Learning from Geometry

Subspaces and manifolds are two powerful models for high dimensional signals. Subspaces model linear correlation and are a good fit to signals generated by physical systems, such as frontal images of human faces and multiple sources impinging at an antenna array. Manifolds model sources that are not linearly correlated, but where signals are determined by a small number of parameters. Examples are images of human faces under different poses or expressions, and handwritten digits with varying styles. However, there will always be some degree of model mismatch between the subspace or manifold model and the true statistics of the source. This dissertation exploits subspace and manifold models as prior information in various signal processing and machine learning tasks.

A near-low-rank Gaussian mixture model measures proximity to a union of linear or affine subspaces. This simple model can effectively capture the signal distribution when each class is near a subspace. This dissertation studies how the pairwise geometry between these subspaces affects classification performance. When model mismatch is vanishingly small, the probability of misclassification is determined by the product of the sines of the principal angles between subspaces. When the model mismatch is more significant, the probability of misclassification is determined by the sum of the squares of the sines of the principal angles. Reliability of classification is derived in terms of the distribution of signal energy across principal vectors. Larger principal angles lead to smaller classification error, motivating a linear transform that optimizes principal angles. This linear transformation, termed TRAIT, also preserves some specific features in each class, being complementary to a recently developed Low Rank Transform (LRT). Moreover, when the model mismatch is more significant, TRAIT shows superior performance compared to LRT.

The manifold model enforces a constraint on the freedom of data variation. Learning features that are robust to data variation is very important, especially when the size of the training set is small. A learning machine with large numbers of parameters, e.g., deep neural network, can well describe a very complicated data distribution. However, it is also more likely to be sensitive to small perturbations of the data, and to suffer from suffer from degraded performance when generalizing to unseen (test) data.

From the perspective of complexity of function classes, such a learning machine has a huge capacity (complexity), which tends to overfit. The manifold model provides us with a way of regularizing the learning machine, so as to reduce the generalization error, therefore mitigate overfiting. Two different overfiting-preventing approaches are proposed, one from the perspective of data variation, the other from capacity/complexity control. In the first approach, the learning machine is encouraged to make decisions that vary smoothly for data points in local neighborhoods on the manifold. In the second approach, a graph adjacency matrix is derived for the manifold, and the learned features are encouraged to be aligned with the principal components of this adjacency matrix. Experimental results on benchmark datasets are demonstrated, showing an obvious advantage of the proposed approaches when the training set is small.

Stochastic optimization makes it possible to track a slowly varying subspace underlying streaming data. By approximating local neighborhoods using affine subspaces, a slowly varying manifold can be efficiently tracked as well, even with corrupted and noisy data. The more the local neighborhoods, the better the approximation, but the higher the computational complexity. A multiscale approximation scheme is proposed, where the local approximating subspaces are organized in a tree structure. Splitting and merging of the tree nodes then allows efficient control of the number of neighbourhoods. Deviation (of each datum) from the learned model is estimated, yielding a series of statistics for anomaly detection. This framework extends the classical {\em changepoint detection} technique, which only works for one dimensional signals. Simulations and experiments highlight the robustness and efficacy of the proposed approach in detecting an abrupt change in an otherwise slowly varying low-dimensional manifold.

Design, Fabrication, and Characterization of Carbon Nanotube Field Emission Devices for Advanced Applications

Carbon nanotubes (CNTs) have recently emerged as promising candidates for electron field emission (FE) cathodes in integrated FE devices. These nanostructured carbon materials possess exceptional properties and their synthesis can be thoroughly controlled. Their integration into advanced electronic devices, including not only FE cathodes, but sensors, energy storage devices, and circuit components, has seen rapid growth in recent years. The results of the studies presented here demonstrate that the CNT field emitter is an excellent candidate for next generation vacuum microelectronics and related electron emission devices in several advanced applications.

