Transfer Learning for Human & AI

In this unhinged rant, I lay out my suspicion that a lot of visualizationsare bullshit . I suspect that bullshit charts take up a large fraction of the time and attention of actual visualization producers and consumers, and yet are apparently absent from academic research into visualization design .…

A combination of geometric and force-agnostic grasp stability metrics yields the highest average success rates of 95.4% for cuboids, 93.1% for cylinders, and62.3% for spheres across wrist position errors between 0 and 7 centimeters . In a second experiment, we show that grasp refinement algorithms trained with contact feedback (contact positions, normals, and forces) perform up to 6.6% better than a baseline that receives no tactile information .…

Metric Multidimensional scaling (MDS) is a classical method for generating meaningful (non-linear) low-dimensional embeddings of high-dimensional data . Despite its ubiquity, our theoreticalunderstanding of MDS remains limited as its objective function is highlynon-convex . In this paper, we prove that minimizing the Kamada-Kawai objectiveis NP-hard and give a provable approximation algorithm for optimizing it, which is a PTAS on low-diameter graphs .…

DeepRare2021 (DR21) is an evolution of a previous version called DeepRare2019 (DR19) based on a common framework . DR21 does not need any training and uses thedefault ImageNet training, 2) is fast even on CPU, 3) is tested on four verydifferent eye-tracking datasets showing that the DR21 is generic and is always in the within the top models on all datasets and metrics .…

International audit standards require the direct assessment of a financial statement’s underlying accounting transactions, referred to as journal entries . We propose a contrastive self-supervisedlearning framework designed to learn audit task invariant accounting datarepresentations to meet this requirement . The framework encompasses deliberateinteracting data augmentation policies that utilize the attributecharacteristics of journal entry data .…

In analyzing and assessing the condition of dynamical systems, it isnecessary to account for nonlinearity . In this work, we propose a reduction scheme relying on the exploitation of anautoencoder as means to infer a latent space from output-only response data .…

The major objective of this work is to study and report the existingontology-driven models for narrative information . The findingsexplicate a comparative view of the narrative models across domains . The findings demonstrate the similarities anddifferences among the elements of the elements in the ontology across domains.…

Rotating machinery is essential to modern life, from power generation totransportation and a host of other industrial applications . accurate remaining useful life (RUL) prediction is essential for maintenance planning and to prevent catastrophic failures . We devise a novelarchitecture and RUL prediction model with uncertainty quantification, termedVisPro, which integrates time-frequency analysis, deep learning imagerecognition, and nonstationary Gaussian process regression .…

This document summarizes the challenges and possible research directions around the shell and its ecosystem, collected during and after the HotOS21Panel on the future of the shell . The goal is to create a snapshot of what anumber of researchers from various disciplines — connected to the shell to varying degrees — think about its future .…

The major objective of this work is to study and report the existingontology-driven models for narrative information . The findingsexplicate a comparative view of the narrative models across domains . The findings demonstrate the similarities anddifferences among the elements of the elements in the ontology across domains.…

International audit standards require the direct assessment of a financial statement’s underlying accounting transactions, referred to as journal entries . We propose a contrastive self-supervisedlearning framework designed to learn audit task invariant accounting datarepresentations to meet this requirement . The framework encompasses deliberateinteracting data augmentation policies that utilize the attributecharacteristics of journal entry data .…

A peridynamics-based finite element method (Peri-FEM) is proposed for the quasi-static fracture analysis . The method is of the consistentcomputational framework with the classical finite elements method . The validity of the proposed method is demonstrated through numerical examples . In this paper, the author proposes a peridynamic element (PE) method to analyze quasional fracture analysis of the structure of a structure .…

Future systems will have to dynamically learn how to act and adapt to face evolving situations with little or no a priori knowledge . Autonomous development includes: learning models ofown capabilities; learning how to . act purposefully towards the achievement of .…

Rotating machinery is essential to modern life, from power generation totransportation and a host of other industrial applications . accurate remaining useful life (RUL) prediction is essential for maintenance planning and to prevent catastrophic failures . We devise a novelarchitecture and RUL prediction model with uncertainty quantification, termedVisPro, which integrates time-frequency analysis, deep learning imagerecognition, and nonstationary Gaussian process regression .…

A peridynamics-based finite element method (Peri-FEM) is proposed for the quasi-static fracture analysis . The method is of the consistentcomputational framework with the classical finite elements method . The validity of the proposed method is demonstrated through numerical examples . In this paper, the author proposes a peridynamic element (PE) method to analyze quasional fracture analysis of the structure of a structure .…

Noise’smain features are expected to strictly depend on the physical platform on which the quantum device is realized, in the form of a distinguishable fingerprint . Noise sources unavoidably affect any quantum technological device . Noise sources are also expected to evolve and change over time .…

A periodic3-dimensional lattice is an infinite set of all integer linear combinations ofbasis vectors in Euclidean 3-space . The resulting spaceLISP consists of infinitely many isometry classes of lattices . In dimension 3,we parameterise this continuous space LISP by six coordinates and introduce newmetrics satisfying the metric axioms and continuity under all perturbations .This…

