arXiv subjects

Shasha Li

Publications and source records attributed to Shasha Li.

69 records · Page 4Linked to original sources

arXiv2020

Connecting the Dots: Detecting Adversarial Perturbations Using Context Inconsistency

There has been a recent surge in research on adversarial perturbations that defeat Deep Neural Networks (DNNs) in machine vision; most of these perturbation-based attacks target object classifiers. Inspired by the observation that humans are able to recognize objects that appear out of place in a scene or along with other unlikely objects, we augment the DNN with a system that learns context consistency rules during training and checks for the violations of the same during testing. Our approach builds a set of auto-encoders, one for each object class, appropriately trained so as to output a discrepancy between the input and output if an added adversarial perturbation violates context consistency rules. Experiments on PASCAL VOC and MS COCO show that our method effectively detects various adversarial attacks and achieves high ROC-AUC (over 0.95 in most cases); this corresponds to over 20% improvement over a state-of-the-art context-agnostic method.

cs.CV↗

arXiv2020

$k$-tree connectivity of line graphs

For a graph $G=(V,E)$ and a set $S\subseteq V(G)$ of size at least $2$, an $S$-Steiner tree $T$ is a subgraph of $G$ that is a tree with $S\subseteq V(T)$. Two $S$-Steiner trees $T$ and $T'$ are internally disjoint (resp. edge-disjoint) if $E(T)\cap E(T')=\emptyset$ and $V(T)\cap V(T')=S$ (resp. if $E(T)\cap E(T')=\emptyset$). Let $κ_G (S)$ (resp. $λ_G (S)$) denote the maximum number of internally disjoint (resp. edge-disjoint) $S$-Steiner trees in $G$. The $k$-tree connectivity $κ_k(G)$ (resp. $k$-tree edge-connectivity $λ_k(G)$) of $G$ is then defined as the minimum $κ_G (S)$ (resp. $λ_G (S)$), where $S$ ranges over all $k$-subsets of $V(G)$. In [H. Li, B. Wu, J. Meng, Y. Ma, Steiner tree packing number and tree connectivity, Discrete Math. 341(2018), 1945--1951], the authors conjectured that if a connected graph $G$ has at least $k$ vertices and at least $k$ edges, then $κ_k(L(G))\geq λ_k(G)$ for any $k\geq 2$, where $L(G)$ is the line graph of $G$. In this paper, we confirm this conjecture and prove that the bound is sharp.

math.CO↗

arXiv2020

A4 : Evading Learning-based Adblockers

Efforts by online ad publishers to circumvent traditional ad blockers towards regaining fiduciary benefits, have been demonstrably successful. As a result, there have recently emerged a set of adblockers that apply machine learning instead of manually curated rules and have been shown to be more robust in blocking ads on websites including social media sites such as Facebook. Among these, AdGraph is arguably the state-of-the-art learning-based adblocker. In this paper, we develop A4, a tool that intelligently crafts adversarial samples of ads to evade AdGraph. Unlike the popular research on adversarial samples against images or videos that are considered less- to un-restricted, the samples that A4 generates preserve application semantics of the web page, or are actionable. Through several experiments we show that A4 can bypass AdGraph about 60% of the time, which surpasses the state-of-the-art attack by a significant margin of 84.3%; in addition, changes to the visual layout of the web page due to these perturbations are imperceptible. We envision the algorithmic framework proposed in A4 is also promising in improving adversarial attacks against other learning-based web applications with similar requirements.

cs.CR↗

arXiv2019

Spin-to-Orbital Conversion of Angular Momentum in Harmonic Generation Driven by Intense Circularly Polarized Beam

A new physical mechanism to achieve spin-to-orbital angular momentum conversion based on the interaction of an intense circularly polarized (CP) laser beam with a plane foil is presented and studied for the first time. It has been verified by both simulation and theoretical analysis that vortex harmonics carrying orbital angular momentum (OAM) are generated after a relativistic CP laser beam, even a Gaussian beam, impinges normally on a plane foil. The generation of this vortex harmonics is attributed to the vortex oscillation of the plasma surface driven harmonically by the vortex longitudinal electric field of the CP beam. During the process of harmonic generation, the spin angular momenta of fundamental-frequency photons are converted to OAM of harmonic photon because of the conservation of total angular momentum. In addition, if an initially vortex beam or a spiral phase plate is used, the OAM of harmonic photon can be more tunable and controllable.