The work presented in this study addresses determining factors that currently confine the performance and application of CNT-FE devices. Characterization studies and improvements to the FE properties of CNTs, along with Micro-Electro-Mechanical Systems (MEMS) design and fabrication, were utilized in achieving these goals. Important performance limiting parameters, including emitter lifetime and failure from poor substrate adhesion, are examined. The compatibility and integration of CNT emitters with the governing MEMS substrate (i.e., polycrystalline silicon), and its impact on these performance limiting parameters, are reported. CNT growth mechanisms and kinetics were investigated and compared to silicon (100) to improve the design of CNT emitter integrated MEMS based electronic devices, specifically in vacuum microelectronic device (VMD) applications.

Improved growth allowed for design and development of novel cold-cathode FE devices utilizing CNT field emitters. A chemical ionization (CI) source based on a CNT-FE electron source was developed and evaluated in a commercial desktop mass spectrometer for explosives trace detection. This work demonstrated the first reported use of a CNT-based ion source capable of collecting CI mass spectra. The CNT-FE source demonstrated low power requirements, pulsing capabilities, and average lifetimes of over 320 hours when operated in constant emission mode under elevated pressures, without sacrificing performance. Additionally, a novel packaged ion source for miniature mass spectrometer applications using CNT emitters, a MEMS based Nier-type geometry, and a Low Temperature Cofired Ceramic (LTCC) 3D scaffold with integrated ion optics were developed and characterized. While previous research has shown other devices capable of collecting ion currents on chip, this LTCC packaged MEMS micro-ion source demonstrated improvements in energy and angular dispersion as well as the ability to direct the ions out of the packaged source and towards a mass analyzer. Simulations and experimental design, fabrication, and characterization were used to make these improvements.

Finally, novel CNT-FE devices were developed to investigate their potential to perform as active circuit elements in VMD circuits. Difficulty integrating devices at micron-scales has hindered the use of vacuum electronic devices in integrated circuits, despite the unique advantages they offer in select applications. Using a combination of particle trajectory simulation and experimental characterization, device performance in an integrated platform was investigated. Solutions to the difficulties in operating multiple devices in close proximity and enhancing electron transmission (i.e., reducing grid loss) are explored in detail. A systematic and iterative process was used to develop isolation structures that reduced crosstalk between neighboring devices from 15% on average, to nearly zero. Innovative geometries and a new operational mode reduced grid loss by nearly threefold, thereby improving transmission of the emitted cathode current to the anode from 25% in initial designs to 70% on average. These performance enhancements are important enablers for larger scale integration and for the realization of complex vacuum microelectronic circuits.

RF MIMO Systems for Wide-Area Indoor Human Motion Monitoring

Human motion monitoring is an important function in numerous applications. In this dissertation, two systems for monitoring motions of multiple human targets in wide-area indoor environments are discussed, both of which use radio frequency (RF) signals to detect, localize, and classify different types of human motion. In the first system, a coherent monostatic multiple-input multiple-output (MIMO) array is used, and a joint spatial-temporal adaptive processing method is developed to resolve micro-Doppler signatures at each location in a wide-area for motion mapping. The downranges are obtained by estimating time-delays from the targets, and the crossranges are obtained by coherently filtering array spatial signals. Motion classification is then applied to each target based on micro-Doppler analysis. In the second system, multiple noncoherent multistatic transmitters (Tx's) and receivers (Rx's) are distributed in a wide-area, and motion mapping is achieved by noncoherently combining bistatic range profiles from multiple Tx-Rx pairs. Also, motion classification is applied to each target by noncoherently combining bistatic micro-Doppler signatures from multiple Tx-Rx pairs. For both systems, simulation and real data results are shown to demonstrate the ability of the proposed methods for monitoring patient repositioning activities for pressure ulcer prevention.

