This work requires comprehensive understanding of actuation and sensing mechanisms of transducers, and use of design tools for the optimization of key performance parameters.

This work requires thorough analysis of wafer-bonding techniques including but not limited to Hertzian contact for bonding dynamics between surfaces.

The aim of the project is to study robust Capon beamforming with diagonal loading. Beamforming aims to capture the component of an incoming wave propagating in a certain direction. The incoming wave is sensed with an array of sensors. The sensor outputs are processed to spatially filter propagating waves in different directions. The operation is similar to low-pass/high-pass filtering where the incoming signal is filtered to enhance certain bands in the spectrum, while suppressing some others. Different from the filtering problem, the input to is composed of very few samples and the accuracy of the beamformer output is very critical in many applications.

This project does not involve any physical implementation, it is conducted on MATLAB. Only one group of students will be admitted (due to lack of time of advisor, sorry!). The students should be **very familiar** with linear algebra (especially the eigendecomposition, orthogonal representations, inner products) and optimization (Lagrange multipliers, gradients and other Calculus topics related with optimization). A basic familiarity with the physics of the problem is useful. This is a highly mathematical project, students enjoying maths are welcomed to apply.

Keywords: Beamforming, Capon Beamformer, MVDR, Robust Beamforming.

Classification is an important problem of machine learning, which is relevant to quite many applications such as face recognition, handwriting analysis, diagnosis of diseases from tissue samples, etc. In machine learning, a classification rule is learned from training data, which is then applied to classify test data. However, in quite many applications the distribution of training data differs from that of test data, which degrades the performance of traditional classification methods. In this project, we will investigate machine learning algorithms that can transfer the knowledge acquired from training data to test data even in settings where these have different distributions.

Summary: Ultra-wide band electromagnetic signals will be numerically simulated by using an open-source GPR simulator. A GPR A-scan signal database will be established for designated solid targets buried in soil. Detection and an introductory level classification of targets are aimed.

Summary: Ultra-wide band electromagnetic signals will be numerically simulated by using an open-source GPR simulator. A GPR A-scan signal database will be established for designated underground volumes filled by liquids with or without plastic boundaries. Detection and an introductory level classification of liquids are aimed.

Brain-like computing is an important area in Artificial Intelligence, Machine Learning and Neurotechnology. Today brain-like computing systems has gained much importance due to diffculties that the conventional computers and programming faced in terms of performance, computation time and information processing style in areas like pattern recognition, machine learning, computer vision, big data or brain computer interfaces. Size and energy consumption is another problem that super computers are faced. Deep Learning, Artificial Neural Networks, Spking Neural Networks, Neuromorphic Computing are approaches all inspired from brain and its processing elements which are called neurons. Although small sized neural networks can be implemented on conventional computers, for medium sized networks GPU programming (CUDA) is used. For large networks or big data even supercomputers are employed. Today there is a tendency for devoloping neuromorphic chips aiming to implement millions of neurons in a small volume with low power consumption. Brain-like computing is also important for constructing more powerful interfaces between machines and human brain/nervous system.