physics.plasm-ph↗

arXiv2018

On the complexity of k-rainbow cycle colouring problems

An edge-coloured cycle is $rainbow$ if all edges of the cycle have distinct colours. For $k\geq 1$, let $\mathcal{F}_{k}$ denote the family of all graphs with the property that any $k$ vertices lie on a cycle. For $G\in \mathcal{F}_{k}$, a $k$-$rainbow$ $cycle$ $colouring$ of $G$ is an edge-colouring such that any $k$ vertices of $G$ lie on a rainbow cycle in $G$. The $k$-$rainbow$ $cycle$ $index$ of $G$, denoted by $crx_{k}(G)$, is the minimum number of colours needed in a $k$-rainbow cycle colouring of $G$. In this paper, we restrict our attention to the computational aspects of $k$-rainbow cycle colouring. First, we prove that the problem of deciding whether $crx_1=3$ can be solved in polynomial time, but that of deciding whether $crx_1 \leq k$ is NP-Complete, where $k\geq 4$. Then we show that the problem of deciding whether $crx_2=3$ can be solved in polynomial time, but those of deciding whether $crx_2 \leq 4$ or $5$ are NP-Complete. Furthermore, we also consider the cases of $crx_3=3$ and $crx_3 \leq 4$. Finally, We prove that the problem of deciding whether a given edge-colouring (with an unbounded number of colours) of a graph is a $k$-rainbow cycle colouring, is NP-Complete for $k=1$, $2$ and $3$, respectively. Some open problems for further study are mentioned.

math.CO↗

arXiv2018

Adversarial Perturbations Against Real-Time Video Classification Systems

Recent research has demonstrated the brittleness of machine learning systems to adversarial perturbations. However, the studies have been mostly limited to perturbations on images and more generally, classification that does not deal with temporally varying inputs. In this paper we ask "Are adversarial perturbations possible in real-time video classification systems and if so, what properties must they satisfy?" Such systems find application in surveillance applications, smart vehicles, and smart elderly care and thus, misclassification could be particularly harmful (e.g., a mishap at an elderly care facility may be missed). We show that accounting for temporal structure is key to generating adversarial examples in such systems. We exploit recent advances in generative adversarial network (GAN) architectures to account for temporal correlations and generate adversarial samples that can cause misclassification rates of over 80% for targeted activities. More importantly, the samples also leave other activities largely unaffected making them extremely stealthy. Finally, we also surprisingly find that in many scenarios, the same perturbation can be applied to every frame in a video clip that makes the adversary's ability to achieve misclassification relatively easy.

cs.LG↗

arXiv2017

Rainbow vertex connection of digraphs

An edge-coloured path is \emph{rainbow} if its edges have distinct colours. An edge-coloured connected graph is said to be \emph{rainbow connected} if any two vertices are connected by a rainbow path, and \emph{strongly rainbow connected} if any two vertices are connected by a rainbow geodesic. The (\emph{strong}) \emph{rainbow connection number} of a connected graph is the minimum number of colours needed to make the graph (strongly) rainbow connected. These two graph parameters were introduced by Chartrand, Johns, McKeon and Zhang in 2008. As an extension, Krivelevich and Yuster proposed the concept of \emph{rainbow vertex-connection}. The topic of rainbow connection in graphs drew much attention and various similar parameters were introduced, mostly dealing with undirected graphs. Dorbec, Schiermeyer, Sidorowicz and Sopena extended the concept of the rainbow connection to digraphs. In this paper, we consider the (strong) rainbow vertex-connection number of digraphs. Results on the (strong) rainbow vertex-connection number of biorientations of graphs, cycle digraphs, circulant digraphs and tournaments are presented.

math.CO↗

arXiv2017

Drug-drug Interaction Extraction via Recurrent Neural Network with Multiple Attention Layers

Drug-drug interaction (DDI) is a vital information when physicians and pharmacists intend to co-administer two or more drugs. Thus, several DDI databases are constructed to avoid mistakenly combined use. In recent years, automatically extracting DDIs from biomedical text has drawn researchers' attention. However, the existing work utilize either complex feature engineering or NLP tools, both of which are insufficient for sentence comprehension. Inspired by the deep learning approaches in natural language processing, we propose a recur- rent neural network model with multiple attention layers for DDI classification. We evaluate our model on 2013 SemEval DDIExtraction dataset. The experiments show that our model classifies most of the drug pairs into correct DDI categories, which outperforms the existing NLP or deep learning methods.

cs.CL↗

arXiv2016

Effector Detection in Social Networks

In a social network, influence diffusion is the process of spreading innovations from user to user. An activation state identifies who are the active users who have adopted the target innovation. Given an activation state of a certain diffusion, effector detection aims to reveal the active users who are able to best explain the observed state. In this paper, we tackle the effector detection problem from two perspectives. The first approach is based on the influence distance that measures the chance that an active user can activate its neighbors. For a certain pair of users, the shorter the influence distance, the higher probability that one can activate the other. Given an activation state, the effectors are expected to have short influence distance to active users while long to inactive users. By this idea, we propose the influence-distance-based effector detection problem and provide a 3-approximation. Second, we address the effector detection problem by the maximum likelihood estimation (MLE) approach. We prove that the optimal MLE can be obtained in polynomial time for connected directed acyclic graphs. For general graphs, we first extract a directed acyclic subgraph that can well preserve the information in the original graph and then apply the MLE approach to the extracted subgraph to obtain the effectors. The effectiveness of our algorithms is experimentally verified via simulations on the real-world social network.