Integrated System Technologies for Modular Trapped Ion Quantum Information Processing

Although trapped ion technology is well-suited for quantum information science, scalability of the system remains one of the main challenges. One of the challenges associated with scaling the ion trap quantum computer is the ability to individually manipulate the increasing number of qubits. Using micro-mirrors fabricated with micro-electromechanical systems (MEMS) technology, laser beams are focused on individual ions in a linear chain and steer the focal point in two dimensions. Multiple single qubit gates are demonstrated on trapped 171Yb+ qubits and the gate performance is characterized using quantum state tomography. The system features negligible crosstalk to neighboring ions (< 3e-4), and switching speeds comparable to typical single qubit gate times (< 2 us). In a separate experiment, photons scattered from the 171Yb+ ion are coupled into an optical fiber with 63% efficiency using a high numerical aperture lens (0.6 NA). The coupled photons are directed to superconducting nanowire single photon detectors (SNSPD), which provide a higher detector efficiency (69%) compared to traditional photomultiplier tubes (35%). The total system photon collection efficiency is increased from 2.2% to 3.4%, which allows for fast state detection of the qubit. For a detection beam intensity of 11 mW/cm2, the average detection time is 23.7 us with 99.885(7)% detection fidelity. The technologies demonstrated in this thesis can be integrated to form a single quantum register with all of the necessary resources to perform local gates as well as high fidelity readout and provide a photon link to other systems.

Evaluation of the Does Reduction Potential Using a Breast Positioning Technique for Organ-based Tube Current Modulated CT Examinations

Purpose: The purpose of this work was to investigate the breast dose saving potential of a breast positioning technique (BP) for thoracic CT examinations with organ-based tube current modulation (OTCM).

Methods: The study included 13 female patient models (XCAT, age range: 27-65 y.o., weight range: 52 to 105.8 kg). Each model was modified to simulate three breast sizes in standard supine geometry. The modeled breasts were further deformed, emulating a BP that would constrain the breasts within 120° anterior tube current (mA) reduction zone. The tube current value of the CT examination was modeled using an attenuation-based program, which reduces the radiation dose to 20% in the anterior region with a corresponding increase to the posterior region. A validated Monte Carlo program was used to estimate organ doses with a typical clinical system (SOMATOM Definition Flash, Siemens Healthcare). The simulated organ doses and organ doses normalized by CTDIvol were compared between attenuation-based tube current modulation (ATCM), OTCM, and OTCM with BP (OTCMBP).

Results: On average, compared to ATCM, OTCM reduced the breast dose by 19.3±4.5%, whereas OTCMBP reduced breast dose by 36.6±6.9% (an additional 21.3±7.3%). The dose saving of OTCMBP was more significant for larger breasts (on average 32, 38, and 44% reduction for 0.5, 1.5, and 2.5 kg breasts, respectively). Compared to ATCM, OTCMBP also reduced thymus and heart dose by 12.1 ± 6.3% and 13.1 ± 5.4%, respectively.

Conclusions: In thoracic CT examinations, OTCM with a breast positioning technique can markedly reduce unnecessary exposure to the radiosensitive organs in the anterior chest wall, specifically breast tissue. The breast dose reduction is more notable for women with larger breasts.

Sparse Sample Detection Using Magnetic Bead Manipulation on a Digital Microfluidic Device

This thesis demonstrates a new way to achieve sparse biological sample detection, which uses magnetic bead manipulation on a digital microfluidic device. Sparse sample detection was made possible through two steps: sparse sample capture and fluorescent signal detection. For the first step, the immunological reaction between antibody and antigen enables the binding between target cells and antibody-­‐‑ coated magnetic beads, hence achieving sample capture. For the second step, fluorescent detection is achieved via fluorescent signal measurement and magnetic bead manipulation. In those two steps, a total of three functions need to work together, namely magnetic beads manipulation, fluorescent signal measurement and immunological binding. The first function is magnetic bead manipulation, and it uses the structure of current-­‐‑carrying wires embedded in the actuation electrode of an electrowetting-­‐‑on-­‐‑dielectric (EWD) device. The current wire structure serves as a microelectromagnet, which is capable of segregating and separating magnetic beads. The device can achieve high segregation efficiency when the wire spacing is 50µμm, and it is also capable of separating two kinds of magnetic beads within a 65µμm distance. The device ensures that the magnetic bead manipulation and the EWD function can be operated simultaneously without introducing additional steps in the fabrication process. Half circle shaped current wires were designed in later devices to concentrate magnetic beads in order to increase the SNR of sample detection. The second function is immunological binding. Immunological reaction kits were selected in order to ensure the compatibility of target cells, magnetic bead function and EWD function. The magnetic bead choice ensures the binding efficiency and survivability of target cells. The magnetic bead selection and binding mechanism used in this work can be applied to a wide variety of samples with a simple switch of the type of antibody. The last function is fluorescent measurement. Fluorescent measurement of sparse samples is made possible of using fluorescent stains and a method to increase SNR. The improved SNR is achieved by target cell concentration and reduced sensing area. Theoretical limitations of the entire sparse sample detection system is as low as 1 Colony Forming Unit/mL (CFU/mL).

