . Its performance properties aredemonstrated for 2d and 3d benchmarks of flow around a cylinder .…

The general adwords problem has remained largely unresolved . We define asubcase called $k-TYPICAL, $k \in \Zplus$ as follows: the total budgetof all the bidders is sufficient to buy $k$ bids for each bidder . We also giverandomized online algorithms for other special cases of adwords .…

We develop deterministic particle schemes to solve non-local scalarconservation laws with congestion . We show that the discrete approximationsconverge to the unique entropy solution with an explicit rate of convergence . We complement our results with somenumerical simulations, among which we show the applicability of the schemes to the multi-species setting .…

$d-logics have not previously been studied in the framework of dynamical systems, which are pairs of a topological space $X$ equipped with a continuous function$f\colon X\to X$. We introduce the logics $wK4C$ and $GLC$ and show that they all have the finite Kripke model property and are sound and complete withrespect to the $d$-semantics in this dynamical setting .…