cs.SI↗

arXiv2011

Note on the complexity of deciding the rainbow connectedness for bipartite graphs

A path in an edge-colored graph is said to be a rainbow path if no two edges on the path have the same color. An edge-colored graph is (strongly) rainbow connected if there exists a rainbow (geodesic) path between every pair of vertices. The (strong) rainbow connection number of $G$, denoted by ($scr(G)$, respectively) $rc(G)$, is the smallest number of colors that are needed in order to make $G$ (strongly) rainbow connected. Though for a general graph $G$ it is NP-Complete to decide whether $rc(G)=2$, in this paper, we show that the problem becomes easy when $G$ is a bipartite graph. Moreover, it is known that deciding whether a given edge-colored (with an unbound number of colors) graph is rainbow connected is NP-Complete. We will prove that it is still NP-Complete even when the edge-colored graph is bipartite. We also show that a few NP-hard problems on rainbow connection are indeed NP-Complete.

cs.CC↗

arXiv2011

Note on the minimal size of a graph with generalized connectivity kappa_3= 2

The concept of generalized $k$-connectivity $κ_{k}(G)$ of a graph $G$ was introduced by Chartrand et al. in recent years. In our early paper, extremal theory for this graph parameter was started. We determined the minimal number of edges of a graph of order $n$ with $κ_{3}= 2$, i.e., for a graph $G$ of order $n$ and size $e(G)$ with $κ_{3}(G)= 2$, we proved that $e(G)\geq (6/5)n$, and the lower bound is sharp by constructing a class of graphs, only for $n\equiv 0 \ (mod \ 5)$ and $n\neq 10$. In this paper, we improve the lower bound to $\lceil(6/5)n\rceil$. Moreover, we show that for all $n\geq 4$ but $n= 9, 10$, there always exists a graph of order $n$ with $κ_{3}= 2$ whose size attains the lower bound $\lceil(6/5)n\rceil$. Whereas for $n= 9, 10$ we give examples to show that $\lceil(6/5)n\rceil+1$ is the best possible lower bound. This gives a clear picture on the minimal size of a graph of order $n$ with generalized connectivity $κ_{3}= 2$.

math.CO↗

arXiv2011

The minimal size of a graph with generalized connectivity $κ_3 = 2$

Let $G$ be a nontrivial connected graph of order $n$ and $k$ an integer with $2\leq k\leq n$. For a set $S$ of $k$ vertices of $G$, let $κ(S)$ denote the maximum number $\ell$ of edge-disjoint trees $T_1,T_2,...,T_\ell$ in $G$ such that $V(T_i)\cap V(T_j)=S$ for every pair $i,j$ of distinct integers with $1\leq i,j\leq \ell$. Chartrand et al. generalized the concept of connectivity as follows: The $k$-$connectivity$, denoted by $κ_k(G)$, of $G$ is defined by $κ_k(G)=$min$\{κ(S)\}$, where the minimum is taken over all $k$-subsets $S$ of $V(G)$. Thus $κ_2(G)=κ(G)$, where $κ(G)$ is the connectivity of $G$. This paper mainly focuses on the minimal number of edges of a graph $G$ with $κ_{3}(G)= 2$. For a graph $G$ of order $v(G)$ and size $e(G)$ with $κ_{3}(G)= 2$, we obtain that $e(G)\geq 6/5v(G)$, and the lower bound is sharp by showing a class of examples attaining the lower bound.

