Transfer Learning for Human & AI

The stylus is equipped with two vibration actuators at the ends, which create a tactile sensation of up and down movement along the stylus . It is also embedded with a DC motor, which is used to create bidirectional rotational torque about the long axis of the pen .…

The ConferencingSpeech 2021 challenge is proposed to stimulate research on multi-channel speech enhancement for video conferencing . The challenge consists of two separate tasks: Task 1 and Task 2 are non-real-time tasks . The baseline system which is a complex ratio mask based neuralnetwork and its experimental results are also presented .…

Self-supervised learning of speech representations has been a very active research area . Most work is focused on a single domain such as read audiobooks for which there exist large quantities of labeled and unlabeled data . Ourexperiments show that using target domain data during pre-training leads to large performance improvements across a variety of setups .…

Current RF sensors lack the computational resources to support practical, in-situtraining for intelligent signal classification . Deepdelay Loop Reservoir Computing (DLR) delivers reductions in form factor, hardware complexity and latency compared to the State-ofthe- Art (SoA) neural nets . DLR enables mobile edge platforms to authenticate andthen track emitters with fast SEI retraining .…

This paper presents an efficient and compact MATLAB code forthree-dimensional stress-based sensitivity analysis . The 146 lines code includes the finite element analysis and p-norm stress sensitivity analysis based on the adjoint method . The complete program for sensitivity analysis is given in the Appendix and is intended for educational purposes only .…

Archive of Formal Proofs (AFP) is an online repository of formal proofs . It serves as a central location for publishing, discovering, and viewing libraries of proofs . Long-term users of the website are generally satisfied with the AFP but that there are anumber of areas, such as navigation, search and script browsing, that need improvement .…

A compiler’s optimizer operates over abstract syntax trees (ASTs), applying rewrite rules to replace subtrees of the AST with more efficient ones . Especially on large source repositories, even simply finding opportunities for a rewrite can be expensive . In this paper, we leverage the need to repeatedly find rewrites, and explore options for making the search faster through indexing and incrementalview maintenance (IVM) We consider bolt-on approaches that makeuse of embedded IVM systems like DBToaster, as well as two new approaches:Label-indexing and TreeToaster .…

In this work, we consider the target detection problem in a sensingarchitecture where the radar is aided by a reconfigurable intelligent surface(RIS) The RIS is modeled as an array of sub-wavelength small reflectiveelements capable of imposing a tunable phase shift to the impinging waves and,ultimately, of providing the radar with an additional echo of the target .…

The codebook design for code-domain non-orthogonal multiple access (CD-NOMA) can be considered as a constellation design for multi-user multi-dimensionalmodulation (MU-MDM) This paper proposes an autoencoder (AE)-basedconstellation design for MU-MDm with the objective of achieving a comparablebit error rate (BER) performance to single-user multidimensional modulation (SU-M) It has been demonstrated that the proposed designachieves a single-users BER bound with only 0.2dB loss, equivalently outperforming the existing CD-NomA designs, while maintaining their overloadingfactor .…

DataPrep.EDA is a novel task-centric EDA system in Python . It allows data scientists to declaratively specify a widerange of EDA tasks in different granularity with a single function call . It is open-sourced as an EDA component of DataPrep:https://://://www.sfu-db/dataprep.com/svsvsv/dataPrep/dataaprep .…

The CSO Classifier takes as input the metadata associated with a research paper (title, abstract, keywords) and returns a selection of research concepts drawn from the CSO Ontology . The approach was evaluated on a gold standard of manually annotated articles .…

A number of studies compare SPARQL federation engines across different performance metrics, including queryruntime, result set completeness and correctness, number of sources selected and number of requests sent . The effect of estimated cardinality errors on the overall query runtime performance must be measured .…

In nominal mission scenarios, geostationary satellites perform end-of-lifeorbit maneuvers to reach suitable disposal orbits . This research investigates the long-term orbitevolution of decommissioned geostations . The model accounts for all the relevant harmonics of the gravity field and solar radiation pressure and third-body perturbations caused by the Moon and the Sun .…

Biological and cultural evolution show a trend towards increasing hierarchical organization, in which entities at one level combine cooperatively to form a new entity at a higher level of organization . John Stewart has proposed a theoreticalframework called Management Theory, which attempts to explain all of the major cooperative transitions in biological and cultural .…

We propose using an adversarial autoencoder (AAE) to replace generativeadversarial network (GAN) in the private aggregation of teacher ensembles . The AAE architecture allows us to obtain good synthetic speech leveraging upona discriminative training of latent vectors . Such synthetic speech is used tobuild a privacy-preserving classifier when non-sensitive data is notsufficiently available in the public domain .…

