Directly optimizing an information retrieval (IR) metric has become a hot topic in the field of learning to rank. Conventional wisdom believes that it is better to train for the loss function on which will be used for evaluation. But we often observe different results in reality. For example, directly optimizing averaged precision achieves higher performance than directly optimizing precision@3 when the ranking results are evaluated in terms of precision@3. This motivates us to combine multiple metrics in the process of optimizing IR metrics. For simplicity we study learning with two metrics. Since we usually conduct the learning process in a restricted hypothesis space, e.g., linear hypothesis space, it is usually difficult to maximize both metrics at the same time. To tackle this problem, we propose a relaxed approach in this paper. Specifically, we incorporate one metric within the constraint while maximizing the other one. By restricting the feasible hypothesis space, we can get a more robust ranking model. Empirical results on the benchmark data set LETOR show that the relaxed approach is superior to the direct linear combination approach, and also outperforms other baselines.
The construction process for a similarity matrix has an important impact on the performance of spectral clustering algorithms. In this paper, we propose a random walk based approach to process the Gaussian kernel similarity matrix. In this method, the pair-wise similarity between two data points is not only related to the two points, but also related to their neighbors. As a result, the new similarity matrix is closer to the ideal matrix which can provide the best clustering result. We give a theoretical analysis of the similarity matrix and apply this similarity matrix to spectral clustering. We also propose a method to handle noisy items which may cause deterioration of clustering performance. Experimental results on real-world data sets show that the proposed spectral clustering algorithm significantly outperforms existing algorithms.
This paper aims to address the problem of modeling human behavior patterns captured in surveillance videos for the application of online normal behavior recognition and anomaly detection. A novel framework is developed for automatic behavior modeling and online anomaly detection without the need for manual labeling of the training data set. The framework consists of the following key components. 1) A compact and effective behavior representation method is developed based on spatial-temporal interest point detection. 2) The natural grouping of behavior patterns is determined through a novel clustering algorithm, topic hidden Markov model (THMM) built upon the existing hidden Markov model (HMM) and latent Dirichlet allocation (LDA), which overcomes the current limitations in accuracy, robustness, and computational efficiency. The new model is a four-level hierarchical Bayesian model, in which each video is modeled as a Markov chain of behavior patterns where each behavior pattern is a distribution over some segments of the video. Each of these segments in the video can be modeled as a mixture of actions where each action is a distribution over spatial-temporal words. 3) An online anomaly measure is introduced to detect abnormal behavior, whereas normal behavior is recognized by runtime accumulative visual evidence using the likelihood ratio test (LRT) method. Experimental results demonstrate the effectiveness and robustness of our approach using noisy and sparse data sets collected from a real surveillance scenario.
Moving object detection in dynamic scenes is a basic task in a surveillance system for sensor data collection. In this paper, we present a powerful background subtraction algorithm called Gaussian-kernel density estimator (G-KDE) that improves the accuracy and reduces the computational load. The main innovation is that we divide the changes of background into continuous and stable changes to deal with dynamic scenes and moving objects that first merge into the background, and separately model background using both KDE model and Gaussian models. To get a temporal-spatial background model, the sample selection is based on the concept of region average at the update stage. In the detection stage, neighborhood information content (NIC) is implemented which suppresses the false detection due to small and un-modeled movements in the scene. The experimental results which are generated on three separate sequences indicate that this method is well suited for precise detection of moving objects in complex scenes and it can be efficiently used in various detection systems.
Media access control (MAC) protocols control how nodes access a shared wireless channel. It is critical to the performance of wireless sensor networks (WSN). An adaptive polling interval and short preamble MAC protocol (AX-MAC) is proposed in this paper. AX-MAC is an asynchronous protocol which composed of two basic features. First, rendezvous between the sender and the receiver is reached by a series of short preambles. Second, nodes dynamically adjust their polling intervals according to network traffic conditions. Threshold parameters used to determine traffic conditions and adjust polling intervals are analyzed based on a Markov chain. Energy consumption and network latency are also discussed in detail. Simulation results indicate that AX-MAC is suited to dynamic network traffic conditions and is superior to both X-MAC and Boost-MAC in energy consumption and latency.
In a hostile environment, sensor nodes may be compromised and then be used to launch various attacks. One severe attack is false data injection which is becoming a serious threat to wireless sensor networks. An attacker uses the compromised node to flood the network and exhaust network resources by injecting a large number of bogus packets. In this paper, we study how to locate the attack node using a framework of packet marking and packet logging. We propose a combined packet marking and logging scheme for traceback (CPMLT). In CPMLT, one packet can be marked by up to
Radio frequency identification (RFID) systems suffer many security risks because they use an insecure wireless communication channel between tag and reader. In this paper, we analyze two recently proposed RFID authentication protocols. Both protocols are vulnerable to tag information leakage and untraceability attacks. For the attack on the first protocol, the adversary only needs to eavesdrop on the messages between reader and tag, and then perform an XOR operation. To attack the second protocol successfully, the adversary may execute a series of carefully designed challenges to determine the tag’s identification.
There is a growing tendency for people in the community of object-oriented methods to use pre- and post-conditions to write formal specifications for operations (methods) of classes. The motivation for trying to take advantage of well established formalism in precisely defining the functionality of operations is laudable, but unfortunately this exercise may be flawed because the use of pre- and post-conditions containing method calls (or similar) with side effects are likely to cause confusion in the interpretation of specifications. This paper analyzes, with comprehensible examples, why using pre-post notation is not effective to specify operations in object-oriented systems in general, discusses existing approaches to using pre-post notation for object-oriented systems, and offers some solutions to the problem.
Substructural type systems are designed from the insight inspired by the development of linear and substructural logics. Substructural type systems promise to control the usage of computational resources statically, thus detect more program errors at an early stage than traditional type systems do. In the past decade, substructural type systems have been deployed in the design of novel programming languages, such as Vault, etc. This paper presents a general typing theory for substructural type system. First, we define a universal semantic framework for substructural types by interpreting them as characteristic intervals composed of type qualifiers. Based on this framework, we present the design of a substructural calculus λ