math.CO↗

arXiv2010

The generalized connectivity of complete bipartite graphs

Let $G$ be a nontrivial connected graph of order $n$, and $k$ an integer with $2\leq k\leq n$. For a set $S$ of $k$ vertices of $G$, let $κ(S)$ denote the maximum number $\ell$ of edge-disjoint trees $T_1,T_2,...,T_\ell$ in $G$ such that $V(T_i)\cap V(T_j)=S$ for every pair $i,j$ of distinct integers with $1\leq i,j\leq \ell$. Chartrand et al. generalized the concept of connectivity as follows: The $k$-$connectivity$, denoted by $κ_k(G)$, of $G$ is defined by $κ_k(G)=$min$\{κ(S)\}$, where the minimum is taken over all $k$-subsets $S$ of $V(G)$. Thus $κ_2(G)=κ(G)$, where $κ(G)$ is the connectivity of $G$. Moreover, $κ_{n}(G)$ is the maximum number of edge-disjoint spanning trees of $G$. This paper mainly focus on the $k$-connectivity of complete bipartite graphs $K_{a,b}$. First, we obtain the number of edge-disjoint spanning trees of $K_{a,b}$, which is $\lfloor\frac{ab}{a+b-1}\rfloor$, and specifically give the $\lfloor\frac{ab}{a+b-1}\rfloor$ edge-disjoint spanning trees. Then based on this result, we get the $k$-connectivity of $K_{a,b}$ for all $2\leq k \leq a+b$. Namely, if $k>b-a+2$ and $a-b+k$ is odd then $κ_{k}(K_{a,b})=\frac{a+b-k+1}{2}+\lfloor\frac{(a-b+k-1)(b-a+k-1)}{4(k-1)}\rfloor,$ if $k>b-a+2$ and $a-b+k$ is even then $κ_{k}(K_{a,b})=\frac{a+b-k}{2}+\lfloor\frac{(a-b+k)(b-a+k)}{4(k-1)}\rfloor,$ and if $k\leq b-a+2$ then $κ_{k}(K_{a,b})=a. $

math.CO↗

arXiv2010

Hardness results on generalized connectivity

Let $G$ be a nontrivial connected graph of order $n$ and let $k$ be an integer with $2\leq k\leq n$. For a set $S$ of $k$ vertices of $G$, let $κ(S)$ denote the maximum number $\ell$ of edge-disjoint trees $T_1,T_2,...,T_\ell$ in $G$ such that $V(T_i)\cap V(T_j)=S$ for every pair $i,j$ of distinct integers with $1\leq i,j\leq \ell$. A collection $\{T_1,T_2,...,T_\ell\}$ of trees in $G$ with this property is called an internally disjoint set of trees connecting $S$. Chartrand et al. generalized the concept of connectivity as follows: The $k$-$connectivity$, denoted by $κ_k(G)$, of $G$ is defined by $κ_k(G)=$min$\{κ(S)\}$, where the minimum is taken over all $k$-subsets $S$ of $V(G)$. Thus $κ_2(G)=κ(G)$, where $κ(G)$ is the connectivity of $G$, for which there are polynomial-time algorithms to solve it. This paper mainly focus on the complexity of the generalized connectivity. At first, we obtain that for two fixed positive integers $k_1$ and $k_2$, given a graph $G$ and a $k_1$-subset $S$ of $V(G)$, the problem of deciding whether $G$ contains $k_2$ internally disjoint trees connecting $S$ can be solved by a polynomial-time algorithm. Then, we show that when $k_1$ is a fixed integer of at least 4, but $k_2$ is not a fixed integer, the problem turns out to be NP-complete. On the other hand, when $k_2$ is a fixed integer of at least 2, but $k_1$ is not a fixed integer, we show that the problem also becomes NP-complete. Finally we give some open problems.

math.CO↗

arXiv2009

Sharp bounds for the generalized connectivity $κ_3(G)$

Let $G$ be a nontrivial connected graph of order $n$ and let $k$ be an integer with $2\leq k\leq n$. For a set $S$ of $k$ vertices of $G$, let $κ(S)$ denote the maximum number $\ell$ of edge-disjoint trees $T_1,T_2,...,T_\ell$ in $G$ such that $V(T_i)\cap V(T_j)=S$ for every pair $i,j$ of distinct integers with $1\leq i,j\leq \ell$. A collection $\{T_1,T_2,...,T_\ell\}$ of trees in $G$ with this property is called an internally disjoint set of trees connecting $S$. Chartrand et al. generalized the concept of connectivity as follows: The $k$-$connectivity$, denoted by $κ_k(G)$, of $G$ is defined by $κ_k(G)=$min$\{κ(S)\}$, where the minimum is taken over all $k$-subsets $S$ of $V(G)$. Thus $κ_2(G)=κ(G)$, where $κ(G)$ is the connectivity of $G$. In general, the investigation of $κ_k(G)$ is very difficult. We therefore focus on the investigation on $κ_3(G)$ in this paper. We study the relation between the connectivity and the 3-connectivity of a graph. First we give sharp upper and lower bounds of $κ_3(G)$ for general graphs $G$, and construct two kinds of graphs which attain the upper and lower bound, respectively. We then show that if $G$ is a connected planar graph, then $κ(G)-1 \leq κ_3(G)\leq κ(G)$, and give some classes of graphs which attain the bounds. In the end we show that the problem whether $κ(G)=κ_3(G)$ for a planar graph $G$ can be solved in polynomial time.

math.CO↗