Transfer Learning for Human & AI

Spatial and channel re-calibration have become powerful concepts in computervision . Their ability to capture long-range dependencies is especially useful for those networks that extract local features, such as CNNs . We propose a set ofre-Calibration blocks that extend Squeeze and Excitation blocks and that can beadded to any network for 3D point cloud analysis that builds a globaldescriptor by hierarchically combining features from multiple localneighborhoods .…

The framework is designed to enable users to design experiments by adeclarative PyYAML configuration file . It helps researchers complete therecently proposed ML Code Completeness Checklist . All the source code,configurations, log files and the trained model weights are publicly available at https://://://github.com/yoshitomo-matsubara/torchdistill…

Open-world learning is a problem where an autonomous agent detects thingssthat it does not know and learns them over time from a non-stationary andnever-ending stream of data . In an open-world environment, the training data and objective criteria are never available at once .…

Different learning rates may lead to differences in optimization and generalization . We find that the large learning rate phase is closely related to the separability of data . We demonstrate that this interpretation can be applied to real settings on MNIST and CIFAR10 datasets with the fact that theoptimal performance is often found in this largelearning rate phase .…

This letter proposes a new approach for online detection of low-qualitysynchrophasor data under both normal and event conditions . The proposed approach does not require any offline studyand it is more effective to detect low quality data with apparentlyindistinguishable profiles .…

In this article, we present new random walk methods to solve flow and transport problems in unsaturated/saturated porous media . The numerical schemes are based on global random walkalgorithms (GRW) which approximate the solution by moving large numbers ofcomputational particles on regular lattices according to specific random walkrules .…

Privacy,like security, is a non-functional property, yet most software design tools are focused on functional aspects . Recent regulations, such as the European General Data Protection Regulation, put stringent constraints on the handling of personal data . In this paper, we provide an explicit algorithm and a proof-of-concept implementation to transform DFDs into PA-DFDs .…

In underwater acoustics, shallow water environments act as modal dispersivewaveguides when considering low-frequency sources . In this context, propagatingsignals can be described as a sum of few modal components, each of them propagating according to its own wavenumber . Estimating these wavenumbers is of interest to understand the propagating environment as well as the emittingssource .…

This paper develops an efficient multi-agent deep reinforcement learning algorithm for cooperative controls in powergrids . The decentralized learning scheme and highsample efficiency also make it viable to large-scale power grids . The developed PowerNetoutperforms a conventional model-based control, as well as several state-of-the-art MARL algorithms, the paper says .…

Genome assembly is a fundamental problem in Bioinformatics, requiring toreconstructing a source genome from an assembly graph built from a set of reads . The goal is to find what is definitely present in all solutions, orwhat is safe . The long-standing open problem offinding all the safe parts of the solutions was recently solved by a majortheoretical result [RECOMB’16].…

This paper studies an edge intelligence-based IoT network in which a set ofedge servers learn a shared model using federated learning (FL) based on thedatasets uploaded from a multi-technology-supported IoT network . The datauploading performance of IoT network and the computational capacity of edgeservers are entangled with each other in influencing the FL model training process .…

existing methods either do not estimate clothing at all or model clothdeformations with simple geometric priors . This leads to noticeable artifacts in their constructions, such as baked-in wrinkles and implausible deformations that defy gravity . We propose a person-specific, learning-based method thatintegrates a finite element-based simulation layer into the training process .…

Machine learning (ML) is included in Self-organizing Networks (SONs) that arekey drivers for enhancing the Operations, Administration, and Maintenance (OAM) activities . Standalone (SA) system is one of the 5Gcommunication tracks that transforms 4G networking to next-generation technology that is based on mobile applications .…

Energy forecasting has a vital role to play in smart grid (SG) systems . Traditional statistical and machine learning-based forecasting methods are extensively investigated in terms of their applicability to energyforecasting . Hybrid methods such as CNN-LSTM are also highly effective to deal with long sequences in energy data .…

MOBA games pose grand challenges to AI systems such as multi-agent, enormous state-action space, complex action control, etc. OpenAI’s Dota AI limits theplay to a pool of only 17 heroes. As a result, full MOBA . games without . restrictions are far from being mastered by any existing AI system.…

Homomorphic encryption is a very useful gradient protection technique used inprivacy preserving federated learning . But existing encrypted federatedlearning systems need a trusted third party to generate and distribute keypairs to connected participants . The proposed method avoidssencryption and decryption of the entire model .…

