Spectral embedding and spectral clustering are common methods for non-linear dimensionality reduction and clustering of complex high dimensional datasets. In this paper we provide a diffusion based probabilistic analy...
详细信息
ISBN:
(纸本)9783540737490
Spectral embedding and spectral clustering are common methods for non-linear dimensionality reduction and clustering of complex high dimensional datasets. In this paper we provide a diffusion based probabilistic analysis of algorithms that use the normalized graph Laplacian. Given the pairwise adjacency matrix of all points in a dataset, we define a random walk on the graph of points and a diffusion distance between any two points. We show that the diffusion distance is equal to the Euclidean distance in the embedded space with all eigenvectors of the normalized graph Laplacian. This identity shows that characteristic relaxation times and processes of the random walk on the graph are the key concept that governs the properties of these spectral clustering and spectral embedding algorithms. Specifically, for spectral clustering to succeed, a necessary condition is that the mean exit times from each cluster need to be significantly larger than the largest (slowest) of all relaxation times inside all of the individual clusters. For complex, multiscale data, this condition may not hold and multiscale methods need to be developed to handle such situations.
Active appearance model (AAM) has been widely used in face tracking and recognition. However, accuracy and efficiency are always two main challenges with the AAM search. The paper therefore proposed a fast appearance-...
详细信息
ISBN:
(纸本)9781424422944
Active appearance model (AAM) has been widely used in face tracking and recognition. However, accuracy and efficiency are always two main challenges with the AAM search. The paper therefore proposed a fast appearance-model based 3D face tracking algorithm to track a face appearance with significant translation, rotation, and scaling activities by using stochastic meta-descent (SMD) optimization scheme to accelerate the appearance model search and to improve the tracking efficiency and accuracy. The proposed algorithm constructs an active face appearance model by using several semantic landmark points extracted from each frame and then processes the appearance model search to approximate the model translating, rotating, and scaling by using the SMD filter to minimize the appearance difference between the current model and the new observation. We compared the results with both a conventional AAM and a Camshift filter and found that our algorithm outperforms both two in terms of efficiency and accuracy in tracking a fast moving, rotating, and scaling face object in a video sequence.
The lack of a written representation for American sign language (ASL) makes it difficult to do something as commonplace as looking up an unknown word in a dictionary. The majority of printed dictionaries organize ASL ...
详细信息
The lack of a written representation for American sign language (ASL) makes it difficult to do something as commonplace as looking up an unknown word in a dictionary. The majority of printed dictionaries organize ASL signs (represented in drawings or pictures) based on their nearest English translation; so unless one already knows the meaning of a sign, dictionary look-up is not a simple proposition. In this paper we introduce the ASL lexicon video dataset, a large and expanding public dataset containing video sequences of thousands of distinct ASL signs, as well as annotations of those sequences, including start/end frames and class label of every sign. This dataset is being created as part of a project to develop a computer vision system that allows users to look up the meaning of an ASL sign. At the same time, the dataset can be useful for benchmarking a variety of computer vision and machine learning methods designed for learning and/or indexing a large number of visual classes, and especially approaches for analyzing gestures and human communication.
In this paper, a simple method is proposed to evolve artificial neural networks(ANNs) using augmenting weight matrix method(AWMM). ANNs' architecture and connection weights can be evolved simultaneously by AWMM, a...
In this paper, a simple method is proposed to evolve artificial neural networks(ANNs) using augmenting weight matrix method(AWMM). ANNs' architecture and connection weights can be evolved simultaneously by AWMM, and their structures incrementally are growing up from minimal structure. It is a non-mating method. It employs 5 mutation operators: add connection, add node, delete connection, delete node, and new initial weight. And the connection weight is trained by the simplified alopex method, which is a correlation based method for solving optimization problem. In AWMM, structural information is encoded to weighting matrix, and the matrix is augmenting as the hidden nodes are added.
Mean-shift algorithm shows robust performances in various object-tracking technologies including face tracking. Due to its robustness and accuracy, mean-shift algorithm is regarded as one of the best ways to apply in ...
详细信息
Mean-shift algorithm shows robust performances in various object-tracking technologies including face tracking. Due to its robustness and accuracy, mean-shift algorithm is regarded as one of the best ways to apply in object-tracking technology in computer vision fields. However, it has a drawback of getting into a bottleneck state when faced with a speedy object moving beyond its window size within one image frame interval time. The time required to calculate mean-shift vector could be much lessened with lesser memory when color model is adjusted to the previously known target information. This paper shows the building process of target-adjusted model with a non-uniform quantization. The target color model dealt in this paper is the one used for deriving mean-shift vector. It is a kernel model containing both the color and distance information. This paper gives scheme to efficiently deal with color information in the model. Through a proper selection of color bins, unimportant color values were reduced to a small amount. As a result, the computing time of the mean-shift vector in face-tracking was shortened while maintaining robustness and accuracy.
