arXiv subjects

Dan Wang

Publications and source records attributed to Dan Wang.

At least 127 records · Page 7Linked to original sources

arXiv2021

On-edge Multi-task Transfer Learning: Model and Practice with Data-driven Task Allocation

On edge devices, data scarcity occurs as a common problem where transfer learning serves as a widely-suggested remedy. Nevertheless, transfer learning imposes a heavy computation burden to resource-constrained edge devices. Existing task allocation works usually assume all submitted tasks are equally important, leading to inefficient resource allocation at a task level when directly applied in Multi-task Transfer Learning (MTL). To address these issues, we first reveal that it is crucial to measure the impact of tasks on overall decision performance improvement and quantify \emph{task importance}. We then show that task allocation with task importance for MTL (TATIM) is a variant of the NP-complete Knapsack problem, where the complicated computation to solve this problem needs to be conducted repeatedly under varying contexts. To solve TATIM with high computational efficiency, we propose a Data-driven Cooperative Task Allocation (DCTA) approach. Finally, we evaluate the performance of DCTA by not only a trace-driven simulation, but also a new comprehensive real-world AIOps case study that bridges model and practice via a new architecture and main components design within the AIOps system. Extensive experiments show that our DCTA reduces 3.24 times of processing time, and saves 48.4\% energy consumption compared with the state-of-the-art when solving TATIM.

cs.DC↗

arXiv2021

Nonadiabatic evolution and thermodynamics of a time-dependent open quantum system

We investigate the dynamic evolution and thermodynamic process of a driven quantum system immersed in a finite-temperature heat bath. A Born-Markovian quantum master equation is formally derived for the time-dependent system with discrete energy levels. This quantum master equation can be applied to situations with a broad range of driving speeds and bath temperatures and thus be used to study the finite-time quantum thermodynamics even when nonadiabatic transition and dissipation coexist. The dissipative Landau-Zener model is analyzed as an example. The population evolution and transition probability of the model reveal the importance of the competition between driving and dissipation beyond the adiabatic regime. Moreover, local maximums of irreversible entropy production occur at intermediate sweep velocity and finite temperature, which the low-dissipation model cannot describe.

quant-ph↗

arXiv2021

Multi-view 3D Reconstruction with Transformer

Deep CNN-based methods have so far achieved the state of the art results in multi-view 3D object reconstruction. Despite the considerable progress, the two core modules of these methods - multi-view feature extraction and fusion, are usually investigated separately, and the object relations in different views are rarely explored. In this paper, inspired by the recent great success in self-attention-based Transformer models, we reformulate the multi-view 3D reconstruction as a sequence-to-sequence prediction problem and propose a new framework named 3D Volume Transformer (VolT) for such a task. Unlike previous CNN-based methods using a separate design, we unify the feature extraction and view fusion in a single Transformer network. A natural advantage of our design lies in the exploration of view-to-view relationships using self-attention among multiple unordered inputs. On ShapeNet - a large-scale 3D reconstruction benchmark dataset, our method achieves a new state-of-the-art accuracy in multi-view reconstruction with fewer parameters ($70\%$ less) than other CNN-based methods. Experimental results also suggest the strong scaling capability of our method. Our code will be made publicly available.

cs.CV↗

arXiv2021

Painlevé IV, $σ-$Form and the Deformed Hermite Unitary Ensembles

We study the Hankel determinant generated by a deformed Hermite weight with one jump $w(z,t,γ)=e^{-z^2+tz}|z-t|^γ(A+Bθ(z-t))$, where $A\geq 0$, $A+B\geq 0$, $t\in\textbf{R}$, $γ>-1$ and $z\in\textbf{R}$. By using the ladder operators for the corresponding monic orthogonal polynomials, and their relative compatibility conditions, we obtain a series of difference and differential equations to describe the relations among $α_n$, $β_n$, $R_n(t)$ and $r_n(t)$. Especially, we find that the auxiliary quantities $R_n(t)$ and $r_n(t)$ satisfy the coupled Riccati equations, and $R_n(t)$ satisfies a particular Painlevé IV equation. Based on above results, we show that $σ_n(t)$ and $\hatσ_n(t)$, two quantities related to the Hankel determinant and $R_n(t)$, satisfy the continuous and discrete $σ-$form equations, respectively. In the end, we also discuss the large $n$ asymptotic behavior of $R_n(t)$, which produce the expansion of the logarithmic of the Hankel determinant and the asymptotic of the second order differential equation of the monic orthogonal polynomials.

