Hybrid learning scheme for data mining applications

Classification of large datasets is a challenging task in Data Mining. In the current work, we propose a novel method that compresses the data and classifies the test data directly in its compressed form. The work forms a hybrid learning approach integrating the activities of data abstraction, frequent item generation, compression, classification and use of rough sets.

Mixed pixel analysis and assessment for flood mapping

The most difficult operation in flood inundation mapping using optical flood images is to map the ‘wet’ areas where trees and houses are partly covered by water. This can be referred to as a typical problem of the presence of mixed pixels in the images. A number of automatic information extracting image classification algorithms have been developed over the years for flood mapping using optical remote sensing images, with most labelling a pixel as a particular class. However, they often fail to generate reliable flood inundation mapping because of the presence of mixed pixels in the images. To solve this problem, spectral unmixing methods have been developed. In this thesis, methods for selecting endmembers and the method to model the primary classes for unmixing, the two most important issues in spectral unmixing, are investigated. We conduct comparative studies of three typical spectral unmixing algorithms, Partial Constrained Linear Spectral unmixing, Multiple Endmember Selection Mixture Analysis and spectral unmixing using the Extended Support Vector Machine method. They are analysed and assessed by error analysis in flood mapping using MODIS, Landsat and World View-2 images. The Conventional Root Mean Square Error Assessment is applied to obtain errors for estimated fractions of each primary class. Moreover, a newly developed Fuzzy Error Matrix is used to obtain a clear picture of error distributions at the pixel level. This thesis shows that the Extended Support Vector Machine method is able to provide a more reliable estimation of fractional abundances and allows the use of a complete set of training samples to model a defined pure class. Furthermore, it can be applied to analysis of both pure and mixed pixels to provide integrated hard-soft classification results. Our research also identifies and explores a serious drawback in relation to endmember selections in current spectral unmixing methods which apply fixed sets of endmember classes or pure classes for mixture analysis of every pixel in an entire image. However, as it is not accurate to assume that every pixel in an image must contain all endmember classes, these methods usually cause an over-estimation of the fractional abundances in a particular pixel. In this thesis, a subset of adaptive endmembers in every pixel is derived using the proposed methods to form an endmember index matrix. The experimental results show that using the pixel-dependent endmembers in unmixing significantly improves performance.

Hybrid peptide hairpins containing alpha- and omega-amino acids: Conformational analysis of decapeptides with unsubstituted beta-, gamma-, and delta-residues at positions 3 and 8

The effects of inserting unsubstituted omega-amino acids into the strand segments of model beta-hairpin peptides was investigated by using four synthetic decapeptides, Boc-Lcu-Val-Xxx-Val-D-Pro-Gly-Leu-Xxx-Val-Val- OMe: pepticle 1 (Xxx=Gly), pepticle 2 (Xxx=beta Gly=beta hGly=homoglycine, beta-glycine), pepticle 3 (Xxx=gamma Abu=gamma-aminobutyric acid), pepticle 4 (Xxx= delta Ava=delta-aminovaleric acid). H-1 NMR studies (500 MHz, methanol) reveal several critical cross-strand NOEs, providing evidence for P-hairpin conformations in peptides 2-4. In peptide 3, the NMR results support the formation of the nucleating turn, however, evidence for cross-strand registry is not detected. Single-crystal X-ray diffraction studies of peptide 3 reveal a beta-hairpin conformation for both molecules in the crystallographic asymmetric unit, stabilized by four cross-strand hydrogen bonds, with the gamma Abu residues accommodated within the strands. The D-Pro-Gly segment in both molecules (A,B) adopts a type II' beta-turn conformation. The circular dichroism spectrum for peptide 3 is characterized by a negative CD band at 229 rim, whereas for peptides 2 and 4, the negative band is centered at 225 nm, suggesting a correlation between the orientation of the amide units in the strand segments and the observed CD pattern.

Polypeptide Helices in Hybrid Peptide Sequences

A new class of polypeptide helices in hybrid sequences containing alpha-, beta-, and gamma-residues is described. The molecular conformations in crystals determined for the synthetic peptides Boc-Leu-Phe-Val-Aib-beta Phe-Leu-Phe-Val-OMe 1 (beta Phe: (S)-beta(3)-homophenylalanine) and Boc-Aib-Gpn-AibGpn-OM2(Gpn:1-(aminomethyl)cycl hexaneacetic acid) reveal expanded helical turns in the hybrid sequences (alpha alpha beta)(n) and (ay), In 1, a repetitive helical structure composed Of C-14 hydrogen-bonded units is observed, whereas 2 provides an example of a repetitive C-12 hydrogen-bonded structure. Using experimentally determined backbone torsion angles for the hydrogen-bonded units formed by hybrid sequences, we have generated energetically favorable hybrid helices. Conformational parameters are provided for C-11, C-12, C-13, C-14, and C-15 helices in hybrid sequences.

