Proper Instance Research Articles

HWMP-based secure communication of multi-agent systems

Multi-agent systems (MAS), known for their autonomy, fault tolerance, and flexible collaboration capabilities, are widely used in various fields. Communication security is one of the most critical problems of MAS tasks. Once a malicious attacker invades agents in the system, it brings massive threats to the entire system. It tampers with routing messages, discards packets, and paralyzes the communication network. To implement collaboration tasks of MASs safely and efficiently, this work makes security improvements on the hybrid wireless mesh protocol (HWMP) with an additional authentication system. We enhance the proactive and reactive mode of the HWMP and apply it to an integrated system, i.e., the MAS formation control as a proper instance. We present a HWMP-based secure communication model (HSCM) for MASs. The theoretical analysis and simulation results demonstrate a significant enhancement in network performance under various threats, including packet loss rate, throughput, etc. The results indicate that HSCM plays a significant role in resisting common attacks (such as black hole and flooding attacks) when compared to multi-agent systems without any security protection. There is an approximate 30% increase in packet delivery rate, a substantial enhancement in throughput, and a reduction in routing overhead. The latency also fluctuates within reasonable bounds. Our solution provides secure communication for formation control under malicious attacks.

A deep learning network based end-to-end image composition

Currently, high-quality image composition largely depends on multiple user interactions and complex manual operations. In particular, the process of composition object extraction and region determination has become a burden that cannot be underestimated, restricting wider applications. Aiming at this problem, we propose an end-to-end image composition method that combines powerful deep-learning-based application modules such as image retrieval and instance segmentation to realize efficient non-interactive image composition. Specifically, the retrieval module, which is based on the attention mechanism, can determine semantically similar material images. Moreover, the content of interest (COI) extraction and optimization procedure is able to select the most proper instance among the material images. Finally, we propose the double-sieving strategy, which locates the best composition position in the target image. Using these effective modules, we carried out niche targeting experiments using an image database with high plausibility. The realistic experimental results illustrate that our method can achieve effective and reasonable end-to-end image composition.

The social (and cultural, and syntactic, and semantic) life of generics

ABSTRACTRecent work in linguistics, philosophy, and psychology suggests that the distinction between generic and specific (singular) reference is foundational to concept formation, and hence of special interest to social scientists. Generics provide the first-language learner with external evidence of the integrity of a word/concept cluster, partially filling in the scaffolding of concepts. As such, they arereplicators,critical to the transmission of concepts across populations and across time. Generics are tacitly normative. As they refer to the constitutive properties of a concept rather than to its object, they tell us what—in a given social setting—a proper instance of the concept should look like. Generics sustain and reproduce social stereotypes, including—and perhaps especially—ethnoracial, class, and gender stereotypes. (Generics, conceptual formation, ethnography, tokenization, materiality)*

Abstract Concept Instantiation with Context Relevance Measurement

In different contexts, one abstract concept (e.g., fruit) may be mapped into different concrete instance sets, which is called abstract concept instantiation. It has been widely applied in many applications, such as web search, intelligent recommendation, etc. However, in most abstract concept instantiation models have the following problems: (1) the neglect of incorrect label and label incompleteness in the category structure on which instance selection relies; (2) the subjective design of instance profile for calculating the relevance between instance and contextual constraint. The above problems lead to false prediction in terms of abstract concept instantiation. To tackle these problems, we proposed a novel model to instantiate the abstract concept. Firstly, to alleviate the incorrect label and remedy label incompleteness in the category structure, an improved random-walk algorithm is proposed, called InstanceRank, which not only utilize the category information, but it also exploits the association information to infer the right instances of an abstract concept. Secondly, for better measuring the relevance between instances and contextual constraint, we learn the proper instance profile from different granularity ones. They are designed based on the surrounding text of the instance. Finally, noise reduction and instance filtering are introduced to further enhance the model performance. Experiments on Chinese food abstract concept set show that the proposed model can effectively reduce false positive and false negative of instantiation results.

Query-Driven Multi-Instance Learning

We introduce a query-driven approach (qMIL) to multi-instance learning where the queries aim to uncover the class labels embodied in a given bag of instances. Specifically, it solves a multi-instance multi-label learning (MIML) problem with a more challenging setting than the conventional one. Each MIML bag in our formulation is annotated only with a binary label indicating whether the bag contains the instance of a certain class and the query is specified by the word2vec of a class label/name. To learn a deep-net model for qMIL, we construct a network component that achieves a generalized compatibility measure for query-visual co-embedding and yields proper instance attentions to the given query. The bag representation is then formed as the attention-weighted sum of the instances' weights, and passed to the classification layer at the end of the network. In addition, the qMIL formulation is flexible for extending the network to classify unseen class labels, leading to a new technique to solve the zero-shot MIML task through an iterative querying process. Experimental results on action classification over video clips and three MIML datasets from MNIST, CIFAR10 and Scene are provided to demonstrate the effectiveness of our method.

Tight Bound on Incnetive Ratio for Sybil Attack in Resource Sharing System

In this paper, we discuss the Sybil attack on a sharing-based economic system where each participant contributes its own resource for all to share. Such an attack is possible especially in the cloud computing model where agents can exchange information with the cloud and obtain aggregated information from it. We are interested in the robustness of the market equilibrium mechanism against such an attack. We adopt the incentive ratio to measure the gain that a participant can make by splitting its identity and reconstructing communication connections with others. On the one hand, we show that no player can increase more than $\sqrt{2}$ times its original share in the market equilibrium solution by characterizing the worst case, in which a strategic agent can obtain the maximal gain in utility by playing the Sybil attack. On the other hand, the bound of $\sqrt{2}$ is proved to be tight by constructing a proper instance. We also simulate on a series of random graphs and observe that the incentive ratio was no more than two in the general setting.

Toward location-aware Web: extraction method, applications and evaluation

Location-based services (LBS) belong to one of the most popular types of services today. However, a recurring issue is that most of the content in LBS has to be created from scratch and needs to be explicitly tagged to locations, which makes existing Web content not directly usable for LBS. In this paper, we aim at making Web sites location-aware and feed this information to LBS. Our approach toward location-aware Web is threefold: First, we present a extraction method: SALT. It receives Web sites as input and equips them with tags. Compared to other approaches, SALT is capable of extracting locations with a precision up to the street level. Performance evaluations further show high applicability for practice. Second, we present three applications for SALT: Webnear.me, Local Browsing and Local Facebook. Webnear.me offers location-aware Web surfing through a mobile Web site and a smartphone app. Local Browsing adds the feature to browse by nearby tags, extracted from Web sites delivered by SALT. Local Facebook extends tagging to social networks, allowing to run SALT on one's own and one's friends' timeline. Finally, we evaluate SALT for technology acceptance of Webnear.me through a formative user study. Through real user data, collected during a 3 months pilot field deployment of Webnear.me, we assess whether SALT is a proper instance of location of a Web site.