Searcharxiv⌕ Search

arXiv subjects

Yang Shi

Publications and source records attributed to Yang Shi.

198 records · Page 11Linked to original sources
arXiv2018

Question Type Guided Attention in Visual Question Answering

Visual Question Answering (VQA) requires integration of feature maps with drastically different structures and focus of the correct regions. Image descriptors have structures at multiple spatial scales, while lexical inputs inherently follow a temporal sequence and naturally cluster into semantically different question types. A lot of previous works use complex models to extract feature representations but neglect to use high-level information summary such as question types in learning. In this work, we propose Question Type-guided Attention (QTA). It utilizes the information of question type to dynamically balance between bottom-up and top-down visual features, respectively extracted from ResNet and Faster R-CNN networks. We experiment with multiple VQA architectures with extensive input ablation studies over the TDIUC dataset and show that QTA systematically improves the performance by more than 5% across multiple question type categories such as "Activity Recognition", "Utility" and "Counting" on TDIUC dataset. By adding QTA on the state-of-art model MCB, we achieve 3% improvement for overall accuracy. Finally, we propose a multi-task extension to predict question types which generalizes QTA to applications that lack of question type, with minimal performance loss.

cs.CV↗
arXiv2018

Bounded Connectivity-Preserving Coordination of Networked Euler-Lagrange Systems

This paper derives sufficient conditions for bounded distributed connectivity-preserving coordination of Euler-Lagrange systems with only position measurements and with system uncertainties, respectively. The paper proposes two strategies that suitably scale conventional gradient-based controls to account for the actuation bounds and to reserve sufficient actuation for damping injection. For output feedback control of networked systems with only position measurements, the paper incorporates a first-order filter to estimate velocities and to inject damping for stability. For networks of uncertain systems, the paper augments conventional linear filter-based adaptive compensation with damping injection to maintain the local connectivity of the network. Analyses based on monotonically decreasing Lyapunov-like functions and Barbalat's lemma lead to sufficient conditions for bounded local connectivity-preserving coordination of Euler-Lagrange networks under the two strategies. The sufficient conditions clarify the interrelationships among the bounded actuations, initial system velocities and initial inter-system distances. Simulation results validate these conditions.

math.OC↗
arXiv2018

Connectivity-Preserving Consensus of Multi-Agent Systems with Bounded Actuation

This paper investigates the impact of bounded actuation on the connectivity-preserving consensus of two classes of multi-agent systems, with kinematic agents and with Euler- Lagrange agents. The investigation establishes that: (1) there exists a class of gradient-based controls which drive kinematic multi-agent systems to connectivity-preserving consensus even if they saturate; (2) actuator saturation restricts the initial states from which Euler-Lagrange multi-agent systems can be synchronized while preserving their local connectivity; (3) Euler-Lagrange multi-agent systems with unbounded actuation can achieve connectivity-preserving consensus without velocity measurements or exact system dynamics; and (4) a proposed indirect coupling control strategy drives Euler-Lagrange multi-agent systems with limited actuation and starting from rest to connectivitypreserving consensus without requiring velocity measurements and including in the presence of uncertain dynamics and timevarying delays.

math.OC↗
arXiv2018

Connectivity-Preserving Coordination Control of Multi-Agent Systems with Time-Varying Delays

This paper presents a distributed position synchronization strategy that also preserves the initial communication links for single-integrator multi-agent systems with time-varying delays. The strategy employs a coordinating proportional control derived from a specific type of potential energy, augmented with damping injected through a dynamic filter. The injected damping maintains all agents within the communication distances of their neighbours, and asymptotically stabilizes the multi-agent system, in the presence of time delays. Regarding the closed-loop single-integrator multi-agent system as a double-integrator system suggests an extension of the proposed strategy to connectivity-preserving coordination of Euler-Lagrange networks with time-varying delays. Lyapunov stability analysis and simulation results validate the two designs.

math.OC↗
arXiv2018

Globally Stable Output Feedback Synchronization of Teleoperation with Time-Varying Delays

This paper presents a globally stable teleoperation control strategy for systems with time-varying delays that eliminates the need for velocity measurements through novel augmented Immersion and Invariance velocity observers. The new observers simplify a recent constructive Immersion and Invariance velocity observer to achieve globally convergent velocity estimation with only $n+2$ states, where $n$ is the number of degrees of freedom of the master and slave robots. They introduce dynamic scaling factors to accelerate the speed of convergence of the velocity estimates and, thus, to limit the energy generated by the velocity estimation errors and to guarantee sufficient estimate-based damping injection to dissipate the energy generated by the time-varying delays. The paper shows that Proportional plus damping control with the simplified and augmented Immersion and Invariance-based velocity observers can synchronize the free master and slave motions in the presence of time-varying delays without using velocity measurements. Numerical results illustrate the estimation performance of the new observers and the stability of a simulated two degrees-of-freedom nonlinear teleoperation system with time-varying delays under the proposed output feedback Proportional plus damping control.

eess.SY↗
arXiv2017

Geometric description of discrete power function associated with the sixth Painlevé equation

