155 sonuçlar
Arama Sonuçları
Listeleniyor 1 - 10 / 155
Yayın Calculating the VC-dimension of decision trees(IEEE, 2009) Aslan, Özlem; Yıldız, Olcay Taner; Alpaydın, Ahmet İbrahim EthemWe propose an exhaustive search algorithm that calculates the VC-dimension of univariate decision trees with binary features. The VC-dimension of the univariate decision tree with binary features depends on (i) the VC-dimension values of the left and right subtrees, (ii) the number of inputs, and (iii) the number of nodes in the tree. From a training set of example trees whose VC-dimensions are calculated by exhaustive search, we fit a general regressor to estimate the VC-dimension of any binary tree. These VC-dimension estimates are then used to get VC-generalization bounds for complexity control using SRM in decision trees, i.e., pruning. Our simulation results shows that SRM-pruning using the estimated VC-dimensions finds trees that are as accurate as those pruned using cross-validation.Yayın On the effect of random sampling jitter on cyclostationarity based spectrum sensing algorithms for cognitive radio(IEEE, 2009) Öner, Mustafa MengüçCognitive radio is an enabling technology, which is expected to lead to a more efficient utilization of the available spectral resources due to its flexibility and its ability to sense its spectral environment. Recently, spectrum sensing methods based on exploiting the cyclostationary characteristics of communication signals have been drawing considerable interest. Imperfections in the cognitive radio receiver that affect the cyclic statistics of a signal of interest may lead to a degradation in the performance of spectrum sensing algorithms based on cyclostationarity. One such typical source of imperfection is random timing jitter in the sampling process. In this work, we explore the effect of random sampling jitter on the second order cyclostationary statistics of wide sense cyclostationary signals. General analytical expressions are derived for the cyclic statistics of sampled signals in the presence of sampling jitter and specific results are provided for two cases of interest. Subsequently, the effect of the jitter on a spectrum sensing algorithm is investigated via simulations.Yayın Phased subarray imaging for low-cost, wideband coherent array imaging(IEEE, 2003) Johnson, Jeremy A.; Oralkan, Ömer; Ergün, Arif Sanlı; Demirci, Utkan; Karaman, Mustafa; Khuri-Yakub, Butrus ThomasThe front-end hardware complexity of conventional full phased array (FPA) imaging is proportional to the number of array elements. Phased subarray (PSA) imaging has been proposed as a method of reducing the hardware complexity-and therefore system cost and size-while achieving near-FPA image quality. A new method is presented for designing the subarray-dependent interpolation filters suitable for wideband PSA imaging. The method was tested experimentally using pulse-echo data of a wire target phantom acquired using a 3.2-cm. 128-element capacitive micromachined ultrasonic transducer (CMUT) array with 85% fractional bandwidth at 3 MHz. A specific PSA configuration using seven 32-element subarrays was compared to FPA imaging, representing a 4-fold reduction in front-end hardware complexity and a 43% decrease in frame rate. For targets near the fixed transmit focal distance, the mean 6-dB lateral resolution was identical to that of FPA, the axial resolution improved by 4%, and the SNR decreased by 5 dB. Measurements were repeated for 10 different PSA configurations with subarray sizes ranging from 4 to 60. The lateral and axial resolutions did not vary significantly with subarray size; both the SNR and contrast-to-noise ratio (CNR) improved with increased subarray size.Yayın A function of direction in a Weyl subspace associated with a set of orthogonal vector fields(World Scientific Publ Co Pte Ltd, 2003) Özdeğer, AbdülkadirLet W-m be an m-dimensional subspace of an n-dimensional Weyl space W-n. Suppose that (1)v, (2)v,..., (m)v are mutually orthogonal smooth vector fields in W-m and that v is a non-tangential smooth vector field defined on W-m. Consider the (m + 1)-dimensional net delta defined by (1)v, (2)v, ..., (m)v, v. In this work, we obtain a function of direction associated with the net delta and define a class of curves on W-m in relation to this function of direction.Yayın Integrated ultrasonic imaging systems based on CMUT arrays: Recent progress(IEEE, 2004) Wygant, Ira O.; Zhuang, Xuefeng; Yeh, David T.; Nikoozadeh, Amin; Oralkan, Ömer; Ergün, Arif Sanlı; Karaman, Mustafa; Khuri-Yakub, Butrus ThomasThis paper describes the development of an ultrasonic imaging system based on a two-dimensional capacitive micromachined ultrasonic transducer (CMUT) array. The transducer array and front-end electronics are designed to fit in a 5-mm endoscopic channel. A custom-designed integrated circuit, which comprises the front-end electronics, will be connected with the transducer elements via through-wafer interconnects and flip-chip bonding. FPGA-based signal-processing hardware will provide real-time three-dimensional imaging. The imaging system is being developed to demonstrate a means of integrating the front-end electronics with the transducer array and to provide a clinically useful technology. Integration of the electronics can improve signal-to-noise ratio, reduce the number of cables connecting the imaging