Question answering (QA) models for reading comprehension exploit unintended dataset biases such as question-contextlexical overlap . This hinders QA models from generalizing to under-representedsamples such as questions with low lexical overlap. To address this problem, we use a synonymreplacement-based approach to augment questions with high lexical overlapping.…

Mean-field games (MFG) were introduced to efficiently analyze approximateNash equilibria in large population settings . We show that entropy regularization provides thenecessary regularity conditions, that are lacking in the standard finite meanfield games . We establish conditions for the existence of a Nashequilibrium in the limiting case as $N$ tends to infinity, and we demonstratethat the Nash equilibrium for the infinite population case is also an$epsilon$-Nash equilibrium .…

We cast a suite of information extraction tasks into a text-to-tripletranslation framework . Instead of solving each task relying on task-specificdatasets and models, we formalize the task as a translation betweentask-specific input text and output triples . We study the zero-shot performance of this framework on open information extraction(OIE2016, NYT, WEB, PENN), relation classification (FewRel and TACRED), andfactual probe (Google-RE and T-REx) The model transfers non-trivially to most tasks and is often competitive with a fully supervised method without the need for any task- specific training .…

Replanners often fail to solve problems involving a large number of misleading plans, i.e., weak plansthat do not lead to strong solutions . Safe-Planner relies on a single-outcomedeterminization to compile a non-deterministic domain to a set of classical domains, and ordering heuristics for ranking the obtained classical domains .…

In analyzing and assessing the condition of dynamical systems, it isnecessary to account for nonlinearity . In this work, we propose a reduction scheme relying on the exploitation of anautoencoder as means to infer a latent space from output-only response data .…

The proposed DVC-P is based on Deep Video Compression (DVC) network, but improves it with perceptual optimization . Adiscriminator network and a mixed loss are employed to help our network tradeoff among distortion, perception and rate . The proposed method can generate videos with higher perceptual quality achieving 12.27% reduction in aperceptual BD-rate equivalent, on average, compared with the baseline DVC, .…

Rotating machinery is essential to modern life, from power generation totransportation and a host of other industrial applications . accurate remaining useful life (RUL) prediction is essential for maintenance planning and to prevent catastrophic failures . We devise a novelarchitecture and RUL prediction model with uncertainty quantification, termedVisPro, which integrates time-frequency analysis, deep learning imagerecognition, and nonstationary Gaussian process regression .…

A peridynamics-based finite element method (Peri-FEM) is proposed for the quasi-static fracture analysis . The method is of the consistentcomputational framework with the classical finite elements method . The validity of the proposed method is demonstrated through numerical examples . In this paper, the author proposes a peridynamic element (PE) method to analyze quasional fracture analysis of the structure of a structure .…

In analyzing and assessing the condition of dynamical systems, it isnecessary to account for nonlinearity . In this work, we propose a reduction scheme relying on the exploitation of anautoencoder as means to infer a latent space from output-only response data .…

Automatic metrics are cheap and reproducible but sometimes poorly correlated with human judgment . In this work, we proposeflexible semiautomatic to automatic summary evaluation metrics, following thePyramid human evaluation method . Semi-automatic Lite2Pyramid retains thereusable human-labeled Summary Content Units (SCUs) for reference(s) but replaces the manual work of judging SCUs’ presence in system summaries with anatural language inference (NLI) model .…

We propose a new approach to temporal inference inspired by the Pearliancausal inference paradigm . Rather than using directed acyclic graphs, we make use of factoredsets, which are sets expressed as Cartesian products . We show that finitefactored sets are powerful tools for inferring temporal relations .…

Metric Multidimensional scaling (MDS) is a classical method for generating meaningful (non-linear) low-dimensional embeddings of high-dimensional data . Despite its ubiquity, our theoreticalunderstanding of MDS remains limited as its objective function is highlynon-convex . In this paper, we prove that minimizing the Kamada-Kawai objectiveis NP-hard and give a provable approximation algorithm for optimizing it, which is a PTAS on low-diameter graphs .…

We present new algorithm for computing union and intersection of alljustifications for a given ontological consequence without first computing theset of all justifications . We show that ourapproach works well in practice for expressive DLs . In particular, the union ofall justifications can be computed much faster than with existingjustification-enumeration approaches .…

Fault injection attacks are known to be vulnerable to physical attackvectors such as fault injection attacks . As of now, these attacks were only used during the inference phase with the intention to cause amisclassification . In this work, we explore a novel attack paradigm by injecting faults during the training phase of a neural network in a way that the resulting network can be attacked during deployment without the necessity of further faulting .…

In this unhinged rant, I lay out my suspicion that a lot of visualizationsare bullshit . I suspect that bullshit charts take up a large fraction of the time and attention of actual visualization producers and consumers, and yet are apparently absent from academic research into visualization design .…

In analyzing and assessing the condition of dynamical systems, it isnecessary to account for nonlinearity . In this work, we propose a reduction scheme relying on the exploitation of anautoencoder as means to infer a latent space from output-only response data .…