In this paper, we consider the discrete power function associated with the sixth Painlevé equation. This function is a special solution of the so-called cross-ratio equation with a similarity constraint. We show in this paper that this system is embedded in a cubic lattice with $\widetilde{W}(3A_1^{(1)})$ symmetry. By constructing the action of $\widetilde{W}(3A_1^{(1)})$ as a subgroup of $\widetilde{W}(D_4^{(1)})$, i.e., the symmetry group of P$_{\rm VI}$, we show how to relate $\widetilde{W}(D_4^{(1)})$ to the symmetry group of the lattice. Moreover, by using translations in $\widetilde{W}(3A_1^{(1)})$, we explain the odd-even structure appearing in previously known explicit formulas in terms of the $τ$ function.

math-ph↗
arXiv2017

Decentralized Charging Control of Electric Vehicles in Residential Distribution Networks

Electric vehicle (EV) charging can negatively impact electric distribution networks by exceeding equipment thermal ratings and causing voltages to drop below standard ranges. In this paper, we develop a decentralized EV charging control scheme to achieve "valley-filling" (i.e., flattening demand profile during overnight charging), meanwhile meeting heterogeneous individual charging requirements and satisfying distribution network constraints. The formulated problem is an optimization problem with a non-separable objective function and strongly coupled inequality constraints. We propose a novel shrunken primal-dual subgradient (SPDS) algorithm to support the decentralized control scheme, derive conditions guaranteeing its convergence, and verify its efficacy and convergence with a representative distribution network model.

math.OC↗
arXiv2017

Robustness Analysis of Asynchronous Sampled-Data Multi-Agent Networks With Time-Varying Delays

In this paper, we study the simultaneous stability problem of a finite number of locally inter-connected linear subsystems under practical constraints, including asynchronous and aperiodic sampling, time-varying delays, and measurement errors. We establish a new Lyapunov-based stability result for such a decentralized system. This system has a particular simple structure of interconnections, but it captures some key characteristics of a large class of intermediate models derived from the consensus analysis of multi-agent systems. The stability result is applicable to the estimation of the maximum allowable inter-sampling periods and time delays based on individual dynamics and coupling structures in the scenarios of consensus control via asynchronous sampling of relative states and asynchronous broadcasting of self-sampled states respectively. The asynchrony of aperiodic sampling and the existence of measurement errors allow the utilization of some kinds of quantizing devices, such as Logarithmic quantizers, in the process of data sampling, and allow the introduction of a period of dwell time after each update of state measurement to eliminate the Zeno behavior of events in event-based control. The extension in the case with input saturations and input delays is also discussed.

math.OC↗
arXiv2017

Compact Tensor Pooling for Visual Question Answering

Performing high level cognitive tasks requires the integration of feature maps with drastically different structure. In Visual Question Answering (VQA) image descriptors have spatial structures, while lexical inputs inherently follow a temporal sequence. The recently proposed Multimodal Compact Bilinear pooling (MCB) forms the outer products, via count-sketch approximation, of the visual and textual representation at each spatial location. While this procedure preserves spatial information locally, outer-products are taken independently for each fiber of the activation tensor, and therefore do not include spatial context. In this work, we introduce multi-dimensional sketch ({MD-sketch}), a novel extension of count-sketch to tensors. Using this new formulation, we propose Multimodal Compact Tensor Pooling (MCT) to fully exploit the global spatial context during bilinear pooling operations. Contrarily to MCB, our approach preserves spatial context by directly convolving the MD-sketch from the visual tensor features with the text vector feature using higher order FFT. Furthermore we apply MCT incrementally at each step of the question embedding and accumulate the multi-modal vectors with a second LSTM layer before the final answer is chosen.

cs.CV↗
arXiv2017

Leak Event Identification in Water Systems Using High Order CRF

Today, detection of anomalous events in civil infrastructures (e.g. water pipe breaks and leaks) is time consuming and often takes hours or days. Pipe breakage as one of the most frequent types of failure of water networks often causes community disruptions ranging from temporary interruptions in services to extended loss of business and relocation of residents. In this project, we design and implement a two-phase approach for leak event identification, which leverages dynamic data from multiple information sources including IoT sensing data (pressure values and/or flow rates), geophysical data (water systems), and human inputs (tweets posted on Twitter). In the approach, a high order Conditional Random Field (CRF) is constructed that enforces predictions based on IoT observations consistent with human inputs to improve the performance of event identifications. Considering the physical water network as a graph, a CRF model is built and learned by the Structured Support Vector Machine (SSVM) using node features such as water pressure and flow rate. After that, we built the high order CRF system by enforcing twitter leakage detection information. An optimal inference algorithm is proposed for the adapted high order CRF model. Experimental results show the effectiveness of our system.

