arXiv subjects

Kai Zhou

Publications and source records attributed to Kai Zhou.

At least 181 records · Page 10Linked to original sources

arXiv2020

Unsupervised Outlier Detection in Heavy-Ion Collisions

We present different methods of unsupervised learning which can be used for outlier detection in high energy nuclear collisions. The UrQMD model is used to generate the bulk background of events as well as different variants of outlier events which may result from misidentified centrality or detector malfunctions. The methods presented here can be generalized to different and novel physics effects. To detect the outliers, dimensional reduction algorithms are implemented, specifically the Principle Component Analysis (PCA) and Autoencoders (AEN). We find that mainly the reconstruction error is a good measure to distinguish outliers from background. The performance of the algorithms is compared using a ROC curve. It is shown that the number of reduced (encoded) dimensions to describe a single event contributes significantly to the performance of the outlier detection task. We find that the model which is best suited to separate outlier events requires a good performance in reconstructing events and at the same time a small number of parameters.

hep-ex↗

arXiv2020

Robust Collective Classification against Structural Attacks

Collective learning methods exploit relations among data points to enhance classification performance. However, such relations, represented as edges in the underlying graphical model, expose an extra attack surface to the adversaries. We study adversarial robustness of an important class of such graphical models, Associative Markov Networks (AMN), to structural attacks, where an attacker can modify the graph structure at test time. We formulate the task of learning a robust AMN classifier as a bi-level program, where the inner problem is a challenging non-linear integer program that computes optimal structural changes to the AMN. To address this technical challenge, we first relax the attacker problem, and then use duality to obtain a convex quadratic upper bound for the robust AMN problem. We then prove a bound on the quality of the resulting approximately optimal solutions, and experimentally demonstrate the efficacy of our approach. Finally, we apply our approach in a transductive learning setting, and show that robust AMN is much more robust than state-of-the-art deep learning methods, while sacrificing little in accuracy on non-adversarial data.

cs.LG↗

arXiv2020

Identifying the nature of the QCD transition in relativistic collision of heavy nuclei with deep learning

Using deep convolutional neural network (CNN), the nature of the QCD transition can be identified from the final-state pion spectra from hybrid model simulations of heavy-ion collisions that combines a viscous hydrodynamic model with a hadronic cascade "after-burner". Two different types of equations of state (EoS) of the medium are used in the hydrodynamic evolution. The resulting spectra in transverse momentum and azimuthal angle are used as the input data to train the neural network to distinguish different EoS. Different scenarios for the input data are studied and compared in a systematic way. A clear hierarchy is observed in the prediction accuracy when using the event-by-event, cascade-coarse-grained and event-fine-averaged spectra as input for the network, which are about 80%, 90% and 99%, respectively. A comparison with the prediction performance by deep neural network (DNN) with only the normalized pion transverse momentum spectra is also made. High-level features of pion spectra captured by a carefully-trained neural network were found to be able to distinguish the nature of the QCD transition even in a simulation scenario which is close to the experiments.

hep-ph↗

arXiv2020

Heavy flavors under extreme conditions in high energy nuclear collisions

Heavy flavor hadrons have long been considered as a probe of the quark gluon plasma created in high energy nuclear collisions. In this paper we review the heavy flavor properties under extreme conditions and the realization in heavy ion experiments. After a short introduction on heavy flavor properties in vacuum, we emphasize the cold and hot nuclear matter effects on heavy flavors, including shadowing effect, Cronin effect and nuclear absorption for the former and Debye screening and regeneration for the latter. Then we discuss, in the frame of transport and coalescence models, these medium induced changes in open and closed heavy flavors in nuclear collisions and the comparison with nucleon-nucleon collisions. Considering the extremely strong electromagnetic and rotational fields generated in non-central nuclear collisions, which are widely studied in recent years, we finally investigate their effects on heavy flavor production and evolution in high energy nuclear collisions.

nucl-th↗

arXiv2020

Efficient Characterization of Dynamic Response Variation Using Multi-Fidelity Data Fusion through Composite Neural Network