High-quality arguments are an essential part of decision-making . We test uncertainty-basedactive learning (AL) methods on two popular argument-strength data sets . Our extensiveempirical evaluation shows that uncertainty based acquisition functions can not surpass the accuracy reached with the random acquisition on these data sets.…

In time-series anomaly detection,detecting anomalous events is more important than detecting individual time-points . The composite F-score ($Fc_1$) is a new metric to overcome these drawbacks . Dynamic scoring functions work much better than static ones for multivariate time series anomaly detection .…

Differential Privacy (DP) models are increasingly used to produce and share synthetic data in a privacy-friendly manner . In this paper, we set out to analyze the impact of DP on these models vis-a-visunderrepresented classes and subgroups of data . We also evaluate the effect of various levels of imbalanceand privacy budgets .…

International audit standards require the direct assessment of a financial statement’s underlying accounting transactions, referred to as journal entries . We propose a contrastive self-supervisedlearning framework designed to learn audit task invariant accounting datarepresentations to meet this requirement . The framework encompasses deliberateinteracting data augmentation policies that utilize the attributecharacteristics of journal entry data .…

We revisit a classical crossword filling puzzle which already appeared in Garey\&Jonhson’s book . We show that the problem remains NP-hard even under very severe structural restrictions . The problem becomes FPT, but the parameter dependence isexponential in $n^2$ We show an algorithm with running time $2^{o(n^1)$ would contradict therandomized ETH .…

We revisit a classical crossword filling puzzle which already appeared in Garey\&Jonhson’s book . We show that the problem remains NP-hard even under very severe structural restrictions . The problem becomes FPT, but the parameter dependence isexponential in $n^2$ We show an algorithm with running time $2^{o(n^1)$ would contradict therandomized ETH .…

A state-space model is proposed,where the hidden state expresses the identity of the current active speaker and the predicted locations of all speakers . The model is implemented as a particle filter . Experiments on a Microsoft rich meeting transcription task show that the proposed joint location tracking and diarisation approach is able to perform comparably with other methods that use location information .…

Weak Supervision (WS) approaches have had widespread success ineasing the bottleneck of labeling training data for machine learning . But proper measurement and analysis of these approaches remain a challenge . We introduce a benchmark platform, \benchmark, for a thoroughand standardized evaluation of WS approaches .…

This document summarizes the challenges and possible research directions around the shell and its ecosystem, collected during and after the HotOS21Panel on the future of the shell . The goal is to create a snapshot of what anumber of researchers from various disciplines — connected to the shell to varying degrees — think about its future .…

We revisit a classical crossword filling puzzle which already appeared in Garey\&Jonhson’s book . We show that the problem remains NP-hard even under very severe structural restrictions . The problem becomes FPT, but the parameter dependence isexponential in $n^2$ We show an algorithm with running time $2^{o(n^1)$ would contradict therandomized ETH .…

A peridynamics-based finite element method (Peri-FEM) is proposed for the quasi-static fracture analysis . The method is of the consistentcomputational framework with the classical finite elements method . The validity of the proposed method is demonstrated through numerical examples . In this paper, the author proposes a peridynamic element (PE) method to analyze quasional fracture analysis of the structure of a structure .…

The widespread deployment of smartphones and location-enabled, networkedin-vehicle devices renders it increasingly feasible to collect streamingtrajectory data of moving objects . The continuous clustering of such data canenable a variety of real-time services, such as identifying representativepaths or common moving trends among objects .…

Graph neural networks (GNNs) have received tremendous attention due to their superiority in learning node representations . However, stacking more convolutional layers significantly decreases the performance of GNNs . We propose a novel orthogonal featuretransformation, named Ortho-GConv, which could generally augment the existing GNN backbones to stabilize the model training and improve the model’s generalization performance .…

Diagnostic imaging plays a critical role in healthcare, serving as afundamental asset for timely diagnosis, disease staging and management . Ultrasoundprobes are becoming increasingly compact and portable, with the market demand for low-cost pocket-sized and (in-body) miniaturized devices expanding . There is a strong trend towards 3D imaging and the use of high-frame-rate imaging schemes; both accompanied by dramatically increasing data rates that pose a heavy burden on the probe-system communication and image reconstruction algorithms .…

We target the problem of synthesizing proofs of semantic equivalence . Our system, S4Eq, achieves 97% proofsuccess on 10,000 pairs of programs while ensuring zero false positives bydesign . We study effectiveness of the system to generate proofs ofincreasing length .…

The paper provides background context, describes theontology creation process, and introduces the Data Sharing Agreements forRestricted and Sensitive Information (DARSI) ontology . DARSI contains four toplevel classes, 20 sub-classes, 33 sub-categories, and 17 simple properties forcategories applicable at various levels .…

Linking research results to grants is an essential prerequisite for aneffective monitoring and evaluation of funding programs . OpenAIRE Research Graph offers a complete dataset of scholarly outputs of from EU Research funding programs. We identify possible improvements and make recommendations on how they can be addressed.…