Dynamic backdoor attacks against federated learning

Federated Learning (FL) is a new machine learning framework, which enablesmillions of participants to collaboratively train machine learning model without compromising data privacy and security . FL does not guarantee that all clients are honest by design, which makes it vulnerable to adversarial attack naturally .…

New Confocal Hyperbola based Ellipse Fitting with Applications to Estimating Parameters of Mechanical Pipes from Point Clouds

The confocal hyperbola is an excellent approximation to the true geometric distance of points to ellipses . The proposed method outperformed the gold standardgeometric method of Ahn . It was observed that the proposed ellipse fitting method achieved almost identical results (and in some cases better) than the gold-standardgeometrificalic and Sampson methods .…

Model Driven Synthesis for Programming Tutors

Wepropose to investigate how we can overcome this problem by using programsynthesis . We generate correct solutions that closely match astudent programs . We give feedback based on the results of a program . We hope to find a way to give constructive feedback to a student working on a beginner’s exercise .…

FedRec Federated Learning of Universal Receivers over Fading Channels

Wireless communications are often subject to fading conditions . We propose a collaborative neuralnetwork-based symbol detection mechanism for downlink fading channels called FedRec . FedRec is based on the maximum a-posteriori probability (MAP)detector . The performance of the resulting FedRec receiver is shown to approach the MAP performance in diverse channel conditions, while inducing asubstantially reduced communication overhead in its training procedure compared to training in a centralized fashion .…

On Relaxed Filtered Krylov Subspace Method for Non Symmetric Eigenvalue Problems

Relaxed filtered Krylov subspace method for computing theeigenvalues with the largest real parts and the corresponding eigenvectors of non-symmetric matrices . We give the convergenceanalysis of the complex Chebyshev polynomial, which plays a significant role inthe polynnomine acceleration technique . In addition, numerical experiments are carried out to show the robustness of the relaxed filtered filtered Kryov subspachod and its great superiority over some state-of-the art iteration methods .…

Quantum communication capacity transition of complex quantum networks

Quantum network is the key to enable distributed quantum information processing . As the single-link communication rate decays exponentially with the distance, to enable reliable end-to-end quantum communication, the number of nodes need to grow with the network scale . For highly connected networks, weidentify a threshold transition in the capacity as the density of network nodes increases .…

Learning of Structurally Unambiguous Probabilistic Grammars

The problem of identifying a probabilistic context free grammar has two aspects: the grammar’s topology and the weights of the grammar . We provide a query learning algorithm for structurally unambiguousprobabilisticcontext-free grammars (SUPCFG) We show that SUWCFG can be represented using co-linear multiplicity tree automata (CMTA) and provide apolynomial learning algorithm that learns CMTAs.…

Does spontaneous motion lead to intuitive Body Machine Interfaces A fitness study of different body segments for wearable telerobotics

Human-Robot Interfaces (HRIs) represent a crucial component in teleroboticsystems . Data-driven approaches select a set of body segments and transform their motion into commands for the robot based on their spontaneous motion patterns . Despite being a versatile and genericmethod, there is no scientific evidence that implementing an interface based on spontaneous motion maximizes its effectiveness.…

Quantitative towers in finite difference calculus approximating the continuum

Multivector fields and differential forms at the continuum level have commutative associative products, a third composition product between them and various operators like $d$ and $*$ which are used to describe many nonlinear problems . The point of this paper is to construct consistent direct and inverse systems of finite dimensional approximations to these structures and to calculate combinatorially how these finite dimensionalmodels differ from their continuum idealizations .…

Recoverable Abortable and Adaptive Mutual Exclusion with Sublogarithmic RMR Complexity

We present the first recoverable mutual exclusion (RME) algorithm that issimultaneously abortable, adaptive to point contention, and with sublogarithmicRMR complexity . We obtain this algorithm byusing the Fetch-And-Add (FAA) primitive, unlike prior work on RME that uses FAS/SWAP . Our key building blocksare: A $D$-process abortable RME algorithm, for $D \leq W$ with $O(1)$ passagecomplexity and $ O(1+F)$ super-passage complexity .…

Echo CGC A Communication Efficient Byzantine tolerant Distributed Machine Learning Algorithm in Single Hop Radio Network

In Echo-CGC, if a worker “agrees” with a combination of priorgradients, it will broadcast the “echo message” instead of the its raw local gradient . The echo message contains a vector of coefficients (of size at most$n$) and the ratio of the magnitude between two gradients (a float) In the radio network, each worker is able to overhearprevious gradients that were transmitted to the parameter server .…