cs.LG↗
arXiv2016

Tensor Contractions with Extended BLAS Kernels on CPU and GPU

Tensor contractions constitute a key computational ingredient of numerical multi-linear algebra. However, as the order and dimension of tensors grow, the time and space complexities of tensor-based computations grow quickly. Existing approaches for tensor contractions typically involves explicit copy and transpose operations. In this paper, we propose and evaluate a new BLAS-like primitive STRIDEDBATCHEDGEMM that is capable of performing a wide range of tensor contractions on CPU and GPU efficiently. Through systematic benchmarking, we demonstrate the advantages of our approach over conventional approaches. Concretely, we implement the Tucker decomposition and show that using our kernels yields 100x speedup as compared to the implementation using existing state-of-the-art libraries.

cs.DC↗
arXiv2016

Reflection groups and discrete integrable systems

We present a method of constructing discrete integrable systems with crystallographic reflection group (Weyl) symmetries, thus clarifying the relationship between different discrete integrable systems in terms of their symmetry groups. Discrete integrable systems are associated with space-filling polytopes arise from the geometric representation of the Weyl groups in the $n$-dimensional real Euclidean space $\mathbb{R}^n$. The "multi-dimensional consistency" property of the discrete integrable system is shown to be inherited from the combinatorial properties of the polytope; while the dynamics of the system is described by the affine translations of the polytopes on the weight lattices of the Weyl groups. The connections between some well-known discrete systems such as the multi-dimensional consistent systems of quad-equations \cite{abs:03} and discrete Painlevé equations \cite{sak:01} are obtained via the geometric constraints that relate the polytope of one symmetry group to that of another symmetry group, a procedure which we call geometric reduction.

nlin.SI↗
arXiv2016

Tensor vs Matrix Methods: Robust Tensor Decomposition under Block Sparse Perturbations

Robust tensor CP decomposition involves decomposing a tensor into low rank and sparse components. We propose a novel non-convex iterative algorithm with guaranteed recovery. It alternates between low-rank CP decomposition through gradient ascent (a variant of the tensor power method), and hard thresholding of the residual. We prove convergence to the globally optimal solution under natural incoherence conditions on the low rank component, and bounded level of sparse perturbations. We compare our method with natural baselines which apply robust matrix PCA either to the {\em flattened} tensor, or to the matrix slices of the tensor. Our method can provably handle a far greater level of perturbation when the sparse tensor is block-structured. This naturally occurs in many applications such as the activity detection task in videos. Our experiments validate these findings. Thus, we establish that tensor methods can tolerate a higher level of gross corruptions compared to matrix methods.

cs.LG↗
arXiv2016

Receding Horizon Consensus of General Linear Multi-agent Systems with Input Constraints: An Inverse Optimality Approach

It is desirable but challenging to fulfill system constraints and reach optimal performance in consensus protocol design for practical multi-agent systems (MASs). This paper investigates the optimal consensus problem for general linear MASs subject to control input constraints. Two classes of MASs including subsystems with semi-stable and unstable dynamics are considered. For both classes of MASs without input constraints, the results on designing optimal consensus protocols are first developed by inverse optimality approach. Utilizing the optimal consensus protocols, the receding horizon control (RHC)-based consensus strategies are designed for these two classes of MASs with input constraints. The conditions for assigning the cost functions distributively are derived, based on which the distributed RHC-based consensus frameworks are formulated. Next, the feasibility and consensus properties of the closed-loop systems are analyzed. It is shown that 1) the optimal performance indices under the inverse optimal consensus protocols are coupled with the network topologies and the system matrices of subsystems, but they are different for MASs with semi-stable and unstable subsystems; 2) the unstable modes of subsystems impose more stringent requirements for the parameter design; 3) the designed RHC-based consensus strategies can make the control input constraints fulfilled and ensure consensus for the closed-loop systems in both cases. But for MASs with semi-stable subsystems, the {\em convergent consensus} can be reached. Finally, two examples are provided to verify the effectiveness of the proposed results.

math.OC↗
arXiv2015

A systematic approach to reductions of type-Q ABS equations

We present a class of reductions of Möbius type for the lattice equations known as Q1, Q2, and Q3 from the ABS list. The deautonomised form of one particular reduction of Q3 is shown to exist on the $A_1^{(1)}$ surface which belongs to the multiplicative type of rational surfaces in Sakai's classification of Painlevé systems. Using the growth of degrees of iterates, all other mappings that result from the class of reductions considered here are shown to be linearisable. Any possible linearisations are calculated explicitly by constructing a birational transformation defined by invariant curves in the blown up space of initial values for each reduction.

nlin.SI↗
arXiv2014

Geometric Reductions of ABS equations on an $n$-cube to discrete Painlevé systems

In this paper, we show how to relate $n$-dimensional cubes on which ABS equations hold to the symmetry groups of discrete Painlevé equations. We here focus on the reduction from the 4-dimensional cube to the $q$-discrete third Painlevé equation, which is a dynamical system on a rational surface of type $A_5^{(1)}$ with the extended affine Weyl group $\widetilde{\mathcal W}\bigl((A_2+A_1)^{(1)}\bigr)$. We provide general theorems to show that this reduction also extends to other discrete Painlevé equations at least of type A.

nlin.SI↗