Using Image Processing Methods to Improve the Detection of Buried Explosive Threats in GPR Data

Current state of the art techniques for landmine detection in ground penetrating radar (GPR) utilize statistical methods to identify characteristics of a landmine response. This research makes use of 2-D slices of data in which subsurface landmine responses have hyperbolic shapes. Various methods from the field of visual image processing are adapted to the 2-D GPR data, producing superior landmine detection results. This research goes on to develop a physics-based GPR augmentation method motivated by current advances in visual object detection. This GPR specific augmentation is used to mitigate issues caused by insufficient training sets. This work shows that augmentation improves detection performance under training conditions that are normally very difficult. Finally, this work introduces the use of convolutional neural networks as a method to learn feature extraction parameters. These learned convolutional features outperform hand-designed features in GPR detection tasks. This work presents a number of methods, both borrowed from and motivated by the substantial work in visual image processing. The methods developed and presented in this work show an improvement in overall detection performance and introduce a method to improve the robustness of statistical classification.

Gaussian Process Kernels for Cross-Spectrum Analysis in Electrophysiological Time Series

Multi-output Gaussian processes provide a convenient framework for multi-task problems. An illustrative and motivating example of a multi-task problem is multi-region electrophysiological time-series data, where experimentalists are interested in both power and phase coherence between channels. Recently, the spectral mixture (SM) kernel was proposed to model the spectral density of a single task in a Gaussian process framework. This work develops a novel covariance kernel for multiple outputs, called the cross-spectral mixture (CSM) kernel. This new, flexible kernel represents both the power and phase relationship between multiple observation channels. The expressive capabilities of the CSM kernel are demonstrated through implementation of 1) a Bayesian hidden Markov model, where the emission distribution is a multi-output Gaussian process with a CSM covariance kernel, and 2) a Gaussian process factor analysis model, where factor scores represent the utilization of cross-spectral neural circuits. Results are presented for measured multi-region electrophysiological data.

Using Coding to Improve Localization and Backscatter Communication Performance in Low-Power Sensor Networks

Backscatter communication is an emerging wireless technology that recently has gained an increase in attention from both academic and industry circles. The key innovation of the technology is the ability of ultra-low power devices to utilize nearby existing radio signals to communicate. As there is no need to generate their own energetic radio signal, the devices can benefit from a simple design, are very inexpensive and are extremely energy efficient compared with traditional wireless communication. These benefits have made backscatter communication a desirable candidate for distributed wireless sensor network applications with energy constraints.

The backscatter channel presents a unique set of challenges. Unlike a conventional one-way communication (in which the information source is also the energy source), the backscatter channel experiences strong self-interference and spread Doppler clutter that mask the information-bearing (modulated) signal scattered from the device. Both of these sources of interference arise from the scattering of the transmitted signal off of objects, both stationary and moving, in the environment. Additionally, the measurement of the location of the backscatter device is negatively affected by both the clutter and the modulation of the signal return.

This work proposes a channel coding framework for the backscatter channel consisting of a bi-static transmitter/receiver pair and a quasi-cooperative transponder. It proposes to use run-length limited coding to mitigate the background self-interference and spread-Doppler clutter with only a small decrease in communication rate. The proposed method applies to both binary phase-shift keying (BPSK) and quadrature-amplitude modulation (QAM) scheme and provides an increase in rate by up to a factor of two compared with previous methods.

