FindOut: Finding Outliers in Very Large Datasets

Finding the rare instances or the outliers is important in many KDD (knowledge discovery and data-mining) applications, such as detecting credit card fraud or finding irregularities in gene expressions. Signal-processing techniques have been introduced to transform images for enhancement, filtering, restoration, analysis, and reconstruction. In this paper, we present a new method in which we apply signal-processing techniques to solve important problems in data mining. In particular, we introduce a novel deviation (or outlier) detection approach, termed FindOut, based on wavelet transform. The main idea in FindOut is to remove the clusters from the original data and then identify the outliers. Although previous research showed that such techniques may not be effective because of the nature of the clustering, FindOut can successfully identify outliers from large datasets. Experimental results on very large datasets are presented which show the efficiency and effectiveness of the proposed approach. Received 7 September 2000 / Revised 2 February 2001 / Accepted in revised form 31 May 2001 Correspondence and offprint requests to: A. Zhang, Department of Computer Science and Engineering, State University of New York at Buffalo, Buffalo, NY 14260, USA. Email: azhang@cse.buffalo.eduau     

Design of all-optical tunable filter based on two-dimensional photonic crystals for WDM (wave division multiplexing) applications

In this paper, all-optical tunable filters based on two-dimensional photonic crystals with very small dimensions for optical telecommunication of WDM technology are designed and simulated. The structure is made of air holes in a dielectric background. The tuning is done by changing resonant defect angle. The channels obtained for this structure will be set in wavelength range of 1550 nm. Created channels are at wavelengths of 1550, 1551, 1552, and 1565 (16 channels); the distance between adjacent channels is 1 nm. Design and simulation of this filter is done by RSOFT software. Quality factor, transmission efficiency, and band gap shows that filter performance is very good.     

Describing a strategy to estimate the CO2-heavy oil minimum miscibility pressure based on the experimental methods

To measure MMP, slim tube method is time consuming, especially for heavy oils. In this study, a strategy is considered to measure the MMP in one of the Iranian heavy oil reservoirs. The strategy steps were done through; partial data of slim tube, coupling numerical model (CNM), full slim tube, and VIT test. The results showed that CNM technique with two experimental data obtains the MMP value faster than running a full slim tube with low error result. However, VIT determined MMP faster and its results illustrated that miscibility pressure for heavy oil occurs at zero IFT, unlike light oils.     

Forming Limit Diagram Prediction of Tailor-Welded Blank Using Experimental and Numerical Methods

The forming limit diagram (FLD) is a useful method for characterizing the formability of sheet metals. In this article, different numerical models were used to investigate the FLD of tailor-welded blank (TWB). TWBs were CO₂ laser-welded samples of interstitial-free (IF) steel sheets with difference in thickness. The results of the numerical models were compared with the experimental FLD as well as with the empirical model proposed by the North American Deep Drawing Research Group. The emphasis of this investigation is to determine the performance of these different approaches in predicting the FLD. These numerical models for FLD are: second derivative of thinning (SDT), effective strain rate (ESR), major strain rate (MSR), thickness strain rate (TSR), and thickness gradient (TG). Results of this research show necking will be happened, when the value of MSR, TSR, ESR criteria is maximum, TG????0.78 and SDT criterion has the first peak in forming process time. The value of dome height of TWB samples at failure was predicted based on the numerical models for samples with different widths. These numerical predictions were compared with the experimental results. The SDT model indicates a better agreement with experimental results in prediction of both the FLD and the limit dome height (LDH) in comparison to the other numerical models. Both numerical and experimental results show that minimum of LDH is happened in plane strain condition.     

Numerical and Experimental Investigation of Temperature Effect on Thickness Distribution in Warm Hydroforming of Aluminum Tubes

Reduction of weight and increase of corrosion resistance are among the advantageous applications of aluminum alloys in automotive industry. Producing complicated components with several parts as a uniform part not only increases their strength but also decreases the production sequences and costs. However, achieving this purpose requires sufficient formability of the material. Tube hydroforming is an alternative process to produce complex products. In this process, the higher the material formability the more uniform will be the thickness distribution. In this research, tube hydroforming of aluminum alloy (AA1050) at various temperatures has been investigated numerically to study temperature effect on thickness distribution of final product. Also a warm hydroforming set-up has been designed and manufactured to evaluate numerical results. According to numerical and experimental results in the case of free bulging, unlike the constrained bulging, increase of the process temperature causes more uniform thickness distribution and therefore increases the material formability.     

