Operational efficiency of solar energy farms requires detailed analytics and information on each panel regarding voltage, current, temperature, and irradiance. Monitoring utility-scale solar arrays was shown to minimize the cost of maintenance and help optimize the performance of photovoltaic (PV) arrays under various conditions. This dissertation describes a project that focuses on the development of machine learning and neural network algorithms. It also describes an 18kW solar array testbed for the purpose of PV monitoring and control. The use of the 18kW Sensor Signal and Information Processing (SenSIP) PV testbed which consists of 104 modules fitted with smart monitoring devices (SMDs) is described in detail. Each of the SMDs has embedded, a wireless transceiver, and relays that enable continuous monitoring, fault detection, and real-time connection topology changes. Data is obtained in real time using the SenSIP PV testbed. Machine learning and neural network algorithms for PV fault classification is are studied in depth. More specifically, the development of a series of customized neural networks for detection and classification of solar array faults that include soiling, shading, degradation, short circuits and standard test conditions is considered. The evaluation of fault detection and classification methods using metrics such as accuracy, confusion matrices, and the Risk Priority Number (RPN) is performed. The examination and assessment the classification performance of customized neural networks with dropout regularizers is presented in detail. The development and evaluation of neural network pruning strategies and illustration of the trade-off between fault classification model accuracy and algorithm complexity is studied. This study includes data from the National Renewable Energy Laboratory (NREL) database and also real-time data collected from the SenSIP testbed at MTW under various loading and shading conditions. The overall approach for detection and classification promises to elevate the performance and robustness of PV arrays.

Dealing with relational data structures is central to a wide-range of applications including social networks, epidemic modeling, molecular chemistry, medicine, energy distribution, and transportation. Machine learning models that can exploit the inherent structural/relational bias in the graph structured data have gained prominence in recent times. A recurring idea that appears in all approaches is to encode the nodes in the graph (or the entire graph) as low-dimensional vectors also known as embeddings, prior to carrying out downstream task-specific learning. It is crucial to eliminate hand-crafted features and instead directly incorporate the structural inductive bias into the deep learning architectures. In this dissertation, deep learning models that directly operate on graph structured data are proposed for effective representation learning. A literature review on existing graph representation learning is provided in the beginning of the dissertation. The primary focus of dissertation is on building novel graph neural network architectures that are robust against adversarial attacks. The proposed graph neural network models are extended to multiplex graphs (heterogeneous graphs). Finally, a relational neural network model is proposed to operate on a human structural connectome. For every research contribution of this dissertation, several empirical studies are conducted on benchmark datasets. The proposed graph neural network models, approaches, and architectures demonstrate significant performance improvements in comparison to the existing state-of-the-art graph embedding strategies.

In many real-world machine learning classification applications, well labeled training data can be difficult, expensive, or even impossible to obtain. In such situations, it is sometimes possible to label a small subset of data as belonging to the class of interest though it is impractical to manually label all data not of interest. The result is a small set of positive labeled data and a large set of unknown and unlabeled data. This is known as the Positive and Unlabeled learning (PU learning) problem, a type of semi-supervised learning. In this dissertation, the PU learning problem is rigorously defined, several common assumptions described, and a literature review of the field provided. A new family of effective PU learning algorithms, the MLR (Modified Logistic Regression) family of algorithms, is described. Theoretical and experimental justification for these algorithms is provided demonstrating their success and flexibility. Extensive experimentation and empirical evidence are provided comparing several new and existing PU learning evaluation estimation metrics in a wide variety of scenarios. The surprisingly clear advantage of a simple recall estimate as the best estimate for overall PU classifier performance is described. Finally, an application of PU learning to the field of solar fault detection, an area not previously explored in the field, demonstrates the advantage and potential of PU learning in new application domains.

A distributed wireless sensor network (WSN) is a network of a large number of lowcost,multi-functional sensors with power, bandwidth, and memory constraints, operating in remote environments with sensing and communication capabilities. WSNs are a source for a large amount of data and due to the inherent communication and resource constraints, developing a distributed algorithms to perform statistical parameter estimation and data analysis is necessary. In this work, consensus based distributed algorithms are developed for distributed estimation and processing over WSNs. Firstly, a distributed spectral clustering algorithm to group the sensors based on the location attributes is developed. Next, a distributed max consensus algorithm robust to additive noise in the network is designed. Furthermore, distributed spectral radius estimation algorithms for analog, as well as, digital communication models are developed. The proposed algorithms work for any connected graph topologies. Theoretical bounds are derived and simulation results supporting the theory are also presented.