Uncertainties in a structure is inevitable, which generally lead to variation in dynamic response predictions. For a complex structure, brute force Monte Carlo simulation for response variation analysis is infeasible since one single run may already be computationally costly. Data driven meta-modeling approaches have thus been explored to facilitate efficient emulation and statistical inference. The performance of a meta-model hinges upon both the quality and quantity of training dataset. In actual practice, however, high-fidelity data acquired from high-dimensional finite element simulation or experiment are generally scarce, which poses significant challenge to meta-model establishment. In this research, we take advantage of the multi-level response prediction opportunity in structural dynamic analysis, i.e., acquiring rapidly a large amount of low-fidelity data from reduced-order modeling, and acquiring accurately a small amount of high-fidelity data from full-scale finite element analysis. Specifically, we formulate a composite neural network fusion approach that can fully utilize the multi-level, heterogeneous datasets obtained. It implicitly identifies the correlation of the low- and high-fidelity datasets, which yields improved accuracy when compared with the state-of-the-art. Comprehensive investigations using frequency response variation characterization as case example are carried out to demonstrate the performance.

stat.ML↗

arXiv2020

Structural Model Updating Using Adaptive Multi-Response Gaussian Process Meta-modeling

Finite element model updating utilizing frequency response functions as inputs is an important procedure in structural analysis, design and control. This paper presents a highly efficient framework that is built upon Gaussian process emulation to inversely identify model parameters through sampling. In particular, a multi-response Gaussian process (MRGP) meta-modeling approach is formulated that can accurately construct the error response surface, i.e., the discrepancies between the frequency response predictions and actual measurement. In order to reduce the computational cost of repeated finite element simulations, an adaptive sampling strategy is established, where the search of unknown parameters is guided by the response surface features. Meanwhile, the information of previously sampled model parameters and the corresponding errors is utilized as additional training data to refine the MRGP meta-model. Two stochastic optimization techniques, i.e., particle swarm and simulated annealing, are employed to train the MRGP meta-model for comparison. Systematic case studies are conducted to examine the accuracy and robustness of the new framework of model updating.

cs.CE↗

arXiv2020

Double-dot interferometer for quantum measurement of Majorana qubits and stabilizers

Motivated by the need of quantum measurement of Majorana qubits and surface-code stabilizers, we analyze the performance of a double-dot interferometer under the influence of environment noise. The double-dot setup design allows accounting for the full multiple tunneling process between the dots through the Majorana island, within a master equation approach. In the co-tunneling regime, which results in a Majorana-mediated effective coupling between the dots, the master equation approach allows us to obtain analytic solutions for the measurement currents. The measurement quality, characterized by figures of merit such as the visibility of measurement signals, is carried out in regard to the unusual decoherence effect rather than `which-path' dephasing. The results obtained in this work are expected to be useful for future experiments of Majorana qubit and stabilizer measurements.

quant-ph↗

arXiv2020

Uncertainty Quantification of Mode Shape Variation Utilizing Multi-Level Multi-Response Gaussian Process

Mode shape information play the essential role in deciding the spatial pattern of vibratory response of a structure. The uncertainty quantification of mode shape, i.e., predicting mode shape variation when the structure is subjected to uncertainty, can provide guidance for robust design and control. Nevertheless, computational efficiency is a challenging issue. Direct Monte Carlo simulation is unlikely to be feasible especially for a complex structure with large number of degrees of freedom. In this research, we develop a new probabilistic framework built upon Gaussian process meta-modeling architecture to analyze mode shape variation. To expedite the generation of input dataset for meta-model establishment, a multi-level strategy is adopted which can blend a large amount of low-fidelity data acquired from order-reduced analysis with a small amount of high-fidelity data produced by high-dimensional full finite element analysis. To take advantage of the intrinsic relation of spatial distribution of mode shape, a multi-response strategy is incorporated to predict mode shape variation at different locations simultaneously. These yield a multi-level, multi-response Gaussian process that can efficiently and accurately quantify the effect of structural uncertainty to mode shape variation. Comprehensive case studies are carried out for demonstration and validation.