This paper presents new parallel algorithms for generating Euclidean minimumspanning trees and spatial clustering hierarchies . Our approach is based on generating a well-separated pair decomposition followed by using Kruskal’s minimum spanning tree algorithm and bichromatic closest paircomputations . We show that our algorithms are theoretically efficient: they have work (number ofoperations) matching their sequential counterparts, and polylogarithmic depth(parallel time) We implement our algorithms and propose a memory optimization that requiresonly a subset of well-Sepated pairs to be computed and materialized, leading to savings in both space (up to 10x) and time .…

The Fisher information matrix plays the role of aRiemannian metric in this framework . This tool helps us obtain Cram\'{e}r-Raolower bound (CRLB) This chapter summarizes the recent results which extend the framework to more general Cram\{e}.r-Rao inequalities . We apply Eguchi’s theory to a generalized form of Czsisz\’ar $f$-divergence to obtain a .general…

Basic Linear Algebra Subprograms (BLAS) is a core library in scientificcomputing and machine learning . FT-BLAS offers high reliability and high performance — faster than Intel MKL, OpenBLAS, and BLISby up to 3.50%, 22.14% and 21.70%, respectively, for routines spanning all three levels of BLAS we benchmarked, even under hundreds of errors injected perminute .…

Tempo is a leaderless SMR protocol that, in comparison to prior solutions, achievessuperior throughput and offers predictable performance even in contended workloads . To achieve these benefits, Tempo timestamps each application commandand executes it only after the timestamp becomes stable, i.e.,…

Event sourced systems are increasing in popularity because they are reliable,flexible, and scalable . In this article, we point a microscope at a softwarearchitecture pattern that is rapidly gaining popularity in industry, but has not received as much attention from the scientific community .…

In this note, we present constructive bijections from Dyck and Motzkinmeanders with catastrophes to Dyck paths avoiding some patterns . As abyproduct, we deduce correspondences from Dycks excursions to restricted Dyck path paths .…

We present a new approach to proving non-termination of non-deterministic integer programs . Our technique is rather simple but efficient . It relies on apurely syntactic reversal of the program’s transition system followed by aconstraint-based invariant synthesis with constraints coming from both the original and reversed transition system .…

Backdoor can be injected into a third-party model and has strongstealthiness in normal situation, thus has been widely discussed . Backdoor attack on deep neural network has been concerned a lot and there comeslots of researches about attack and defense around backdoor in DNN .…

We propose an automatic parameter selection strategy for variational imagesuper-resolution of blurred and down-sampled images . By exploiting particularproperties of the operators describing the problem in the frequency domain, our strategy selects the optimal parameter as the one optimising a suitableresidual whiteness measure .…

Many mission-critical applications of machine learning (ML) in the real-world require a quality assurance (QA) process before the decisions or predictions of an ML model can be deployed . QA4ML users have to view a non-trivial amount of data and perform many input actions to correct errors made by the MLmodel .…

We proposed a platform for selecting and configuring automatically an appropriate cloud environment that meets a set of consumer and provider requirements . We based on the DesignScience Research Methodology to conduct this work . The platformbased on the principles of dynamic software product lines (SPL), Agent-orientedsoftware engineering, and the MAPE-k reference model .…

Understanding students’ programming misconceptions is critical . We propose catalogs ofantipatterns for two programming languages: C and Python . We analyzed the codes of 166 CS1 engineering students when they were codingsolutions to programming exercises . We catalog 41 CS1antipAtterns from 95 cataloged misconceptions in C and .…

A key challenge in Imitation Learning (IL) is that optimal state actionsdemonstrations are difficult for the teacher to provide . As an alternative to state action demonstrations, the teacher can provide corrective feedback such as their preferences or rewards . Instead we propose that in orderto learn from a diversity of scenarios we need to learn a variety of feedback that is weakly correlated with the teacher’s true cost function .…

Study tested feasibility of using ‘off the shelf’ information extraction algorithms to identify medical concepts from Italian clinical notes . We used MetaMap to map medical concepts to the Unified Medical Language System (UMLS) Results in EXP1 showed that theItalian UMLS Metathesaurus sources covered 91% of the medical terms of the’Disorders’ semantic group, as found in the studied dataset .…

The Review Bomb is a phenomenon consisting of a massiveattack by groups of Internet users on a website that displays users’ review on products . The Bomb occurs suddenly and for a shorttime, because in this way it leverages the notorious problem of cold-start: ifreviews are submitted by a lot of fresh new accounts, it makes hard to justifypreventative measures .…

The first chapter is a critical review and a case study in eBusiness, withspecial attention to the digital currencies resource and its possibilities . The last chapter s purpose was todevelop a measuring and modelling framework, an instrument of IBSQ for theSouth African banking sector .…

1.3 million household fires, 3,400 civilian deaths, and 23 billion dollars indamage, a fire department is called to respond every 24 seconds . Many firefighters are injured during search and rescue operations due to hiddendangers . Fire-retardant water runoff pollution can threaten human health .…