In recent years, the widespread use of deep neural networks (DNNs) has facilitated great improvements in performance for computer vision tasks like image classification and object recognition. In most realistic computer vision applications, an input image undergoes some form of image distortion such as blur and additive noise during image acquisition or transmission. Deep networks trained on pristine images perform poorly when tested on such distortions. DNN predictions have also been shown to be vulnerable to carefully crafted adversarial perturbations. Specifically, so-called universal adversarial perturbations are image-agnostic perturbations that can be added to any image and can fool a target network into making erroneous predictions. This work proposes selective DNN feature regeneration to improve the robustness of existing DNNs to image distortions and universal adversarial perturbations.

In the context of common naturally occurring image distortions, a metric is proposed to identify the most susceptible DNN convolutional filters and rank them in order of the highest gain in classification accuracy upon correction. The proposed approach called DeepCorrect applies small stacks of convolutional layers with residual connections at the output of these ranked filters and trains them to correct the most distortion-affected filter activations, whilst leaving the rest of the pre-trained filter outputs in the network unchanged. Performance results show that applying DeepCorrect models for common vision tasks significantly improves the robustness of DNNs against distorted images and outperforms other alternative approaches.

In the context of universal adversarial perturbations, departing from existing defense strategies that work mostly in the image domain, a novel and effective defense which only operates in the DNN feature domain is presented. This approach identifies pre-trained convolutional features that are most vulnerable to adversarial perturbations and deploys trainable feature regeneration units which transform these DNN filter activations into resilient features that are robust to universal perturbations. Regenerating only the top 50% adversarially susceptible activations in at most 6 DNN layers and leaving all remaining DNN activations unchanged can outperform existing defense strategies across different network architectures and across various universal attacks.

An analysis is presented of a network of distributed receivers encumbered by strong in-band interference. The structure of information present across such receivers and how they might collaborate to recover a signal of interest is studied. Unstructured (random coding) and structured (lattice coding) strategies are studied towards this purpose for a certain adaptable system model. Asymptotic performances of these strategies and algorithms to compute them are developed. A jointly-compressed lattice code with proper configuration performs best of all strategies investigated.

A critical problem for airborne, ship board, and land based radars operating in maritime or littoral environments is the detection, identification and tracking of targets against backscattering caused by the roughness of the sea surface. Statistical models, such as the compound K-distribution (CKD), were shown to accurately describe two separate structures of the sea clutter intensity fluctuations. The first structure is the texture that is associated with long sea waves and exhibits long temporal decorrelation period. The second structure is the speckle that accounts for reflections from multiple scatters and exhibits a short temporal decorrelation period from pulse to pulse. Existing methods for estimating the CKD model parameters do not include the thermal noise power, which is critical for real sea clutter processing. Estimation methods that include the noise power are either computationally intensive or require very large data records.



This work proposes two new approaches for accurately estimating all three CKD model parameters, including noise power. The first method integrates, in an iterative fashion, the noise power estimation, using one-dimensional nonlinear curve fitting,

with the estimation of the shape and scale parameters, using closed-form solutions in terms of the CKD intensity moments. The second method is similar to the first except it replaces integer-based intensity moments with fractional moments which have been shown to achieve more accurate estimates of the shape parameter. These new methods can be implemented in real time without requiring large data records. They can also achieve accurate estimation performance as demonstrated with simulated and real sea clutter observation datasets. The work also investigates the numerically computed Cram\'er-Rao lower bound (CRLB) of the variance of the shape parameter estimate using intensity observations in thermal noise with unknown power. Using the CRLB, the asymptotic estimation performance behavior of the new estimators is studied and compared to that of other estimators.

In the past half century, low-power wireless signals from portable radar sensors, initially continuous-wave (CW) radars and more recently ultra-wideband (UWB) radar systems, have been successfully used to detect physiological movements of stationary human beings.

The thesis starts with a careful review of existing signal processing techniques and state of the art methods possible for vital signs monitoring using UWB impulse systems. Then an in-depth analysis of various approaches is presented.

Robust heart-rate monitoring methods are proposed based on a novel result: spectrally the fundamental heartbeat frequency is respiration-interference-limited while its higher-order harmonics are noise-limited. The higher-order statistics related to heartbeat can be a robust indication when the fundamental heartbeat is masked by the strong lower-order harmonics of respiration or when phase calibration is not accurate if phase-based method is used. Analytical spectral analysis is performed to validate that the higher-order harmonics of heartbeat is almost respiration-interference free. Extensive experiments have been conducted to justify an adaptive heart-rate monitoring algorithm. The scenarios of interest are, 1) single subject, 2) multiple subjects at different ranges, 3) multiple subjects at same range, and 4) through wall monitoring.

A remote sensing radar system implemented using the proposed adaptive heart-rate estimation algorithm is compared to the competing remote sensing technology, a remote imaging photoplethysmography system, showing promising results.