physics.data-an↗

arXiv2020

A Markovian influence graph formed from utility line outage data to mitigate large cascades

We use observed transmission line outage data to make a Markov influence graph that describes the probabilities of transitions between generations of cascading line outages, where each generation of a cascade consists of a single line outage or multiple line outages. The new influence graph defines a Markov chain and generalizes previous influence graphs by including multiple line outages as Markov chain states. The generalized influence graph can reproduce the distribution of cascade size in the utility data. In particular, it can estimate the probabilities of small, medium and large cascades. The influence graph has the key advantage of allowing the effect of mitigations to be analyzed and readily tested, which is not available from the observed data. We exploit the asymptotic properties of the Markov chain to find the lines most involved in large cascades and show how upgrades to these critical lines can reduce the probability of large cascades.

physics.soc-ph↗

arXiv2020

Bayesian estimates of transmission line outage rates that consider line dependencies

Transmission line outage rates are fundamental to power system reliability analysis. Line outages are infrequent, occurring only about once a year, so outage data are limited. We propose a Bayesian hierarchical model that leverages line dependencies to better estimate outage rates of individual transmission lines from limited outage data. The Bayesian estimates have a lower standard deviation than estimating the outage rates simply by dividing the number of outages by the number of years of data, especially when the number of outages is small. The Bayesian model produces more accurate individual line outage rates, as well as estimates of the uncertainty of these rates. Better estimates of line outage rates can improve system risk assessment, outage prediction, and maintenance scheduling.

stat.AP↗

arXiv2019

ICDAR 2019 Robust Reading Challenge on Reading Chinese Text on Signboard

Chinese scene text reading is one of the most challenging problems in computer vision and has attracted great interest. Different from English text, Chinese has more than 6000 commonly used characters and Chinesecharacters can be arranged in various layouts with numerous fonts. The Chinese signboards in street view are a good choice for Chinese scene text images since they have different backgrounds, fonts and layouts. We organized a competition called ICDAR2019-ReCTS, which mainly focuses on reading Chinese text on signboard. This report presents the final results of the competition. A large-scale dataset of 25,000 annotated signboard images, in which all the text lines and characters are annotated with locations and transcriptions, were released. Four tasks, namely character recognition, text line recognition, text line detection and end-to-end recognition were set up. Besides, considering the Chinese text ambiguity issue, we proposed a multi ground truth (multi-GT) evaluation method to make evaluation fairer. The competition started on March 1, 2019 and ended on April 30, 2019. 262 submissions from 46 teams are received. Most of the participants come from universities, research institutes, and tech companies in China. There are also some participants from the United States, Australia, Singapore, and Korea. 21 teams submit results for Task 1, 23 teams submit results for Task 2, 24 teams submit results for Task 3, and 13 teams submit results for Task 4. The official website for the competition is http://rrc.cvc.uab.es/?ch=12.

cs.CV↗

arXiv2019

A machine learning study to identify spinodal clumping in high energy nuclear collisions

The coordinate and momentum space configurations of the net baryon number in heavy ion collisions that undergo spinodal decomposition, due to a first-order phase transition, are investigated using state-of-the-art machine-learning methods. Coordinate space clumping, which appears in the spinodal decomposition, leaves strong characteristic imprints on the spatial net density distribution in nearly every event which can be detected by modern machine learning techniques. On the other hand, the corresponding features in the momentum distributions cannot clearly be detected, by the same machine learning methods, in individual events. Only a small subset of events can be systematically differentiated if only the momentum space information is available. This is due to the strong similarity of the two event classes, with and without spinodal decomposition. In such scenarios, conventional event-averaged observables like the baryon number cumulants signal a spinodal non-equilibrium phase transition. Indeed the third-order cumulant, the skewness, does exhibit a peak at the beam energy ($\mathrm{E}_{\mathrm{lab}}= 3-4$ A GeV), where the transient hot and dense system created in the heavy ion collision reaches the first-order phase transition.