Unsupervised domain adaptation (UDA) studies the problem of transferring models trained on one labeled source domain to another unlabeled target domain . CycleEmotionGAN++ is a novel end-to-end cycle-consistent adversarial model . Cygany proposes a dynamic emotional semantic consistency loss to preserve the emotionlabels of the source images .…

This paper presents redundancy resolution and disturbance rejection optimization in Hybrid Cable-Driven Robots (HCDRs) Compared to existing approaches, this paper provides the first solution (TOAUJ-basedmethod) for HCDRs that can solve the redundancy resolution problem as well as the disturbance rejection problem .…

Cytoarchitecture is a basic principle of the microstructural organization of the brain . The new workflow does not require preceding 3D-reconstruction of sections, and is robust against histological artefacts . It is based on a DeepConvolutional Neural Network (CNN) which is trained on a pair of sectionimages with annotations, with a large number of un-annotated sections inbetween .…

We construct fast, structure-preserving iterations for computing the signdecomposition of a unitary matrix $A$ with no eigenvalues equal to $\pm i$ We use our iterations to construct a stable spectraldivide-and-conquer algorithm . Our iterationsrely on a recently discovered formula for the best (in the minimax sense)unimodular rational approximant of the scalar function $S =\operatorname{sign}(A) = A(A(A^2)^{-1/2) times a matrix $N = (A#2) + (A/N) The iterations converge significantly faster than Pad\’e iterations.…

Continuum bone remodelling is an important tool for predicting the effects of mechanical stimuli on bone density evolution . While the modelling of onlycancellous bone is considered in many studies based on continuum boneremodelling, this work presents an approach of modelling also cortical bone and the interaction of both bone types .…

This paper deals with the compensation of nonlinearities in dynamical systems . The compensation approach is formulated for static and dynamical contexts . The core idea is to rewrite the model as analgebraic polynomial whose roots are potential compensation inputs .…

This paper proposes a novel multilevel inverter for automotive applications . The topology consists of a modular DCDC converter and a tap selector . The DC-DC converter is capable of self-balancing its modules and thus does not require large capacitors which yields a high power density .…

Interest in non-linear impulsive systems (NIS) has been growing due to the impact in application problems such as disease treatments (diabetes, HIV,influenza, among many others) This paper studies the asymptotic stability of controlequilibrium orbits for NSI, based on the underlying discrete time system .…

Two families of block graphs withcliques of fixed size, the block-path and block-starlike graphs, are introduced . Cases A and B of classification for both families were considered, as well as the behavior of the algebraic connectivity for particular cases ofblock-path graphs .…

Contract scheduling is a general technique that allows to design a system with interruptible capabilities, given an algorithm that is not necessarilyinterruptible . Previous work on this topic has largely assumed that the interruption is a worst-case deadline that is unknown to the scheduler .…

We present algorithms that extend the path-based hierarchical drawing framework and give experimental results . Our algorithms run in $O(km)$ time,where $k$ is the number of paths and $m$ is number of edges of the graph . We also provide somecomparison to a well known .…

This paper presents A3D, an adaptive 3D network that can infer at a widerange of computational constraints with one-time training . Instead of training multiple models in a grid-search manner, it generates good configurations by trading off between network width and spatio-temporal resolution .…

Paper proposed a new method to locate the source of forced oscillation that involves resonance with natural oscillation modes . The new method is based on comparing the oscillation mode shape of the forced oscillations with that of the natural oscillations .…

The acceleration race of digital computing technologies seems to be steeringtoward impasses — technological, economical and environmental — a condition that has spurred research efforts in alternative, “neuromorphic” (brain-like)computing technologies . The idea of exploiting nonlinear physical phenomena “directly” for non-digital computing has been explored under names like “unconventional computing”, “natural computing”,”physical computing”, or “in-materio computing” I stake out how a general concept of “computing” can bedeveloped which comprises digital, neuromorphic, unconventional and possible future paradigms .…

We show that for $k, q = O(1)$ there is a deterministic, polynomial time algorithm to approximately count thenumber of satisfying assignments . For the special case of properly $q$-coloring $k$-uniform hypergraphs, the term $Delta^{7$ improves the previously best known $\Delta^{60}$ for deterministic algorithms [Moitra,J.ACM,…

This paper presents a new deep learning-based model for fault detection in offshore wind turbines . The proposed model uses the nonlinear relationships among multiple sensor variables and thetemporal dependency of each sensor on others that considerably increases the performance of the model .…

This work studies the stability of multi-class queuing networks under a class of centralized or decentralized model data-independent (MDI) control policies . Controlactions include routing, sequencing, and holding . For stabilizable single-class networks, we show that a centralized, stabilizing MDI control policy exists .…

