Transfer Learning for Human & AI

Smart contracts are dependent on oracle systems for their adoption andusability . We collect ChainLink usage data on the . Ethereum network using a modified . Ethereum client and running a full node . We present our findings and insights surrounding the .…

UDDSKETCH is a recent algorithm for accurate tracking of quantiles in datastreams . It is derived from the DDSKetCH algorithm . We show how to compress and fuse data streams (or datasets) by using UDDSkETCH data summaries that are fused into a new summary related to the union of the union .…

Existing multi-camera SLAM systems assume synchronized shutters for all cameras, which is often not the case in practice . In this work, we propose ageneralized SLAM formulation which accounts for asynchronoussensor observations . Our framework integrates a continuous-time motion model torelate information across asynchronous multi-frames during tracking, localmapping, and loop closing .…

This paper considers the inverse problem of recovering both the unknown,spatially-dependent conductivity $a(x)$ and the nonlinear reaction term $f(u)$ in a reaction-diffusion equation from overposed data . We prove bothuniqueness results and the convergence of iteration schemes designed to recover these coefficients .…

Deep learning has demonstrated the important roles in improving the systemperformance and reducing computational complexity for $5$G-and-beyond networks . We propose a new channel estimation method with the assistance of deep learning in order to support the least squares estimation, which is alow-cost method but having relatively high channel estimation errors .…

Many existing deep learning models are vulnerable to adversarial examplesthat are imperceptible to humans . We propose to search for deep neural architectures thatare robust to five types of well-known adversarial attacks using amulti-objective evolutionary algorithm . To reduce the computational cost, anormalized error rate of a randomly chosen attack is calculated as the robustness for each newly generated neural architecture at each generation .…

This paper proposes a service restoration model for unbalanced distributionsystems and inverter-dominated microgrids . After extreme events, the damaged distribution systems can be sectionalized into several isolated MGs to restore critical loads and tripped distributed generations by black-start DGs . The high penetration of inverter based DGs reduce the system inertia, which results in low-inertia issues and large frequency fluctuation during therestoration .…

graph-based SLAM system for 2D LiDAR sensor in natural environments . System takes advantage of robustweighting scheme, sliding-windowed optimization, fast scan-matcher and parallelcomputing . Simulated and experimental results confirm the feasibility and efficiency in the overall design of the proposed system .…

In this paper, we report the results obtained from the acceleration of multi-binary64-type multiple precision matrix multiplication with AVX2 . We implement SIMDized EFT functions, whichsimultaneously compute with four binary64 numbers on x86_64 computingenvironment . We also develop SIMDize DD, TD, and QDadditions and multiplications .…

Predictive processing has become an influential framework in cognitivesciences . This framework turns the traditional view of perception upside down,claiming that the main flow of information processing is realized in a top-downhierarchical manner . In the field of cognitive robotics there is no clear-cut distinction on whichschemes have been implemented and under which assumptions .…

A network is a composition of many communities, i.e., sets of nodes and edges with stronger relationships, with distinct and overlapping properties . Community detection is crucial for various reasons, such as serving as afunctional unit of a network that captures local interactions among nodes .…

The daily activities performed by a disabled or elderly person can bemonitored by a smart environment, and the acquired data can be used to learn apredictive model of user behavior . We use state-of-the-art deep neuralnetwork-based techniques to learn predictive human activity models in the local, centralized, and federated learning settings .…

In this paper, we propose a method to discover multiscale chemical reactionsgoverned by the law of mass action from data . We use one matrix torepresent the stoichiometric coefficients for both the reactants and products in a system without catalysis reactions .…

At construction sites, multiple construction machines will usually work together and side-by-side, causing large dynamic occlusions in the cameras’ view . We propose a motionsegmentation method to efficiently extract static parts from crowded dynamicscenes to enable robust tracking of camera ego-motion .…

Cellular-Vehicle-to-Everything (C-V2X) communication has become an integral component of Intelligent Transportation Systems . It is unfeasible to physically conduct a V2X communication experiment to test the performance of this technology . Networkemulators can provide more realistic and repeatable results for testingvehicular communication .…

Artificial intelligence (AI)-driven zero-touch network slicing is envisaged as a promising cutting-edge technology to harness the full potential ofheterogeneous 5G and beyond 5G (B5G) communication systems . We present a novel actor-Critic-based network slicing approach called, prioritized twin delayeddistributional deep deterministic policy gradient (D-TD3)}.…

