The availability of large gene expression microarray data has brought along many challenges for biological data mining. Many different clustering methods have been proposed and widely used to analyze gene expression data. The underlying concept allows to identify sets of genes sharing similar expression patterns across subsets of samples, and its usefulness has been demonstrated for different organisms and data sets. Currently, there are several biclustering methods that use different techniques; however, it is not clear how to compare the resulted biclusters with respect to biological relevance. So far, there are no available guidelines for choosing a biclustering technique from available ones. In this work, we propose two new Mean Squared Residue (MSR) based biclustering methods. The first method is a dual biclustering algorithm which finds a set of biclusters using a greedy approach. The second method combines dual biclustering algorithm with quadratic programming. The dual biclustering algorithm reduces the size of the matrix, so that the quadratic program can find an optimal bicluster reasonably fast. We also describe the comparison method, explain how we handle bicluster’s overlap and how we treat missing data.

A 16-level instrument in which the input voltage is sampled and quantised to yield direct decimal readouts of 1/T∫v^2dt . and T is described. This is an improved version of an instrument previously constructed; The upper frequency limit has been extended from 5 Hz to about 2 kHz. The readout of the instrument can be in any code, the decimal code being implemented in the instrument described. The original error analysis has been extended. An extensive analysis of the overall error characteristics was carried out theoretically and the results were confirmed experimentally. The instrument is capable of measuring the mean square value of periodic waveform to within 2%. Normal distribution noise of standard deviation between 1V and 3V can be measured with similar accuracy. The accuracy and the upper frequency limit are determined by the ' aperture time'of the sampling process. The errors arising in the sampling instrument depend on the number of levels used. / Thesis / Master of Engineering (MEngr)

The presented thesis investigates the identification of burst locations in water distribution systems (WDS) by analysis of field and simulation experimental data. This required the development of a new hybrid method of burst detection and sizing, and also a burst location identification algorithm. Generally, existing practice relies on a combination of some simple procedure and experience of the involved staff and cannot be easily automated. The practical methods are based on direct manifestation of burst on the surface or on systematically surveying suspected areas e.g. by using listening sticks, such methods are very time consuming. The proposed burst location algorithm is based on comparing data by means of statistical analysis of field data with results of water network simulation. An extended network hydraulic simulator is used to model pressure dependent leakage terms. The presence of a burst changes the flow pattern and also pressure at network nodes which may be used to estimate the burst size and its location. The influence of such random factors as demand flows and background leakage on the process of burst detection is also considered. The field data is from a generalised fixed area and variable area (FAVOR) test where inlet pressure is being stepped up and down and the following variables are measured: inlet flow, inlet pressure (head) and pressure at a number of selected sensitive nodes. The method has three stages and uses two different models, one is inlet flow model (IFM) to represent the total inlet flow and another is the extended hydraulic model to simulate different burst locations. Initially the presence of a potential burst is investigated. If this is confirmed precise values of the demand, background leakage flow and burst flow in IFM are subsequently estimated. They are used to identify the burst site at the third stage of the method. The method can be easily adapted for practical use. It requires data from experiments carried out at night between 1am and 5am and involves placing typically about 20 temporary loggers to collect the measurements during this period. It also requires the availability of a hydraulic model which normally is in the possession of a water company. The program has been implemented in the Matlab package and is easy to use. The current methodology is tuned to identify a single burst but can be generalised to identify locations of multiple bursts.

No description available.

Nonlinear parametric response of a flexible cantilever beam is simulated. In the simulations, lateral response of the beam due to an imposed axial harmonic base displacement excitation is calculated. The response frequency is approximately half the input frequency. The transient simulations include the assumption of damping proportional to the square of the velocity along the beam. Velocity-squared damping is realistic for situations in which fluid forces resisting the structural motion are significant. The commercial finite element software, ANSYS, is used to perform the simulations. A flexible method is developed and implemented in this work, based on the ANSYS Parametric Design Language, for including the quadratic damping assumption in the analysis. Variation of steady state response amplitude is examined for a range of quadratic damping coefficients over a range of axial base excitation frequencies. Further, a definition of phase angle of the response with the respect to the input is proposed for these nonlinear cases in which the input frequency is an integer multiple of the response frequency. The response phase with respect to excitation is studied over a range of damping coefficients and excitation frequencies. In addition, numerical solutions of nonlinear dynamic systems obtained from the implicit finite element method and the explicit dynamics finite element method are compared. The nonlinear dynamic systems considered are a flexible beam subjected to axial base excitation and also lateral excitations. The studies comparing explicit and implicit method results include cases of stress-stiffening and large deflections.