Multiple conformational states in crystals and in solution in alpha gamma hybrid peptides. Fragility of the C-12 helix in short sequences

The conformational properties of foldamers generated from alpha gamma hybrid peptide sequences have been probed in the model sequence Boc-Aib-Gpn-Aib-Gpn-NHMe. The choice of alpha-aminoisobutyryl (Aib) and gabapentin (Gpn) residues greatly restricts sterically accessible coil formational space. This model sequence was anticipated to be a short segment of the alpha gamma C-12 helix, stabilized by three successive 4 -> 1 hydrogen bonds, corresponding to a backbone-expanded analogue of the alpha polypeptide 3(10)-helix. Unexpectedly, three distinct crystalline polymorphs were characterized in the solid state by X-ray diffraction. In one form, two successive C-12 hydrogen bonds were obtained at the N-terminus, while a novel C-17 hydrogen-bonded gamma alpha gamma turn was observed at the C-terminus. In the other two polymorphs, isolated C-9 and C-7 hydrogen-bonded turns were observed at Gpn (2) and Gpn (4). Isolated C-12 and C-9 turns were also crystallographically established in the peptides Boc-Aib-Gpn-Aib-OMe and Boc-Gpn-Aib-NHMe, respectively. Selective line broadening of NH resonances and the observation of medium range NH(i)<-> NH(i+2) NOEs established the presence of conformational heterogeneity for the tetrapeptide in CDCl3 solution. The NMR results are consistent with the limited population of the continuous C-12 helix conformation. Lengthening of the (alpha gamma)(n) sequences in the nonapeptides Boc-Aib-Gpn-Aib-Gpn-Aib-Gpn-Aib-Gpn-Xxx (Xxx = Aib, Leu) resulted in the observation of all of the sequential NOEs characteristic of an alpha gamma C-12 helix. These results establish that conformational fragility is manifested in short hybrid alpha gamma sequences despite the choice of conformationally constrained residues, while stable helices are formed on chain extension.

Hybrid moving block bootstrap for stochastic simulation of multi-site multi-season streamflows

The Hybrid approach introduced by the authors for at-site modeling of annual and periodic streamflows in earlier works is extended to simulate multi-site multi-season streamflows. It bears significance in integrated river basin planning studies. This hybrid model involves: (i) partial pre-whitening of standardized multi-season streamflows at each site using a parsimonious linear periodic model; (ii) contemporaneous resampling of the resulting residuals with an appropriate block size, using moving block bootstrap (non-parametric, NP) technique; and (iii) post-blackening the bootstrapped innovation series at each site, by adding the corresponding parametric model component for the site, to obtain generated streamflows at each of the sites. It gains significantly by effectively utilizing the merits of both parametric and NP models. It is able to reproduce various statistics, including the dependence relationships at both spatial and temporal levels without using any normalizing transformations and/or adjustment procedures. The potential of the hybrid model in reproducing a wide variety of statistics including the run characteristics, is demonstrated through an application for multi-site streamflow generation in the Upper Cauvery river basin, Southern India. (C) 2004 Elsevier B.V. All rights reserved.

A hybrid computer for X-ray crystallography

A hybrid computer for structure factor calculations in X-ray crystallography is described. The computer can calculate three-dimensional structure factors of up to 24 atoms in a single run and can generate the scatter functions of well over 100 atoms using Vand et al., or Forsyth and Wells approximations. The computer is essentially a digital computer with analog function generators, thus combining to advantage the economic data storage of digital systems and simple computing circuitry of analog systems. The digital part serially selects the data, computes and feeds the arguments into specially developed high precision digital-analog function generators, the outputs of which being d.c. voltages, are further processed by analog circuits and finally the sequential adder, which employs a novel digital voltmeter circuit, converts them back into digital form and accumulates them in a dekatron counter which displays the final result. The computer is also capable of carrying out 1-, 2-, or 3-dimensional Fourier summation, although in this case, the lack of sufficient storage space for the large number of coefficients involved, is a serious limitation at present.