Design and fabrication of a continuous tunable microstrip lowpass filter with wide stopband and sharp response

Tunable microstrip lowpass filters (LPFs) with a good performance are used in most telecommunication systems. In this article, a third‐order circuit with stepped impedance resonator is used to design LPF, and suppression cells are used to improve the performance of the filter in stopbands up to 20 GHz. The ?3 dB cut‐off frequency can be controlled in the range of 0.35 to 1 GHz with the tuning range of 96%. Rejection stopbands of 30 dB (type‐I) and 20 dB (type‐II) can be extended to more than 20 GHz. For the passband, the insertion loss variations are in the range of 0.35 to 2.12 dB. The proposed tunable LPF has a sharp roll‐off rate (42‐96 dB/GHz). The varactor‐tuned microstrip LPF is designed, simulated, fabricated, and measured. The results of simulation and measurement for the proposed LPF are in good agreement. The size of the proposed tunable LPF is small and the filter dimensions are equal to 0.267λg × 0.13λg.     

WaveCluster: a wavelet-based clustering approach for spatial data in very large databases

Many applications require the management of spatial data in a multidimensional feature space. Clustering large spatial databases is an important problem, which tries to find the densely populated regions in the feature space to be used in data mining, knowledge discovery, or efficient information retrieval. A good clustering approach should be efficient and detect clusters of arbitrary shape. It must be insensitive to the noise (outliers) and the order of input data. We propose WaveCluster, a novel clustering approach based on wavelet transforms, which satisfies all the above requirements. Using the multiresolution property of wavelet transforms, we can effectively identify arbitrarily shaped clusters at different degrees of detail. We also demonstrate that WaveCluster is highly efficient in terms of time complexity. Experimental results on very large datasets are presented, which show the efficiency and effectiveness of the proposed approach compared to the other recent clustering methods. Received June 9, 1998 / Accepted July 8, 1999     

Activation of Ziegler‐Natta catalysts by organohalide promoters: A combined experimental and density functional theory study

Several organohalide (HC) compounds were tested as promoters for a Ti‐based Ziegler‐Natta (Z‐N) catalyst at different polymerization conditions. Results show that the intensity of the promoting effect depends on the nature and amount of the promoters. A proper amount, especially optimum amount, of aliphatic type organohalides leads to a strong productivity, and aromatic ones leads to a weak productivity improvement; however, 3‐chloro‐1‐propene poison the catalyst even at lower HC/Ti molar ratios. Among studied compounds, chlorocyclohexane has the best activity promotion effect at HC/Ti molar ratio of 128, and with this as a promoter, the activity increases over 85%. To understand more details about this phenomenon, the rate of polymerization during time and the effect of polymerization conditions (the temperature and hydrogen partial pressure) on the performance of chlorocyclohexane as the most effective promoter were studied. Finally, to explore the mechanism of reactivation of catalyst species by organohalides, molecular modeling was employed and a new oxidation‐addition mechanism was proposed, which basically consisting of homolytic breaking of C? Cl bond in organohalides. It was found that reoxidation of the catalyst, restoring active center, by Cl rich organohalides is energetically more favored. © 2011 Wiley Periodicals, Inc. J Appl Polym Sci, 2012     

Supporting Content-Based Retrieval in Large Image Database Systems

In this paper, we investigate approaches to supporting effective and efficient retrieval of image data based on content. We firstintroduce an effective block-oriented image decomposition structure which can be used to represent image content inimage database systems. We then discuss theapplication of this image data model to content-based image retrieval.Using wavelet transforms to extract image features,significant content features can be extracted from image datathrough decorrelating the data in their pixel format into frequency domain. Feature vectors ofimages can then be constructed. Content-based image retrievalis performed by comparing the feature vectors of the query imageand the decomposed segments in database images.Our experimental analysis illustrates that the proposed block-oriented image representationoffers a novel decomposition structure to be used tofacilitate effective and efficient image retrieval.