In projection-based model order reduction, a reduced-order approximation of the original full-order system is obtained by projecting it onto a reducedsubspace that contains its dominant characteristics . As the order of the reduced model is increased, the deviation in the satisfaction of the optimality conditions decays .…

EU needs an effective exit strategy from the crisis, with a special attention to SMEs that represent the 99% of the enterprises active in the Europeanproduction system . Innovation appears to be a key factor to reinvigorate the EU industrial system .…

Developers from different teams or organizations, co-located or distributed, making changes to the same source code files or areas, through pull requeststhat are active in the same time period, is an essential part of developing software systems . ConE (Concurrent Edit Detector) proactively detects concurrentedits to help mitigate problems caused by them .…

Modern self-driving autonomy systems heavily rely on deep learning . As aconsequence, their performance is influenced significantly by the quality and richness of the training data . It is of key importance to have a mechanism to identify “whatto label” Active learning approaches identify examples to label, but theirinterestingness is tied to a fixed model performing a particular task .…

The typical ad-hoc routing protocols aim atfinding the shortest path, lead to significant performance degradation because of the 3-dimension highly-dynamic nature of UAV networks . This paper proposes OPAR, anoptimized predictive and adaptive routing protocol, to face this challenging problem .…

In this paper we investigate the problem of the sudden disconnection of anedge in a discrete-time multi-agent consensus network . If the graph remainsstrongly connected, the system still achieves consensus, but ingeneral, unless the information exchange between each pair of agents issymmetric, the agents’ states converge to a drifted value of the original consensus value .…

We implement a user-extensible ad hoc connection between the Lean proofassistant and the computer algebra system Mathematica . The proof assistant library serves as adatabase of mathematical knowledge that the CAS can display and explore . In the other direction, we import and process Leandeclarations from within the system .…

The problem of state estimation for a system of coupled hyperbolic PDEs andODEs with Lipschitz nonlinearities with boundary measurements is considered . Aninfinite dimensional observer with a linear boundary injection term is used to solve the state estimation problem . The observer is designed to achieve global exponentialstability of estimation error with respect to a suitable norm .…

End-to-end models have achieved impressive results on the task of automaticspeech recognition (ASR) For low-resource ASR tasks, however, labeled data canhardly satisfy the demand of the demand . We fuse a pre-trained acoustic encoder and apre-trained linguistic encoder into an ASR model .…

In this work we propose an efficient parallelization of multiple-precisionTaylor series method with variable stepsize and fixed order . With 256 CPU cores in Nestum cluster, Sofia,Bulgaria, we succeed to obtain a correct reference solution in the rather longtime interval – [0,11000].…

5G cellular networks are being deployed all over the world and thisarchitecture supports ultra-dense network (UDN) deployment . Small cells have avery important role in providing 5G connectivity to the end users . Exponentialincreases in devices, data and network demands make it mandatory for the service providers to manage handovers better .…

We formulate a free probabilistic analog of the Wasserstein manifold on $\mathbb{R}^d$ We use it to study smooth non-commutative transport ofmeasure . The points of the free Wasserststein manifold$\mathscr{W}(\mathbb {R}^{*d})$ are smooth tracial non-comutative functions $V$ with quadratic growth at $infty .…

High-performance computing (HPC) clusters are widely used in-house at scientific and academic research institutions . This Ten Simple Rules guide is aimed at helping you identify ways to improve your utilisation of HPC, avoiding common pitfalls that can negatively impact other users .…

This paper proposes a trilevel neural architecture search (NAS) method forefficient single image super-resolution (SR) Unlike current NAS methods, we exploit a sortedsparsestmax activation to let the three-level neural structures contribute sparsely . Our NAS algorithm provides SR models that are significantly lighter in terms of the number of parameters and FLOPS withPSNR value comparable to the current state-of-the-art .…

LaneRCNN is a graph-centric motion forecasting model . It uses a specially designed graph encoder to learn a local lane graphrepresentation per actor . We also develop an interaction module which permits efficient message passing among local graph representations within a shared global lanegraph .…

Principal Component Analysis (PCA) is well known for its capability ofdimension reduction and data compression . The vectorization of images makes some correlation constraints of neighboring pixels and spatial information lost . We used small neighborhoods of each pixel to form compounded pixels and use a tensorial version of PCA, called TPCA, to compress and reconstruct acompounded image of compounded pixels .…