math-ph↗

arXiv2020

Is Image Encoding Beneficial for Deep Learning in Finance? An Analysis of Image Encoding Methods for the Application of Convolutional Neural Networks in Finance

In 2012, SEC mandated all corporate filings for any company doing business in US be entered into the Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system. In this work we are investigating ways to analyze the data available through EDGAR database. This may serve portfolio managers (pension funds, mutual funds, insurance, hedge funds) to get automated insights into companies they invest in, to better manage their portfolios. The analysis is based on Artificial Neural Networks applied to the data.} In particular, one of the most popular machine learning methods, the Convolutional Neural Network (CNN) architecture, originally developed to interpret and classify images, is now being used to interpret financial data. This work investigates the best way to input data collected from the SEC filings into a CNN architecture. We incorporate accounting principles and mathematical methods into the design of three image encoding methods. Specifically, two methods are derived from accounting principles (Sequential Arrangement, Category Chunk Arrangement) and one is using a purely mathematical technique (Hilbert Vector Arrangement). In this work we analyze fundamental financial data as well as financial ratio data and study companies from the financial, healthcare and IT sectors in the United States. We find that using imaging techniques to input data for CNN works better for financial ratio data but is not significantly better than simply using the 1D input directly for fundamental data. We do not find the Hilbert Vector Arrangement technique to be significantly better than other imaging techniques.

q-fin.ST↗

arXiv2020

On Spectral Properties of Signed Laplacians with Connections to Eventual Positivity

Signed graphs have appeared in a broad variety of applications, ranging from social networks to biological networks, from distributed control and computation to power systems. In this paper, we investigate spectral properties of signed Laplacians for undirected signed graphs. We find conditions on the negative weights under which a signed Laplacian is positive semidefinite via the Kron reduction and multiport network theory. For signed Laplacians that are indefinite, we characterize their inertias with the same framework. Furthermore, we build connections between signed Laplacians, generalized M-matrices, and eventually exponentially positive matrices.

eess.SY↗

arXiv2020

Organizing Virtual Conferences through Mirrors: The ACM e-Energy 2020 Experience

The emergence of the world-wide COVID-19 pandemic has forced academic conferences to be held entirely in a virtual manner. While prior studies have advocated the merits of virtual conferences in terms of energy and cost savings, organizers are increasingly facing the prospect of planning and executing them systematically, in order to deliver a rich conference-attending-experience for all participants. Starting from March 2020, tens of conferences have been held virtually. Past conferences have revealed numerous challenges, from budget planning, to selecting the supporting virtual platforms. Among these, two special challenges were identified: 1) how to deliver talks to geo-distributed attendees and 2) how to stimulate social interactions among attendees. These are the two important goals of an academic conference. In this paper, we advocate a mirror program approach for academic conferences. More specifically, the conference program is executed in multiple parallel (mirrored) programs, so that each mirror program can fit a different time zone. This can effectively address the first challenge.

cs.CY↗

arXiv2020

Terahertz Strong-Field Physics in Light-Emitting Diodes for Terahertz Detection and Imaging

Intense terahertz (THz) electromagnetic fields have been utilized to reveal a variety of extremely nonlinear optical effects in many materials through nonperturbative driving of elementary and collective excitations. However, such nonlinear photoresponses have not yet been discovered in light-emitting diodes (LEDs), letting alone employing them as fast, cost effective,compact, and room-temperature-operating THz detectors and cameras. Here we report ubiquitously available LEDs exhibited gigantic and fast photovoltaic signals with excellent signal-to-noise ratios when being illuminated by THz field strengths >50 kV/cm. We also successfully demonstrated THz-LED detectors and camera prototypes. These unorthodox THz detectors exhibited high responsivities (>1 kV/W) with response time shorter than those of pyroelectric detectors by four orders of magnitude. The detection mechanism was attributed to THz-field-induced nonlinear impact ionization and Schottky contact. These findings not only help deepen our understanding of strong THz field-matter interactions but also greatly contribute to the applications of strong-field THz diagnosis.

physics.optics↗

arXiv2020

Controlling a Networked SIS Model via a Single Input over Undirected Graphs

This paper formulates and studies the problem of controlling a networked SIS model using a single input in which the network structure is described by a connected undirected graph. A necessary and sufficient condition on the values of curing and infection rates for the healthy state to be exponentially stable is obtained via the analysis of signed Laplacians when the control input is the curing budget of a single agent. In the case when the healthy state is stabilizable, an explicit expression for the minimum curing budget is provided. The utility of the algorithm is demonstrated using a simulation over a network of cities in the northeastern United States.