Oncolytic Adenoviruses with Radiation Therapy for Treatment of Prostate Cancer

Prostate cancer is the most common cancer in males. Although many patients with localized disease can be cured with surgery and radiotherapy, advanced disease and especially castration resistant metastatic disease remains incurable, with a median life expectancy of less than 18 months. Oncolytic adenoviruses (Ads) are a new promising treatment against cancer due to their innate capacity to kill cancer cells. Viral replication in tumor cells leads to oncolysis and production of a multiplicity of new virions that are capable of further destroying cancerous tissue. Oncolytic Ads can be modified for tumor targeted infection and replication and be armed with therapeutic transgenes to maximize the oncolytic effect. Worldwide, clinical trials with oncolytic Ads have demonstrated good safety while the antitumor efficacy remains to be improved. Importantly, the best responses have been reported when oncolytic adenoviruses have been combined with standard cancer treatments, such as chemotherapy and radiation. Further, a challenge in many virotherapy approaches has been the monitoring of virus replication in vivo. Reporter genes have been extensively used as transgenes to evaluate the biodistribution of the virus and activity of specific promoters. However, these techniques are often limited to preclinical evaluation and not amenable to human use. The aim of the thesis was to find and develop new oncolytic Ads with maximum efficacy against metastatic, castration resistant prostate cancer and study them in vitro and in vivo combined to different forms of radiation therapy. Using combination therapy, we were aiming for better antitumor efficacy with reduced side effects. Capsid modified Ads for enhanced transduction were studied. Serotype 3 targeted chimera, Ad5/3, was found to have enhanced infectivity for prostate cancer and was used for developing new viruses for the study. Correlation between Ad-encoded marker peptide secretion and simultaneous viral replication was evaluated and the effects of radiotherapy on viral replication were studied in detail. We found that the repair of double strand breaks caused by ionizing radiation was inhibited by adenoviral proteins and led to autophagic cell death. Both subcutaneous models and intrapulmonary tumor models mimicking metastatic, aggressive disease were used in vivo. Virus efficacy was evaluated by intratumoral injections. Also, intravenous administration was evaluated to study the effectiveness in metastatic disease. Oncolytic adenovirus treatment led to significant tumor growth control and increased the survival rate of the mice. These results were further improved when oncolytic Ads were combined with radiation therapy. Oncolytic Ads expressing human sodium/iodide transporter (hNIS) as a transgene were evaluated for their oncolytic potency and for the functionality of hNIS in vitro and in vivo. Monitoring of viral replication was also assessed using different imaging modalities relative to clinical use. SPECT imaging of tumor-bearing mice was evaluated and combined with simultaneous CT-scanning to obtain important anatomical information on biodistribution, also in a three-dimensional form. It was shown that hNIS-expressing adenoviruses could harbour a bi-functional transgene allowing for localization and imaging of viral replication. Targeted radiotherapy was applied by systemic radioiodide administration and resulted in iodide accumulation into Ad-infected tumor. The combination treatment showed significantly enhanced antitumor efficacy in mice bearing prostate cancer tumors. In summary, the results presented above aim to provide new treatment modalities for castration resistant prostate cancer. Molecular insights were provided for better understanding of the benefits of combined radiation therapy and oncolytic adenoviruses, which will hopefully facilitate the translation of the approach into clinical use for humans.

An exact similarity solution in radiation-gas-dynamics

We have found an exact similarity solution of the point explosion problem in the case when the total energy of the shock wave that is produced is not constant but decreases with time and when the loss due to radiation escape is significant. We have compared the results of our exact solution with those of exact numerical solutions of Elliot and Wang and have explained the cause why our solution differs from theirs in certain aspects.

A class of one-dimensional steady-state flows in radiation-gas-dynamics

The study of steady-state flows in radiation-gas-dynamics, when radiation pressure is negligible in comparison with gas pressure, can be reduced to the study of a single first-order ordinary differential equation in particle velocity and radiation pressure. The class of steady flows, determined by the fact that the velocities in two uniform states are real, i.e. the Rankine-Hugoniot points are real, has been discussed in detail in a previous paper by one of us, when the Mach number M of the flow in one of the uniform states (at x=+∞) is greater than one and the flow direction is in the negative direction of the x-axis. In this paper we have discussed the case when M is less than or equal to one and the flow direction is still in the negative direction of the x-axis. We have drawn the various phase planes and the integral curves in each phase plane give various steady flows. We have also discussed the appearance of discontinuities in these flows.

Strong shock with radiation near the surface of a star

The propagation of a shock wave, originating in a stellar interior, is considered when it approaches the surface of the star and assumes a self-similar character, "forgetting" its initial conditions. The flow behind the shock is assumed to be spatially isothermal rather than adiabatic to simulate the conditions of large radiative transfer near the stellar surface. The adiabatic and isothermal flows behind such a shock are compared. The exact shock-propagation laws, obtained by solving the equations in similarity variables, for different values of the parameter δ in the undisturbed density law, ρ0 ∝ xδ, and γ, the ratio of specific heats, are compared with the approximate values calculated by Whitham's characteristic rule and the two show a generally good agreement.

Possible use of self-calibration to reduce systematic uncertainties in determining distance-redshift relation via gravitational radiation from merging binaries

By observing mergers of compact objects, future gravity wave experiments would measure the luminosity distance to a large number of sources to a high precision but not their redshifts. Given the directional sensitivity of an experiment, a fraction of such sources (gold plated) can be identified optically as single objects in the direction of the source. We show that if an approximate distance-redshift relation is known then it is possible to statistically resolve those sources that have multiple galaxies in the beam. We study the feasibility of using gold plated sources to iteratively resolve the unresolved sources, obtain the self-calibrated best possible distance-redshift relation and provide an analytical expression for the accuracy achievable. We derive the lower limit on the total number of sources that is needed to achieve this accuracy through self-calibration. We show that this limit depends exponentially on the beam width and give estimates for various experimental parameters representative of future gravitational wave experiments DECIGO and BBO.

HFinFET: A Scalable, High Performance, Low Leakage Hybrid n-Channel FET

In this letter, we propose the design and simulation study of a novel transistor, called HFinFET, which is a hybrid of an HEMT and a FinFET, to obtain excellent performance and good OFF-state control. Followed by the description of the design, 3-D device simulation has been performed to predict the characteristics of the device. The device has been benchmarked against published state of the art HEMT as well as planar and nonplanar Si n-MOSFET data of comparable gate length using standard benchmarking techniques.