We demonstrate methods for collecting gaseous and aerosolized particles into microfluidic channels. Surface tension creates gas-liquid interfaces to permit an analyte to transfer from a gaseous environment to a liquid...
详细信息
We demonstrate methods for collecting gaseous and aerosolized particles into microfluidic channels. Surface tension creates gas-liquid interfaces to permit an analyte to transfer from a gaseous environment to a liquid microfluidic channel. In one device, hydrophobic/hydrophilic boundaries create a virtual channel of liquid in which analyte is collected. In the other device, pinning points create circular pillars of air within a mircofluidic channel. Multiple methods demonstrate feasibility of these devices. Nesslerpsilas Reagent (NR) with gaseous ammonia (NH 3 ) showed the real-time and collection-for-later-analysis acquisition abilities. Deionized (DI) water with varying concentrations of gaseous ammonia (30%, 15%, and 7.5%) showed that the devices proportionally changed impedance (~50% per step) in real-time. Aerosolized solid particles showed that these devices were also able to collect larger sample analytes.
Density functional theory can accurately predict chemical and mechanical properties of nanostructures, although at a high computational cost. A quasicontinuum-like framework is proposed to substantially increase the s...
详细信息
Since the appearance in 1993, first approaching the Shannon limit, the turbo codes gave a new direction for the channel encoding field, especially since they were adopted for multiple norms of telecommunications, such...
详细信息
Since the appearance in 1993, first approaching the Shannon limit, the turbo codes gave a new direction for the channel encoding field, especially since they were adopted for multiple norms of telecommunications, such as deeper communication. To obtain an excellent performance, it is necessary to design robust turbo code interleaver. In this research, we investigated genetic algorithms as a promising optimization method to find good performing interleavers for large frame sizes. In this paper, we present our work, compare with several previous approaches and present experimental results.
Typical enterprise and military software systems consist of millions of lines of code with complicated dependence on diverse library abstractions. Manually debugging these codes imposes developers overwhelming workloa...
详细信息
Typical enterprise and military software systems consist of millions of lines of code with complicated dependence on diverse library abstractions. Manually debugging these codes imposes developers overwhelming workload and difficulties. To address software quality concerns efficiently, this paper proposes an ontology-based static analysis approach to automatically detect bugs in the source code of Java programs. First, we elaborate bug list collected, classify bugs into different categories, and translate bug patterns into SWRL (semantic Web rule language) rules using an ontology tool, Protege. An ontology model of Java program is created according to Java program specification using Protege as well. Both SWRL rules and the program ontology model are exported in OWL (Web ontology language) format. Second, Java source code under analysis is parsed into the abstract syntax tree (AST), which is automatically mapped to the individuals of the program ontology model. SWRL bridge takes in the exported OWL file (representing the SWRL rules model and program ontology model) and the individuals created for the Java code, conduits to Jess (a rule engine), and obtains inference results indicating any bugs. We perform experiments to compare bug detection capability with well-known FindBugs tool. A prototype of bug detector tool is developed to show the validity of the proposed static analysis approach.
One of the most known applications of Discrete Optimization is on scheduling. In contrast, one of the most known applications of Continuous Nonlinear Optimization is on the control of dynamic systems. In this paper, w...
详细信息
ISBN:
(纸本)9789955282839
One of the most known applications of Discrete Optimization is on scheduling. In contrast, one of the most known applications of Continuous Nonlinear Optimization is on the control of dynamic systems. In this paper, we combine both views, solving scheduling problems as dynamic systems, modeled as discrete-time nonlinear optimal control problems with state and control continuous variables subjected to upper and lower bounds. The proposed formulation has the following advantages over discrete (mixedinteger) models: a smaller number of variables is employed, and no 0-1 variable is needed. Therefore, the scheduling problem can be solved as a standard continuous nonlinear program. Complementarity constraints are used to represent scheduling decisions, defining a nonconvex problem, which can be solved with Global Optimization (GO) and Nonlinear programming (NLP) methods. Applications with a continuous process background are discussed, such as the ones from petroleum and water & wastewater industries, because they pose challenging issues, with a combination of nonlinear and combinatorial aspects. One example we discuss in detail is the crude oil scheduling in ports, with tanks, pipelines, jetties, and tanker vessels and blending operations. The recent literature on this problem is rich in mixedinteger linear programming (MILP) models, therefore we developed a procedure to reformulate certain mixed-integer constraints as complementarity constraints, discarding the associated binary variables. The resulting NLP model is equivalent to the original MILP, in a sense that a feasible point in the NLP is also a feasible point in the MILP. A number of numerical cases are discussed to illustrate the validity of this approach. Despite obtaining good results with the NLP approach, we acknowledge that the MILP has the desirable feature of having only global optima, whereas the NLP is non-convex. Therefore, we present an hybrid NLP-MILP scheme that uses the NLP to generate new MILP inte
暂无评论