Parametric 3D models have enabled a wide variety of tasks in computer graphics and vision, such as modeling human bodies, faces, and hands . However, construction of these parametric models is often tedious, as it requires manual tweaking . To this end, we propose Neural ParametricModels (NPMs), a novel, learned alternative to traditional, parametric 3d models, which does not require hand-crafted, object-specific constraints .…

Displaying tactile feedback through a touchscreen via electrovibration has potential applications in mobile devices, consumer electronics, home appliances, and automotive industry though our knowledge and understanding on the underlying contact mechanics is very limited . The main cause of the increasein friction during full slip is due to an increase in real contact area and the .reduction…

Mobile traffic is dominated by content services (e.g., video), which typically use recommendation systems . NFR increase the network performance at the cost of being unfair towards certain contents when compared to the standard recommendations . Fair-NFR can achieve high network gains (similar to non-fair) with little unfairness, authors say .…

This paper studies the problem of decentralized continuum deformation coordination of multi-agent systems aided by cooperative localization . We treatagents as particles inside a triangular continuum (deformable body) in a2-Dmotion space . Followers distributed inside theleading tri-angle acquire continuum deformed in a decentralized fashion byintegrating cooperative localization and local communication .…

Discretization of the uniform norm of functions from a given finitedimensional subspace of continuous functions is studied . We pay special attention to the case of trigonometric polynomials with frequencies from anarbitrary finite set with fixed cardinality . We prove a general result, which connects the upper bound on the number of sampling points in the discretization .…

Successful forms of human-AI partnership have rarely been demonstrated in real-world settings . Lumilo smartglasses help teachers help their students in AI-supported classrooms . Smartglasses provide real-time analytics about students’ learning, metacognition, and behavior . Results from a field study conducted in K-12 classrooms indicate that students learn more when teachers and AI tutors work together during class .…

This work investigates the asymptotic behaviour of the gradient approximationmethod called the generalized simplex gradient (GSG) This method has an errorbound that at first glance seems to tend to infinity as the number of sample points increases . But with some careful construction, we show that this is not the case .…

The Internet of Value (IOV) with its distributed ledger technology (DLT)underpinning has created new forms of lending markets . Lending protocols are gaining traction, holding an aggregate liquidity supply of over $40 billion . We discuss the persisting reliance of DeFi lending on the traditional financial system, and conclude with the outlook of the lending market in the IOV era .…

In this paper, we pose the current state-of-the-art voice conversion (VC)systems as two-encoder-one-decoder models . After comparing these models, we combine the best features and propose Assem-VC . This paper also introduces the GTAfinetuning in VC, which significantly improves the quality and the speakersimilarity of the outputs .…

This paper proposes a deep learning framework for classification of BBC television programmes using audio . The audio is firstly transformed intospectrograms, which are fed into a pre-trained convolutional Neural Network . We achieve an average classification accuracy of 93.7% over 14-fold cross validation .…

Visual Question Answering systems target answering open-ended textual questions given input images . They are a testbed for learning high-levelreasoning with a primary use in HCI, for instance assistance for the visually impaired . Recent research has shown that state-of-the-art models tend toproduce answers exploiting biases and shortcuts in the training data .…

We present a novel reformulation of balanced truncation, a classical model reduction method . The principal innovation comes through theuse of system response data that has been either measured or computed, withoutreference to any prescribed realization of the original model .…

The Council of Europe’s Committee of Ministers adopted the terms of reference for the Ad Hoc Committee on Artificial Intelligence (CAHAI) The CAHAI is charged with examining the feasibility and potential elements of a legal framework for the design, development, and deployment of AI systems .…

Previous studies have shown that training data instance selectionbased on nearest neighborhood (NN) information can lead to better performance in cross project defect prediction . Locality Sensitive Hashing (LSH) can be affective as exact methods . LSH and GIS favor recall more than precision, however, in environments where precision is considered more important NN-Filter should be considered .…

A Convolutional Neural Network (CNN) model is able to accuratelydiscriminate between 4 different phases of neurological status in a non-Electroencephalogram(EEG) dataset recorded in an experiment in which subjects are exposed to physical, cognitive and emotional stress . We show that CNN models outperformstradic classification methods such as SVM, and RF, in comparison to other methods, in robustnessto noise by 97.46% accuracy on a noisy dataset .…

In many real datasets such as social media streams and cyber data sources,graphs change over time through a graph update stream of edge insertions anddeletions . Detecting critical patterns in such dynamic graphs plays animportant role in various application domains such as fraud detection, cybersecurity, and recommendation systems for social networks .…

In this work, we give the first large-scalestudy of performance predictors by analyzing 31 techniques ranging fromlearning curve extrapolation, to weight-sharing, to supervised learning, to “zero-cost” proxies . We test a number of correlation- and rank-based measures in a variety of settings, as well as the ability of each technique to speed up predictor-based NAS frameworks .…