Designing Human Robot Coexistence Space

When the human-robot interactions become ubiquitous, the environments surrounding these interactions will have significant impact on the safety and comfort of the human and the effectiveness and efficiency of the robot . The key enabling technique is a motion planner that can evaluate hundreds of similar motion planning problems .…

Debiasing Convolutional Neural Networks via Meta Orthogonalization

The Meta Orthogonalization method encourages CNN representations of different concepts to be orthogonal to one another inactivation space while maintaining strong downstream task performance . It significantly mitigates model bias and is competitive against current debiasing methods, authors say . The method is based on existing work on debiases word embeddings and model interpretability, they say .…

Full Attitude Intelligent Controller Design of a Heliquad under Complete Failure of an Actuator

A neural network-based control allocation is designed to provide complete control authority even under a complete loss of one actuator . Heliquad is a multi-coptersimilar to Quadcopter, with four actuators diagonally symmetric from the center . Each actuator has two control inputs; the first input changes thepropeller blades collective pitch (also called variable pitch) and the otherinput changes the rotation speed.…

Crowdsharing Wireless Energy Services

We propose a novel self-sustained ecosystem for energy sharing in the IoTenvironment . We leverage energy harvesting, wireless power transfer, and crowd-sourcing that facilitate the development of an energy crowdsharingframework to charge IoT devices . The ubiquity of IoT devices, coupled with the potential ability for sharing energy, provides new and exciting opportunitiesto crowdsource wireless energy .…

Performance Analysis and Optimization of a UAV Enabled Two Way Relaying Network under FSMH NC and PNC Schemes

Unmanned aerial vehicles (UAVs) have played an important role in wireless communications due to the advantages such as highly controllable mobility in three-dimensional (3D) space, swift deployment, line-of-sight (LoS)aerial-ground links, and so on . In this paper, we consider a UAV-enabledtwo-way relaying system where the UAV relay assists the information exchangebetween two ground users (GUs) under three different schemes, i.e.,…

Safety Synthesis Sans Specification

We define the problem of learning a transducer from a target language$U$ containing possibly conflicting transducers . We argue that this is a natural question in many situations in hardware and software verification . We devise a learning algorithm for this problem and show that its time and query complexity is polynomial with respect to the rank of the target language, its incompatibility measure, and the maximal length of a given counterexample .…

Respiratory Distress Detection from Telephone Speech using Acoustic and Prosodic Features

Speech samples are collected from de-identified telemedicine phonecalls from a healthcare provider in Bangladesh . The recordings include conversational speechsamples of patients talking to doctors showing mild or severe respiratorydistress or asthma symptoms . We hypothesize that respiratory distress may alterspeech features such as voice quality, speaking pattern, loudness, and speech-pause duration .…

Speech enhancement guided by contextual articulatory information

Previous studies have confirmed the effectiveness of leveraging articulatoryinformation to attain improved speech enhancement (SE) performance . In this study, we propose an SE system that incorporates contextual articulatory information . Such information is obtained using broad phone class (BPC) end-to-end automatic speech recognition(ASR) Meanwhile, two training strategies are developed to train the SE systembased on the BPC-based ASR: multitask-learning and deep-feature trainingstrategies .…

Co optimisation and Settlement of Power Gas Coupled System in Day ahead Market under Multiple Uncertainties

The interdependency of power systems and natural gas systems is beingreinforced by the emerging power-to-gas facilities (PtGs), and the existing gas-fired generators . To jointly improve the efficiency and security underdiverse uncertainties from renewable energy resources and load demands, it is essential to co-optimise these two energy systems for day-ahead marketclearance .…

Nonlinear Cooperative Control of Double Drone Bar Transportation System

Two drones are required to lift the load together, which brings significant challenges to the control problem . A nonlinear cooperative control method is proposed, with both rigorousstability analysis and experimental results demonstrating its greatperformance . Without the need to distinguish the identities between the leaderand the follower, the proposed method successfully realizes effective controlfor the two drones separately, mainly owning to the deep analysis for the system dynamics and the elaborate design for the control law .…

A Distributed Privacy Preserving Learning Dynamics in General Social Networks

In this paper, we study a distributed privacy-preserving learning problem ingeneral social networks . Each agent is allowed to interact with its peersthrough multi-hop communications but with its privacy preserved . We propose a four-staged distributed social learning algorithm . In a nutshell, our algorithm proceeds iteratively, and in every round, each agent randomly perturbs its adoption for privacy preservation .…