probe to a separate processing unit, and provide a means of connecting electronics to large two-dimensional transducer arrays. This paper describes the imaging system architecture and the progress we have made on implementing each of its components: a 16×16 CMUT array, custom-designed integrated circuits, a flip-chip bonding technique, and signal-processing hardware.Yayın Co-array optimization of CMUT arrays for forward-looking IVUS(IEEE, 2009-09) Tekeş, Coşkun; Karaman, Mustafa; Değertekin, Fahrettin LeventThe ring annular array structure is a preferred configuration for implementing Forward-Looking IVUS (FLIVUS) catheters as it allows for volumetric imaging as well as use of a guidewire at the center. CMUT technology is promising for these arrays especially with the flexibility of locating array elements on the circular donut area efficiently. To take advantage of this flexibility, in this study, we introduce a new co-array sampling strategy that improves imaging performance while keeping the number of firings at a level suitable for real-time imaging. The presented co-array sampling strategy is based on the idea of adjusting the element density of the co-array rings in radial direction to suppress side lobes. In non-uniform sampling of the co-array with a given number of firings, the inter-element distances are adjusted both in radial and angular direction from inner ring to outer ring to fit a given apodization function. To test the imaging performance, we performed numerical simulations of the co-array with non-uniform sampling fitting to the raised-cosine apodization. The simulation results shows that sidelobe level can be reduced more than 10 dB by using non-uniform co-array sampling. This approach does use uniform weighting of all Tx and Rx elements in beamforming, and hence does not cause any SNR loss for apodization.Yayın Spectral coding of mesh geometry with a hierarchical set partitioning algorithm(Spie-Int Soc Optical Engineering, 2008) Konur, Umut; Bayazıt, Uluğ; Ateş, Hasan Fehmi; Gürgen, Sadık FikretThis work proposes a progressive mesh geometry coder, which expresses geometry information in terms of spectral coefficients obtained through a transformation and codes these coefficients using a hierarchical set partitioning algorithm that assigns right priorities to those coefficients at all bit planes. The spectral transformation used is the one proposed in [8] where the spectral coefficients are obtained by projecting the mesh geometry on an orthonormal basis determined by mesh topology. The set partitioning method used in coding, treats spectral coefficients belonging to the three spatial coordinates with the right priority at all bit planes and realizes a truly embedded system by achieving implicit bit allocation via joint coding the zeroes of coefficients at the bit planes. The experiments performed on common irregular meshes reveal that the rate-distortion performance of the coder is significantly superior to the coding system proposed in [8].Yayın Theoretical calculation of the kinetic coefficient of normal crystal growth(Trans Tech Publications Ltd, 2004) Dimitrov, Ventzislav IvanovAn expression for the velocity u of migration of a diffuse simple crystal-melt interface has been derived on the basis of the theory of atomic mobility in supercooled liquids: u = K-0 (T / T-m) DeltaT, where DeltaT = T-m - T the undercooling below the melting point T-m; K-0 is the kinetic coefficient of atomic attachment, which is used in models of crystal growth. It has been calculated for a number of metals. u(max) = K0Tm / 4 is the theoretical limit of the velocity of crystal growth. For a number of FCC metals the theoretical limit of crystal growth has been found to be of order of 200 m/s. The crystal growth kinetics has been shown to be limited by the atomic self-diffusion in the interface, for which the strong dependence on the orientation of the crystal/melt interface has been explained.Yayın A designing practice and two coding practices for extreme programming (XP)(Springer Verlag, 2003) Yıldız, Mustafa; Kuru, SelahattinThis paper introduces three new XP practices and reports the experience of applying them to web based software development. These are issue- based programming, comment-first coding and just in time code ownership. The example project is development of an on-line student information and registration software for a university.Yayın Driver recognition using gaussian mixture models and decision fusion techniques(Springer-Verlag Berlin, 2008) Benli, Kristin Surpuhi; Düzağaç, Remzi; Eskil, Mustafa TanerIn this paper we present our research in driver recognition. The goal of this study is to investigate the performance of different classifier fusion techniques in a driver recognition scenario. We are using solely driving behavior signals such as break and accelerator pedal pressure, engine RPM, vehicle speed; steering wheel angle for identifying the driver identities. We modeled each driver using Gaussian Mixture Models, obtained posterior probabilities of identities and combined these scores using different fixed mid trainable (adaptive) fusion methods. We observed error rates is low as 0.35% in recognition of 100 drivers using trainable combiners. We conclude that the fusion of multi-modal classifier results is very successful in biometric recognition of a person in a car setting.