A Log-Structured Merge (LSM) Tree is a natural fit for a lifecycle-aware storage engine due to its high write throughput and level-oriented structure . LASER is almost 5x faster than Postgres (a purerow-store) and two orders of magnitude faster than MonetDB for real-time data analytics workloads .…

Alavi, Malde, Schwenk, and Erd{\”o}s conjectured that for trees theindependence polynomial is unimodal . More accurately, considering trees with up to 20 vertices, we showedthat their independence polynomials are log-concave .…

The positdata format is proposed as a drop-in replacement for IEEE 754 float format . The average error in LuKa withSoftPosit is an order of magnitude lower than with SoftFloat . We then present the integration of the hardware implementation of a posit adder and multiplier in a RISC-V open-source platform .…

A big, diverse and balanced training data is the key to the success of deepneural network training . Separable Batch Normalization(SepBN) module with a Cross-protocol Network Training (CNT) strategy for robustfacial landmark localization . SepBN considerssthat the samples of a training dataset may belong to different sub-domains .…

Steganography methods tend to embed more and more secretbits in the cover images . Most of these methods are designed to embed secretinformation in such a way that the change in the visual quality of the resulting stego image is not detectable .…

This work considers the following extension of the tree-depth problem: for agiven input graph $G$ and integers $k$ and $b$ find a rooted forest $F$ ofheight at most $k and $width$ of height $K$ The problem is NP-hard and obtaining a polynomial-time additive$2b$-approximation algorithm .…

In this letter, we present an intelligent reflecting surface (IRS) selection strategy for multiple IRSs aided multiuser multiple-input single-output (MISO) systems . A two stage user-IRS assignment algorithm is proposed . The first stage of the proposed algorithm employs awell-known Gale Shapley matching designed for the stable marriage problem .…

Pruning has become a promising technique used to compress and accelerateneural networks . The prevailing filter channel pruning method removes the entire filterchannel . In this study, we proposed kernel cluster pruning (KCP) to prune denselabeling networks . We developed a clustering technique to identify the leastrepresentational kernels in each layer .…

A set of colored graphs are compatible, if for every color $i$ the number ofvertices of the same color is the same in every graph . A simultaneous embedding of $k$ compatibly colored graphs, each with $n$ vertices, can always be computed with a sublinear number of bends per edge .…

A new trend with the wide-spread of deep neuralnetwork applications is on-device AI . It is to process neural networks on mobile devices or edge/IoT devices instead of cloud servers . We propose NNStreamer, a software system that handles neural networks asfilters of stream pipelines, applying the stream processing paradigm to deepneural network applications .…

Audio-visual speech enhancement system is regarded to be one of promisingsolutions for isolating and enhancing speech of desired speaker . Conventional methods focus on predicting clean speech spectrum via a naive convolutionneural network based encoder-decoder architecture . Paper proposes attentionalaudio-visual multi-layer feature fusion model, in which soft thresholdattention unit are applied on feature mapping at every layer of decoder .…

SIMPLE is a combination of a fast feedforward network and a hyper-parameteroptimizer . SIMPLE runs in 3 seconds while the newer algorithms take 6 hours toterminate . Since it runs so fast, it is more amenable to being tuned by ouroptimizer.…

A common approach to the automatic detection of mispronunciation works byrecognizing the phonemes produced by a student and comparing it to the expected pronunciation of a native speaker . This approach makes two simplifyingassumptions: a) phoneme can be recognized from speech with high accuracy, b)there is a single correct way for a sentence to be pronounced .…

A significant amount of disease transmission occurs through human-to-human or social contacts . Understanding who interacts with whom intime and space is essential for disease transmission modelling, prediction, and assessment of prevention strategies in different environments and special settings . This paper presents a methodology for generating contacts matrices by using high fidelity simulations which mimic actual workflow and movements of individuals in time and space .…

Multichannel blind source separation aims to recover latent sources from multichannel mixture without priors . A state-of-art blind sourceseparation method called independent low-rank matrix analysis (ILRMA) unifiedindependent vector analysis (IVA) and nonnegative matrix factorization (NMF) However, speech spectra modeled by NMF may not find a compact representation .…

A floating robotic emulation platform for a virtual demonstration of satellite motion in space is presented . The robotic platform is characterized by its friction-less, levitating, yet planar motion over a hyper-smooth surface . The entire design, including 3D printingCAD model and different testbed elements, is provided in an open-sourcerepository and a test campaign is used to showcase its capabilities and illustrate its operations .…