4
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Archived Theses
MIMO for 5th generation heterogeneous communication networks: Modeling and performance analysis
The day-to-day increasing demand of high data rate and connectivity across all of the devices has lead us to the verge of remanipulating the current wireless deployment for the future generation networks. Current 4G deployment either uses small heterogeneous cellular networks(HetNets) or multi-antenna approach to fulfill the demand. Amalgamate of these two features can lead to a significant improvement. This piece of work is a contribution to the analysis of this combined network paradigm. This dissertation proposes a tractable analysis for use of MIMO in HetNets. A downlink analysis by deploying K-tier BSs, with different transmit power and geographical density across all k-tiers is studied. Along with it, localization of different class of BSs is carried out through well behaved independent Point Process, namely Poisson Point Process(PPP). All of this baseline model setup allow the derivation of analytical expression for coverage probability, average rate, signal to interference and noise ratio(SINR) and area spectral efficiency. Insightful results obtained shows that increasing the BSs does not necessarily increase the coverage probability, but it increases by adding less loaded small BSs. For the MIMO based HetNets with a fair spatial model, a tractable framework shows that spreading the many small SISO systems is preferable over fewer MIMO BSs. But with large number of independent SISO systems, it is more preferable to use MIMO BSs with same number of antennas.
Integrative clustering approaches for cancer subtype discovery
Cancer is a heterogeneous disease and identification of cancer subtypes is critical for personalized treatment and drug development. Recently, cancer genome projects have produced multiple types of high-throughput data for thousands of cancer patients. Exploiting the complementary information between different data types can improve finding subtypes. In the first part of this thesis, we apply multi-view kernel k-means to integrate multiple genomic datasets (i.e.,gene expression, DNA methylation and miRNA expression) on two cancer datasets. We show that combining kernels (i.e., that correspond to different views) with learned weights give better clusters compared to combining the kernels uniformly or using each data set independently. We also demonstrate an improved performance compared to existing models that integrate the same data types. In terms of biological significance, Kaplan-Meier analysis shows that our discovered clusters have distinct survival profiles with statistically significant log-rank test p-values. In the second part, we target the high dimensionality problem of genomic datasets by extending a single-view sparse k-means framework to multi-view setting. This extension allows us to perform integrative clustering and feature selection simultaneously by learning both feature weights and view weights. We confirm that performing feature selection improves the clusters for the majority of the datasets. Altogether, our results indicate that integration of multiple genomic characterizations and the application of feature selection enable the discovery of subtypes that improve over current patient stratifications.
A comparative study of kalman filter and its variants in acoustic echo cancellation
In telecommunication, when a far-end speaker's voice is reflected in the near-end, this reflection causes communication problems and it is known as acoustic echo. Despite its wide coverage in the literature, acoustic echo cancellation (AEC) has been mostly examined for linear systems. The number of approaches to mitigate echo in non-linear systems is not sufficient. In this thesis, Kalman filter and its variants such as Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF) and Cubature Kalman Filter (CKF) have been compared for this purpose.
Some machine learning techniques for medical diagnosis
There is a huge increase in the amount of biological data obtained with the great breakthroughs in technology, and the amount of this data increases exponentially every day. The use of data mining algorithms and machine learning methods has become widespread in the field of health, because analysis and interpretation of biological data is very difficult with traditional methods. In this study, biochemistry and hormone values are used to make some prediction on medical diagnosis with some machine learning methods such as linear regression, generalized linear regression, deep learning, random forests, gradient boosted trees and support vector machines. With the analysis of the laboratory results according to the reference ranges, it is tried to reveal some new relations of biochemistry and hormone parameters. Finally, the results are presented and the performance of the model is evaluated.