nucl-th↗

arXiv2019

Kinetic approach to a relativistic BEC with inelastic processes

The phenomenon of Bose-Einstein condensation is investigated in the context of the Color-Glass-Condensate description of the initial state of ultrarelativistic heavy-ion collisions. For the first time, in this paper we study the influence of particle-number changing $2 \leftrightarrow 3$ processes on the transient formation of a Bose-Einstein Condensate within an isotropic system of scalar bosons by including $2 \leftrightarrow 3$ interactions of massive bosons with constant and isotropic cross sections, following a Boltzmann equation. The one-particle distribution function is decomposed in a condensate part and a non-zero momentum part of excited modes, leading to coupled integro-differential equations for the time evolution of the condensate and phase-space distribution function, which are then solved numerically. Our simulations converge to the expected equilibrium state, and only for $σ_{23}/σ_{22} \ll 1$ we find that a Bose-Einstein condensate emerges and decays within a finite lifetime in contrast to the case where only binary scattering processes are taken into account, and the condensate is stable due to particle-number conservation. Our calculations demonstrate that Bose-Einstein Condensates in the very early stage of heavy-ion collisions are highly unlikely, if inelastic collisions are significantly participating in the dynamical gluonic evolution.

hep-ph↗

arXiv2019

Scalar Quantization as Sparse Least Square Optimization

Quantization can be used to form new vectors/matrices with shared values close to the original. In recent years, the popularity of scalar quantization for value-sharing applications has been soaring as it has been found huge utilities in reducing the complexity of neural networks. Existing clustering-based quantization techniques, while being well-developed, have multiple drawbacks including the dependency of the random seed, empty or out-of-the-range clusters, and high time complexity for a large number of clusters. To overcome these problems, in this paper, the problem of scalar quantization is examined from a new perspective, namely sparse least square optimization. Specifically, inspired by the property of sparse least square regression, several quantization algorithms based on $l_1$ least square are proposed. In addition, similar schemes with $l_1 + l_2$ and $l_0$ regularization are proposed. Furthermore, to compute quantization results with a given amount of values/clusters, this paper designed an iterative method and a clustering-based method, and both of them are built on sparse least square. The paper shows that the latter method is mathematically equivalent to an improved version of k-means clustering-based quantization algorithm, although the two algorithms originated from different intuitions. The algorithms proposed were tested with three types of data and their computational performances, including information loss, time consumption, and the distribution of the values of the sparse vectors, were compared and analyzed. The paper offers a new perspective to probe the area of quantization, and the algorithms proposed can outperform existing methods especially under some bit-width reduction scenarios, when the required post-quantization resolution (number of values) is not significantly lower than the original number.

cs.LG↗

arXiv2019

Adversarial Robustness of Similarity-Based Link Prediction

Link prediction is one of the fundamental problems in social network analysis. A common set of techniques for link prediction rely on similarity metrics which use the topology of the observed subnetwork to quantify the likelihood of unobserved links. Recently, similarity metrics for link prediction have been shown to be vulnerable to attacks whereby observations about the network are adversarially modified to hide target links. We propose a novel approach for increasing robustness of similarity-based link prediction by endowing the analyst with a restricted set of reliable queries which accurately measure the existence of queried links. The analyst aims to robustly predict a collection of possible links by optimally allocating the reliable queries. We formalize the analyst problem as a Bayesian Stackelberg game in which they first choose the reliable queries, followed by an adversary who deletes a subset of links among the remaining (unreliable) queries by the analyst. The analyst in our model is uncertain about the particular target link the adversary attempts to hide, whereas the adversary has full information about the analyst and the network. Focusing on similarity metrics using only local information, we show that the problem is NP-Hard for both players, and devise two principled and efficient approaches for solving it approximately. Extensive experiments with real and synthetic networks demonstrate the effectiveness of our approach.