Additionally, this work analyzes the use of frequency modulation and bi-phase waveform coding for the transmitted (interrogating) waveform for high precision range estimation of the transponder location. Compared to previous methods, optimal lower range sidelobes are achieved. Moreover, since both the transmitted (interrogating) waveform coding and transponder communication coding result in instantaneous phase modulation of the signal, cross-interference between localization and communication tasks exists. Phase discriminating algorithm is proposed to make it possible to separate the waveform coding from the communication coding, upon reception, and achieve localization with increased signal energy by up to 3 dB compared with previous reported results.

The joint communication-localization framework also enables a low-complexity receiver design because the same radio is used both for localization and communication.

Simulations comparing the performance of different codes corroborate the theoretical results and offer possible trade-off between information rate and clutter mitigation as well as a trade-off between choice of waveform-channel coding pairs. Experimental results from a brass-board microwave system in an indoor environment are also presented and discussed.

Coding Strategies for X-ray Tomography

This work focuses on the construction and application of coded apertures to compressive X-ray tomography. Coded apertures can be made in a number of ways, each method having an impact on system background and signal contrast. Methods of constructing coded apertures for structuring X-ray illumination and scatter are compared and analyzed. Apertures can create structured X-ray bundles that investigate specific sets of object voxels. The tailored bundles of rays form a code (or pattern) and are later estimated through computational inversion. Structured illumination can be used to subsample object voxels and make inversion feasible for low dose computed tomography (CT) systems, or it can be used to reduce background in limited angle CT systems.

On the detection side, coded apertures modulate X-ray scatter signals to determine the position and radiance of scatter points. By forming object dependent projections in measurement space, coded apertures multiplex modulated scatter signals onto a detector. The multiplexed signals can be inverted with knowledge of the code pattern and system geometry. This work shows two systems capable of determining object position and type in a 2D plane, by illuminating objects with an X-ray `fan beam,' using coded apertures and compressive measurements. Scatter tomography can help identify materials in security and medicine that may be ambiguous with transmission tomography alone.

Micro-Fabricated Surface Trap and Cavity Integration for Trapped Ion Quantum Computing

Atomic ions trapped in micro-fabricated surface traps can be utilized as a physical platform with which to build a quantum computer. They possess many of the desirable qualities of such a device, including high fidelity state preparation and readout, universal logic gates, long coherence times, and can be readily entangled with each other through photonic interconnects. The use of optical cavities integrated with trapped ion qubits as a photonic interface presents the possibility for order of magnitude improvements in performance in several key areas of their use in quantum computation. The first part of this thesis describes the design and fabrication of a novel surface trap for integration with an optical cavity. The trap is custom made on a highly reflective mirror surface and includes the capability of moving the ion trap location along all three trap axes with nanometer scale precision. The second part of this thesis demonstrates the suitability of small micro-cavities formed from laser ablated fused silica substrates with radii of curvature in the 300-500 micron range for use with the mirror trap as part of an integrated ion trap cavity system. Quantum computing applications for such a system include dramatic improvements in the photonic entanglement rate up to 10 kHz, the qubit measurement time down to 1 microsecond, and the measurement error rates down to the 10e-5 range. The final part of this thesis details a performance simulator for exploring the physical resource requirements and performance demands to scale such a quantum computer to sizes capable of performing quantum algorithms beyond the limits of classical computation.

A Molecular-scale Programmable Stochastic Process Based On Resonance Energy Transfer Networks: Modeling And Applications

While molecular and cellular processes are often modeled as stochastic processes, such as Brownian motion, chemical reaction networks and gene regulatory networks, there are few attempts to program a molecular-scale process to physically implement stochastic processes. DNA has been used as a substrate for programming molecular interactions, but its applications are restricted to deterministic functions and unfavorable properties such as slow processing, thermal annealing, aqueous solvents and difficult readout limit them to proof-of-concept purposes. To date, whether there exists a molecular process that can be programmed to implement stochastic processes for practical applications remains unknown.