math.OC↗

arXiv2020

Application of Deep Neural Networks to assess corporate Credit Rating

Recent literature implements machine learning techniques to assess corporate credit rating based on financial statement reports. In this work, we analyze the performance of four neural network architectures (MLP, CNN, CNN2D, LSTM) in predicting corporate credit rating as issued by Standard and Poor's. We analyze companies from the energy, financial and healthcare sectors in US. The goal of the analysis is to improve application of machine learning algorithms to credit assessment. To this end, we focus on three questions. First, we investigate if the algorithms perform better when using a selected subset of features, or if it is better to allow the algorithms to select features themselves. Second, is the temporal aspect inherent in financial data important for the results obtained by a machine learning algorithm? Third, is there a particular neural network architecture that consistently outperforms others with respect to input features, sectors and holdout set? We create several case studies to answer these questions and analyze the results using ANOVA and multiple comparison testing procedure.

q-fin.RM↗

arXiv2020

CHAIN: Concept-harmonized Hierarchical Inference Interpretation of Deep Convolutional Neural Networks

With the great success of networks, it witnesses the increasing demand for the interpretation of the internal network mechanism, especially for the net decision-making logic. To tackle the challenge, the Concept-harmonized HierArchical INference (CHAIN) is proposed to interpret the net decision-making process. For net-decisions being interpreted, the proposed method presents the CHAIN interpretation in which the net decision can be hierarchically deduced into visual concepts from high to low semantic levels. To achieve it, we propose three models sequentially, i.e., the concept harmonizing model, the hierarchical inference model, and the concept-harmonized hierarchical inference model. Firstly, in the concept harmonizing model, visual concepts from high to low semantic-levels are aligned with net-units from deep to shallow layers. Secondly, in the hierarchical inference model, the concept in a deep layer is disassembled into units in shallow layers. Finally, in the concept-harmonized hierarchical inference model, a deep-layer concept is inferred from its shallow-layer concepts. After several rounds, the concept-harmonized hierarchical inference is conducted backward from the highest semantic level to the lowest semantic level. Finally, net decision-making is explained as a form of concept-harmonized hierarchical inference, which is comparable to human decision-making. Meanwhile, the net layer structure for feature learning can be explained based on the hierarchical visual concepts. In quantitative and qualitative experiments, we demonstrate the effectiveness of CHAIN at the instance and class levels.

cs.CV↗

arXiv2019

CHIP: Channel-wise Disentangled Interpretation of Deep Convolutional Neural Networks

With the widespread applications of deep convolutional neural networks (DCNNs), it becomes increasingly important for DCNNs not only to make accurate predictions but also to explain how they make their decisions. In this work, we propose a CHannel-wise disentangled InterPretation (CHIP) model to give the visual interpretation to the predictions of DCNNs. The proposed model distills the class-discriminative importance of channels in networks by utilizing the sparse regularization. Here, we first introduce the network perturbation technique to learn the model. The proposed model is capable to not only distill the global perspective knowledge from networks but also present the class-discriminative visual interpretation for specific predictions of networks. It is noteworthy that the proposed model is able to interpret different layers of networks without re-training. By combining the distilled interpretation knowledge in different layers, we further propose the Refined CHIP visual interpretation that is both high-resolution and class-discriminative. Experimental results on the standard dataset demonstrate that the proposed model provides promising visual interpretation for the predictions of networks in image classification task compared with existing visual interpretation methods. Besides, the proposed method outperforms related approaches in the application of ILSVRC 2015 weakly-supervised localization task.

cs.CV↗

arXiv2019

Towards the same line of liquid-liquid phase transition of dense hydrogen from various theoretical predictions

For a long time, there have been huge discrepancies between different models and experiments concerning the liquid-liquid phase transition (LLPT) in dense hydrogen. In this work, we present the results of extensive calculations of the LLPT in dense hydrogen using the most expensive first-principle path-integral molecular dynamics simulations available. The nonlocal density functional rVV10 and hybrid functional PBE0 are used to improve the description of the electronic structure of hydrogen. Of all the density functional theory calculations available, we report the most consistent results through quantum Monte Carlo simulations and coupled electron-ion Monte Carlo simulations of the LLPT in dense hydrogen. The critical point of the first-order LLPT is estimated above 2000 K according to the equation of state. Moreover, the metallization pressure obtained from the jump of dc electrical conductivity almost coincides with the plateau of equation of state.