This paper presents a new approach to visual zero-shot slot filling . The approach extends previous approaches by reformulating the slot filling task asQuestion Answering . The multi-task approach facilitates the incorporation of a large number of refinements and transfer learning across similar tasks .…

Artificial neural networks are often very complex and too deep for a human to understand . In this paper, we try to find a way to explain anetwork and extract a human-readable equation that describes the model . In some cases, laws of physics, for example, the pattern can be described by relatively simplemathematical expressions .…

This paper compares networks of foreign fighters who joined the Islamic Stateof Iraq and Syria (ISIS) from Europe and the Arabian Peninsula . In places where recruitment needs to be hidden from legal scrutiny, recruitment networks are more secretive . In areas where recruitment could occur more freely, recruitment is more hierarchical .…

Machine learning has the potential to aid in mitigating the human effects of climate change . Previous applications of machine learning to tackle the humaneffects in climate change include informing individuals of their carbon footprint and strategies to reduce it .…

We give a constant factor polynomial time pseudo-approximation algorithm formin-sum clustering with or without outliers . The algorithm is allowed to exclude an arbitrarily small constant fraction of the points . Our results apply to instances of points in real space, as well as to points in a metricspace, where the number of clusters, and also the dimension if relevant, isarbitrary (part of the input, not an absolute constant) The approximation guarantee growswith $\frac{1-\eps) n’$ points .…

Hindsight Network Credit Assignment (HNCA) works by assigning credit to eachneuron’s stochastic output based on how it influences the output of its immediate children in the network . We prove that HNCA provides unbiased gradient estimates while reducing variance compared to the REINFORCE estimator .…

This paper investigates the combination of two model predictive control concepts . The effectiveness of the proposed method is validatedvia numerical simulations at different scenarios on a three-level neutral pointclamped permanent magnet synchronous generator wind turbine system and comparedto otherlong-horizon model predictive controls methods .…

Process mining offers a promising way to help analyze resource grouping, authors say . Organizations need to acquire an accurate and timelyunderstanding of human resource grouping . Authors propose a novel framework built upon a richer definition of organizationalmodels coupling resource grouping with process execution knowledge .…

We present a novel technique for solving the problem of safe control for ageneral class of nonlinear, control-affine systems subject to parametric modeluncertainty . Invoking Lyapunov analysis and the notion of fixed-time stability(FxTS) We introduce a parameter adaptation law which guarantees convergence of the estimates of unknown parameters in the system dynamics to their true values within a fixed time independent of the initial parameter estimation error .…

Neuronal circuits internally regulate electrical signaling via a host of homeostatic mechanisms . Two prominent mechanisms, synaptic scaling andstructural plasticity, are believed to maintain average activity within an operating range . However, both mechanisms operate on relatively slow timescales and thus face fundamental limits due to delays .…

This paper aims to develop a model-free loop-shaping method in controldesign . The core idea is to convert the model matching problem to atrajectory tracking problem . The proposed method is validated through numerical simulation on athird order plant . This method does not require the model of the controlled plant; hence it provides better performance of loop-Shaping controllesign.…

Health-Focused Optimal Power Flow (HF-OPF) proposed to take into account equipment health in operational and physical constraints . The paper addresses theneed for understanding the relationship between health condition index and theoperational constraints in OPF problems . The results show that health conditioninflicts high cost of generation and can lead to infeasibility even with lesscritical faults .…

The closed-loop performance of model predictive controllers (MPCs) issensitive to the choice of prediction models, controller formulation, and tuning parameters . MPC tuning is typically donemanually to satisfy (probabilistic) constraints . In this work, we demonstrate ageneral approach for automating the tuning of MPC under uncertainty .…

The Quantum Alternating Operator Ansatz (QAOA+) framework has recently gained attention due to its ability to solve discrete optimization problems on noisyintermediate-scale quantum (NISQ) devices . We design a technique in this framework totackle a few problems over maximal matchings in graphs .…

A fundamental challenge in learning to control an unknown dynamical system is to reduce model uncertainty by making measurements while maintaining safety . In our framework, the state of the system is required to stay within a given safety region under the (possibly repeated) action of all systems that are consistent with the information gathered so far .…

In this paper, we investigate the global convergence of gradient-based policyoptimization methods for quadratic optimal control of discrete-time Markovian jump linear systems (MJLS) Despite the non-convexity of the resultant problem, we are still able to identify several useful properties such ascoercivity, gradient dominance, and almost smoothness .…