cs.AI↗

arXiv2019

Multi-layer Attention Mechanism for Speech Keyword Recognition

As an important part of speech recognition technology, automatic speech keyword recognition has been intensively studied in recent years. Such technology becomes especially pivotal under situations with limited infrastructures and computational resources, such as voice command recognition in vehicles and robot interaction. At present, the mainstream methods in automatic speech keyword recognition are based on long short-term memory (LSTM) networks with attention mechanism. However, due to inevitable information losses for the LSTM layer caused during feature extraction, the calculated attention weights are biased. In this paper, a novel approach, namely Multi-layer Attention Mechanism, is proposed to handle the inaccurate attention weights problem. The key idea is that, in addition to the conventional attention mechanism, information of layers prior to feature extraction and LSTM are introduced into attention weights calculations. Therefore, the attention weights are more accurate because the overall model can have more precise and focused areas. We conduct a comprehensive comparison and analysis on the keyword spotting performances on convolution neural network, bi-directional LSTM cyclic neural network, and cyclic neural network with the proposed attention mechanism on Google Speech Command datasets V2 datasets. Experimental results indicate favorable results for the proposed method and demonstrate the validity of the proposed method. The proposed multi-layer attention methods can be useful for other researches related to object spotting.

cs.LG↗

arXiv2019

Schubert Class and cyclotomic nilHecke algebras

Let $\ell, n$ be positive integers such that $\ell\geq n$. Let $\mathbb{G}_{n,\ell}$ be the Grassmannian which consists of the set of $n$-dimensional subspaces of $\mathbb{C}^{\ell}$. There is a $\mathbb{Z}$-graded algebra isomorphism between the cohomology $H^*(\mathbb{G}_{n,\ell},\mathbb{Z})$ of $\mathbb{G}_{n,\ell}$ and a natural $\mathbb{Z}$-form $B$ of the $\mathbb{Z}$-graded basic algebra of the type $A$ cyclotomic nilHecke algebra $H_{\ell,n}^{(0)}=\langleψ_1,\cdots,ψ_{n-1},y_1,\cdots,y_n\rangle$. In this paper, we show that the isomorphism can be chosen such that the image of each (geometrically defined) Schubert class $(a_1,\cdots,a_{n})$ coincides with the basis element $b_{\mathbfλ}$ constructed by Jun Hu and Xinfeng Liang by purely algebraic method, where $0\leq a_1\leq a_2\leq\cdots\leq a_{n}\leq \ell-n$ with $a_i\in\mathbb{Z}$ for each $i$, $\mathbfλ$ is the $\ell$-multipartition of $n$ associated to $(\ell+1-(a_{n}+n), \ell+1-(a_{n-1}+n-1),\cdots,\ell+1-(a_1+1))$. A similar correspondence between the Schubert class basis of the cohomology of the Grassmannian $\mathbb{G}_{\ell-n,\ell}$ and the $b_λ$'s basis of the natural $\mathbb{Z}$-form $B$ of the $\mathbb{Z}$-graded basic algebra of $H_{\ell,n}^{(0)}$ is also obtained. As an application, we obtain a second version of Giambelli formula for Schubert classes.

math.RT↗

arXiv2019

Interpretable deep learning for nuclear deformation in heavy ion collisions

The structure of heavy nuclei is difficult to disentangle in high-energy heavy-ion collisions. The deep convolution neural network (DCNN) might be helpful in mapping the complex final states of heavy-ion collisions to the nuclear structure in the initial state. Using DCNN for supervised regression, we successfully extracted the magnitude of the nuclear deformation from event-by-event correlation between the momentum anisotropy or elliptic flow ($v_2$) and total number of charged hadrons ($dN_{\rm ch}/dη$) within a Monte Carlo model. Furthermore, a degeneracy is found in the correlation between collisions of prolate-prolate and oblate-oblate nuclei. Using the Regression Attention Mask algorithm which is designed to interpret what has been learned by DCNN, we discovered that the correlation in total-overlapped collisions is sensitive to only large nuclear deformation, while the correlation in semi-overlapped collisions is discriminative for all magnitudes of nuclear deformation. The method developed in this study can pave a way for exploration of other aspects of nuclear structure in heavy-ion collisions.

nucl-th↗