physics.comp-ph↗

arXiv2019

Distributed Machine Learning through Heterogeneous Edge Systems

Many emerging AI applications request distributed machine learning (ML) among edge systems (e.g., IoT devices and PCs at the edge of the Internet), where data cannot be uploaded to a central venue for model training, due to their large volumes and/or security/privacy concerns. Edge devices are intrinsically heterogeneous in computing capacity, posing significant challenges to parameter synchronization for parallel training with the parameter server (PS) architecture. This paper proposes ADSP, a parameter synchronization scheme for distributed machine learning (ML) with heterogeneous edge systems. Eliminating the significant waiting time occurring with existing parameter synchronization models, the core idea of ADSP is to let faster edge devices continue training, while committing their model updates at strategically decided intervals. We design algorithms that decide time points for each worker to commit its model update, and ensure not only global model convergence but also faster convergence. Our testbed implementation and experiments show that ADSP outperforms existing parameter synchronization models significantly in terms of ML model convergence time, scalability and adaptability to large heterogeneity.

cs.DC↗

arXiv2019

An Optimal Stable Selective Model Inversion for Nonminimum-phase Systems

Stably inverting a dynamic system model is the foundation of numerous servo designs. Existing inversion techniques have provided accurate model approximations that are often highly effective in feedforward controls. However, when the inverse is implemented in a feedback system, additional considerations are needed for assuring causality, closed-loop stability, and robustness. In pursuit of bridging the gap between the best model matching and a robust feedback performance under closed-loop constraints, this paper provides a modern review of frequency-domain model inversion techniques and a new treatment of unstable zeros. We provide first a pole-zero-map-based intuitive inverse tuning for motion control systems. Then for general nonminimum-phase and unstable systems, we propose an optimal inversion algorithm that can attain model accuracy at the frequency regions of interest and meanwhile constrain noise amplification elsewhere to guarantee system robustness. The design goals are achieved by a multi-objective H infinity formulation and all-pass factorization that consider model matching, causality of transfer functions, frequency-domain gain constraints, and factorization of unstable system modes in a unified scheme. The proposed algorithm is validated on motion control systems and complex high-order systems.

eess.SY↗

arXiv2019

On the Phases of a Complex Matrix

In this paper, we define the phases of a complex sectorial matrix to be its canonical angles, which are uniquely determined from a sectorial decomposition of the matrix. Various properties of matrix phases are studied, including those of compressions, Schur complements, matrix products, and sums. In particular, by exploiting a notion known as the compound numerical range, we establish a majorization relation between the phases of the eigenvalues of $AB$ and the phases of $A$ and $B$. This is then applied to investigate the rank of $I + AB$ with phase information of $A$ and $B$, which plays a crucial role in feedback stability analysis. A pair of problems: banded sectorial matrix completion and decomposition is studied. The phases of the Kronecker and Hadamard products are also discussed.

math.RA↗

arXiv2019

High spatial resolution diffraction diagnostics for intense terahertz sources

We propose a high resolution spatial diagnostic method via inserting a millimeter-gap grating into the collimated terahertz beam to monitor the minute variation of the terahertz beam in strong-field terahertz sources, which is difficult to be resolved in conventional terahertz imaging systems. To verify the method, we intentionally fabricate tiny variations of the terahertz beam through tuning the iris for the infrared pumping beam before the tilted pulse-front pumping (TPFP) setups. The phenomena can be well explained by the the theory based on tilted pulse front technique and terahertz diffraction. We believe our observation not only help further understand the mechanism of intense terahertz generation, but also may be useful for strong-field terahertz applications.

physics.optics↗

arXiv2019

Mobile APP User Attribute Prediction by Heterogeneous Information Network Modeling

User-based attribute information, such as age and gender, is usually considered as user privacy information. It is difficult for enterprises to obtain user-based privacy attribute information. However, user-based privacy attribute information has a wide range of applications in personalized services, user behavior analysis and other aspects. this paper advances the HetPathMine model and puts forward TPathMine model. With applying the number of clicks of attributes under each node to express the user's emotional preference information, optimizations of the solution of meta-path weight are also presented. Based on meta-path in heterogeneous information networks, the new model integrates all relationships among objects into isomorphic relationships of classified objects. Matrix is used to realize the knowledge dissemination of category knowledge among isomorphic objects. The experimental results show that: (1) the prediction of user attributes based on heterogeneous information networks can achieve higher accuracy than traditional machine learning classification methods; (2) TPathMine model based on the number of clicks is more accurate in classifying users of different age groups, and the weight of each meta-path is consistent with human intuition or the real world situation.

cs.LG↗