No description available.

<p>It has been observed by researchers that periodic auditory stimuli can cause the activities in different brain areas to be periodically synchronized. Fast auditory stimuli have been shown to cause the brain sources to synchronize at the rate of stimuli. Brain sources respond to them not only by increase in local synchronization, but also in the global synchronization of cortical regions often regarded as functional connectivity. Spectral power and coherence are often used to characterize such neural synchronization. Beta band oscillations have been reported to underlie the neural mechanism during repetitive auditory stimuli. Cortical generators of these underlying beta oscillations were investigated in several studies based on MEG measurements. This research is intended to investigate (1) EEG can be used to detect and localize neural sources changing in power and coherence and (2) beta oscillations underlie such neural synchronization during fast repetitive auditory stimuli based on EEG measurements. The procedure of this study consists of several steps. First, the minimum variance (MV) scalar beamformer, an adaptive spatial filter, is used to estimate the temporal signals in the brain source space, given EEG recordings. The analysis of the estimated source temporal signals then consists of two stages firstly the power analysis and secondly the coherence analysis. The dynamics of power and coherence is investigated instantaneously over time and in the lower beta frequency band [14,20Hz]. This is done by detecting the most prominent changes in the two spectral parameters through singular value decomposition (SVD). Two coherence measures imaginary component (IC) and magnitude-squared coherence (MSC) are employed and compared in terms of their performance both mathematically and experimentally. In the simulations, we show the capability of using EEG to detect and localize power co-variations and dynamic functional connectivity in the cortical regions. We also perform the procedure on the recorded real data from subjects passively listening to rhythmic auditory stimuli. Beta oscillations are found to underlie the neural activity to percept auditory stimuli. This is shown by localization of auditory cortices and detection of power co-variation in this frequency band. We demonstrate the feasibility of using EEG to identify coupled and co-activated brain sources similar to those obtained from MEG signals in the previous studies. These include auditory and motor regions which were found to be functionally coherent and have a functional role in the auditory perception. The superiority of IC over MSC measure is proven mathematically and validated in both simulations and real data experiments.</p> / Master of Applied Science (MASc)

No / We have studied the performance of multidimensional signaling techniques using wavelets based modulation within an orthogonally multiplexed communication system. The discrete wavelets transform and wavelet packet modulation techniques have been studied using Daubechies 2 and 8, Biothogonal1.5 and 3.1 and reverse Biorthognal 1.5 and 3.1 wavelets in the presence of Rayleigh multipath fading channels with AWGN. Results showed that DWT based systems outperform WPM systems both in terms of BER vs. SNR performance as well as processing. The performances of two different equalizations techniques, namely zero forcing (ZF) and minimum mean square error (MMSE), were also compared using DWT. When the channel is modeled using Rayleigh multipath fading, AWGN and ISI both techniques yield similar performance.

A key challenge for maximizing the world’s food supply is crop yield prediction. In this study, three machine models are used to predict the fresh weight (yield) of lettuce plants that are grown inside indoor farms hydroponically using the vertical farming infrastructure, namely, support vector regressor (SVR), random forest regressor (RFR), and deep neural network (DNN).The climate data, nutrient data, and plant growth data are passed as input to train the models to understand the growth pattern based on the available features. The study of evaluation metrics majorly covers Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), R-squared, and Adjusted R-squared values.The results of the project have shown that the Random Forest with all the features is the best model having the best results with the least cross-validated MAE score and good cross-validated Adjusted R-squared value considering that the error of the prediction is minimal. This is followed by the DNN model with minor differences in the resulting values. The Support Vector Regressor (SVR) model gave a very poor performance with a huge error value that cannot be afforded in this scenario. In this study, we have also compared various evaluating metrics mentioned above and considered the cross-validated MAE and cross-validated Adjusted R-squared metrics. According to our study, MAE had the lowest error value, which is less sensitive to the outliers and adjusted R-squared value helps to understand the variance of the target variable with the predictor variable and adjust the metric to prevent the issues of overfitting.

ITC/USA 2007 Conference Proceedings / The Forty-Third Annual International Telemetering Conference and Technical Exhibition / October 22-25, 2007 / Riviera Hotel & Convention Center, Las Vegas, Nevada / The accuracy of channel estimation plays a crucial role in the demodulation of data symbols sent across an unknown wireless medium. In this work a new analytical expression for the channel estimation error of a multiple input multiple output (MIMO) system is obtained when the wireless medium is continuously changing in the temporal domain. Numerical examples are provided to illustrate our findings.