In this dissertation, a fully specified Resonance Energy Transfer (RET) network between chromophores is accurately fabricated via DNA self-assembly, and the exciton dynamics in the RET network physically implement a stochastic process, specifically a continuous-time Markov chain (CTMC), which has a direct mapping to the physical geometry of the chromophore network. Excited by a light source, a RET network generates random samples in the temporal domain in the form of fluorescence photons which can be detected by a photon detector. The intrinsic sampling distribution of a RET network is derived as a phase-type distribution configured by its CTMC model. The conclusion is that the exciton dynamics in a RET network implement a general and important class of stochastic processes that can be directly and accurately programmed and used for practical applications of photonics and optoelectronics. Different approaches to using RET networks exist with vast potential applications. As an entropy source that can directly generate samples from virtually arbitrary distributions, RET networks can benefit applications that rely on generating random samples such as 1) fluorescent taggants and 2) stochastic computing.

By using RET networks between chromophores to implement fluorescent taggants with temporally coded signatures, the taggant design is not constrained by resolvable dyes and has a significantly larger coding capacity than spectrally or lifetime coded fluorescent taggants. Meanwhile, the taggant detection process becomes highly efficient, and the Maximum Likelihood Estimation (MLE) based taggant identification guarantees high accuracy even with only a few hundred detected photons.

Meanwhile, RET-based sampling units (RSU) can be constructed to accelerate probabilistic algorithms for wide applications in machine learning and data analytics. Because probabilistic algorithms often rely on iteratively sampling from parameterized distributions, they can be inefficient in practice on the deterministic hardware traditional computers use, especially for high-dimensional and complex problems. As an efficient universal sampling unit, the proposed RSU can be integrated into a processor / GPU as specialized functional units or organized as a discrete accelerator to bring substantial speedups and power savings.

Peer Mentoring and Professionalism

Peer mentoring has been a success for everyone involved resulting in a ‘win-win-win’ situation for mentors, mentees and university schools and departments (Andrews and Clark, 2011). Mentors have the opportunity to develop key transferable skills such as communication and leadership, which in turn can enhance their employability opportunities. There is also potential to increase and develop social and academic confidence. For mentees the benefits include the opportunity to gain advice, encouragement and support during the transition period from school/college/work to university along with the opportunity to gain an insight into the stages of university life by learning the "rules of the game". Through peer mentor schemes University schools and departments are meeting the demand to support student success while assisting student transition and reducing attrition. This paper will focus on the peer mentor scheme set up in the School of Electronics, Electrical Engineering and Computer Science at Queen’s University Belfast specifically the development of employability skills through company involvement in the scheme.

How Good Am I? Crude Self-Assessment vs. Examination

With the new academic year structure encouraging more in-term assessment to replace end-of-year examinations one of the problems we face is assessing students and keeping track of individual student learning without overloading the students and staff with excessive assessment burdens.
In the School of Electronics, Electrical Engineering and Computer Science, we have constructed a system that allows students to self-assess their capability on a simple Yes/No/Don’t Know scale against fine grained learning outcomes for a module. As the term progresses students update their record as appropriately, including selecting a Learnt option to reflect improvements they have gained as part of their studies.
In the system each of the learning outcomes are linked to the relevant teaching session (lectures and labs) and to online resources that students can access at any time. Students can structure their own learning experience to their needs and preferences in order to attain the learning outcomes.
The system keeps a history of the student’s record, allowing the lecturer to observe how the students’ abilities progress over the term and to compare it to assessment results. The system also keeps of any of the resource links that student has clicked on and the related learning outcome.
The initial work is comparing the accuracy of the student self-assessments with their performance in the related questions in the traditional end-of-year examination.

Detecting stealthy attacks: Efficient monitoring of suspicious activities on computer networks

Stealthy attackers move patiently through computer networks - taking days, weeks or months to accomplish their objectives in order to avoid detection. As networks scale up in size and speed, monitoring for such attack attempts is increasingly a challenge. This paper presents an efficient monitoring technique for stealthy attacks. It investigates the feasibility of proposed method under number of different test cases and examines how design of the network affects the detection. A methodological way for tracing anonymous stealthy activities to their approximate sources is also presented. The Bayesian fusion along with traffic sampling is employed as a data reduction method. The proposed method has the ability to monitor stealthy activities using 10-20% size sampling rates without degrading the quality of detection.