State of the art methods for vital signs monitoring are fundamentally related to process the phase variation due to vital signs motions. Their performance are determined by a phase calibration procedure. Existing methods fail to consider the time-varying nature of phase noise. There is no prior knowledge about which of the corrupted complex signals, in-phase component (I) and quadrature component (Q), need to be corrected. A precise phase calibration routine is proposed based on the respiration pattern. The I/Q samples from every breath are more likely to experience similar motion noise and therefore they should be corrected independently. High slow-time sampling rate is used to ensure phase calibration accuracy. Occasionally, a 180-degree phase shift error occurs after the initial calibration step and should be corrected as well. All phase trajectories in the I/Q plot are only allowed in certain angular spaces. This precise phase calibration routine is validated through computer simulations incorporating a time-varying phase noise model, controlled mechanic system, and human subject experiment.

Multiple-input multiple-output systems have gained focus in the last decade due to the benefits they provide in enhancing the quality of communications. On the other hand, full-duplex communication has attracted remarkable attention due to its ability to improve the spectral efficiency compared to the existing half-duplex systems. Using full-duplex communications on MIMO co-operative networks can provide us solutions that can completely outperform existing systems with simultaneous transmission and reception at high data rates.

This thesis considers a full-duplex MIMO relay which amplifies and forwards the received signals, between a source and a destination that do not a have line of sight. Full-duplex mode raises the problem of self-interference. Though all the links in the system undergo frequency flat fading, the end-to-end effective channel is frequency selective. This is due to the imperfect cancellation of the self-interference at the relay and this residual self-interference acts as intersymbol interference at the destination which is treated by equalization. This also leads to complications in form of recursive equations to determine the input-output relationship of the system. This also leads to complications in the form of recursive equations to determine the input-output relationship of the system.

To overcome this, a signal flow graph approach using Mason's gain formula is proposed, where the effective channel is analyzed with keen notice to every loop and path the signal traverses. This gives a clear understanding and awareness about the orders of the polynomials involved in the transfer function, from which desired conclusions can be drawn. But the complexity of Mason's gain formula increases with the number of antennas at relay which can be overcome by the proposed linear algebraic method. Input-output relationship derived using simple concepts of linear algebra can be generalized to any number of antennas and the computation complexity is comparatively very low.

For a full-duplex amplify-and-forward MIMO relay system, assuming equalization at the destination, new mechanisms have been implemented at the relay that can compensate the effect of residual self-interference namely equal-gain transmission and antenna selection. Though equal-gain transmission does not perform better than the maximal ratio transmission, a trade-off can be made between performance and implementation complexity. Using the proposed antenna selection strategy, one pair of transmit-receive antennas at the relay is selected based on four selection criteria discussed. Outage probability analysis is performed for all the strategies presented and detailed comparison has been established. Considering minimum mean-squared error decision feedback equalizer at the destination, a bound on the outage probability has been obtained for the antenna selection case and is used for comparisons. A cross-over point is observed while comparing the outage probabilities of equal-gain transmission and antenna selection techniques, as the signal-to-noise ratio increases and from that point antenna selection outperforms equal-gain transmission and this is explained by the fact of reduced residual self-interference in antenna selection method.

Both two-way relays (TWR) and full-duplex (FD) radios are spectrally efficient, and their integration shows great potential to further improve the spectral efficiency, which offers a solution to the fifth generation wireless systems. High quality channel state information (CSI) are the key components for the implementation and the performance of the FD TWR system, making channel estimation in FD TWRs crucial.

The impact of channel estimation on spectral efficiency in half-duplex multiple-input-multiple-output (MIMO) TWR systems is investigated. The trade-off between training and data energy is proposed. In the case that two sources are symmetric in power and number of antennas, a closed-form for the optimal ratio of data energy to total energy is derived. It can be shown that the achievable rate is a monotonically increasing function of the data length. The asymmetric case is discussed as well.

Efficient and accurate training schemes for FD TWRs are essential for profiting from the inherent spectrally efficient structures of both FD and TWRs. A novel one-block training scheme with a maximum likelihood (ML) estimator is proposed to estimate the channels between the nodes and the residual self-interference (RSI) channel simultaneously. Baseline training schemes are also considered to compare with the one-block scheme. The Cramer-Rao bounds (CRBs) of the training schemes are derived and analyzed by using the asymptotic properties of Toeplitz matrices. The benefit of estimating the RSI channel is shown analytically in terms of Fisher information.

To obtain fundamental and analytic results of how the RSI affects the spectral efficiency, one-way FD relay systems are studied. Optimal training design and ML channel estimation are proposed to estimate the RSI channel. The CRBs are derived and analyzed in closed-form so that the optimal training sequence can be found via minimizing the CRB. Extensions of the training scheme to frequency-selective channels and multiple relays are also presented.

Simultaneously sensing and transmission in an FD cognitive radio system with MIMO is considered. The trade-off between the transmission rate and the detection accuracy is characterized by the sum-rate of the primary and the secondary users. Different beamforming and combining schemes are proposed and compared.