<p>Recent developments in Artificial Intelligence (AI) and Machine Learning (ML) have not only revolutionized engineering but also the way humanity foresees the future with machines. From self-driving cars to large language models and ChatGPT, AI and ML will continue to redefine the boundaries of innovations and reshape the way we interact with the world. The anticipated benefits are transformative, enabling enhanced productivity, improved decision-making, and the potential for significant cost savings. These developments in AI/ML and the promise for improved reliability, anomaly detection, efficient operation, etc., have unavoidably caught the attention of nuclear engineers. Advancing nuclear predictive models and providing real-time support with regard to operation and maintenance are just a few of the potential tasks AI/ML could provide assistance. Microreactors is just one example of future nuclear systems where semi-autonomous operation and fully digital instrumentation and control with AI/ML-based decision support would be required for cost-effective deployment in remote areas.</p><p>However, the world of nuclear engineering is skeptical of the direct application of AI/ML at nuclear facilities mostly due to limited past experience, potential high risk for false negatives, and limited amount of available data to demonstrate widespread applicability with high confidence. In order to curb these worries and take advantage of recent public interest in AI/ML, publicly available, real-time datasets need to be created. In this thesis, a universal AI/ML dataset is developed takes advantage of the recent digitization of Purdue University Reactor One (PUR-1) and using real-time data directly from PUR-1. The expectation is to follow the paradigm of the AI/ML community where open datasets (e.g., Kaggle, ImageNet, etc.) were the stepping stone towards new algorithms, facilitating collaborative problem-solving, and driving breakthroughs in the field of AI/ML through open competitions and knowledge sharing.</p><p>PUR-1 is capable of providing real-time research data to the second for over 2000 different parameters ranging from physical components such as neutron flux and control rod positions to calculated signals such as the system change rate. The proposed Purdue Reactor Integrated Machine Learning dataset (PRIMaL), as described in the thesis herein, includes ten signals handpicked to create simple and of various degree of complexity AI/ML benchmarks related directly to the nuclear field, with the goal of kickstarting both a new-founded interest in the nuclear field by AI/ML professionals and building faith in AI/ML amongst nuclear engineers. To the best of our knowledge, PRIMaL is the first curated AI/ML benchmark based on real reactor data and focused on nuclear applications, aiming to advance safety, efficiency, and innovation in the nuclear industry while promoting the responsible and secure use of AI/ML technologies.</p><p>To confirm the validity of the dataset and provide a simple example on how to use the dataset for AI/ML benchmarking, an example problem of classifying shutdown data as gang lowers or SCRAM was performed using three ML algorithms: support vector machine, random forest, and logistic regression. This binary classification problem was repeated 288 times for each algorithm, varying the balance ratio of the SCRAMs to gang lowers, the time prior to the shutdown, and the time after the shutdown the algorithms have access to. The sample problem was a success, as the algorithms were able to distinguish SCRAMs and gang lowers with reasonable accuracy in all cases. Future work would include gathering more data from PUR-1 for the database, as further testing with different sized balanced datasets lead to unusually high accuracy due to the smaller sample size.</p>

<p>This thesis explores the usage of machine learning (ML) algorithms and software-defined radio (SDR) hardware for the detection of signal jamming and message injection cyberattacks against unmanned aerial vehicle (UAV) wireless communications. In the first work presented in this thesis, a real-time ML solution for classifying four types of jamming attacks is proposed for implementation with a UAV using an onboard Raspberry Pi computer and HackRF One SDR. Also presented in this thesis is a multioutput multiclass convolutional neural network (CNN) model implemented for the purpose of identifying the direction in which a jamming sample is received from, in addition to detecting and classifying the jamming type. Such jamming types studied herein are barrage, single-tone, successive-pulse, and protocol-aware jamming. The findings of this chapter forms the basis of a reinforcement learning (RL) approach for UAV flightpath modification as the next stage of this research. The final work included in this thesis presents a ML solution for the binary classification of three different message injection attacks against ADS-B communication systems, namely path modification, velocity drift and ghost aircraft injection attacks. The collective results of these individual works demonstrate the viability of artificial-intelligence (AI) based solutions for cybersecurity applications with respect to UAV communications.</p>

<p>In this thesis, we study optimization problems which have a combinatorial aspect to them. Search space for such problems quickly grows large - exponentially - with respect to the problem dimension. Thus, exhaustive search becomes intractable and we need good relaxations to solve combinatorial problems efficiently. Another challenge arises due to the high dimensionality of such problems and lack of large number of samples. Our aim is to come up with innovative approaches that solve the problem in polynomial time and sample complexity. We discuss three combinatorial optimization problems and provide continuous relaxations for them. Our continuous relaxations involve both convex and nonconvex (invex) relaxations. Furthermore, we provide efficient first order algorithms to solve a general class of invex problems with provable convergence rate guarantees. The three combinatorial problems we study in this work are – learning the directed structure of a Bayesian network using blackbox data, fair sparse regression on a biased dataset where bias depends upon a hidden binary attribute and mixed linear regression. We propose convex relaxation for the first problem, while the other two are solved using invex relaxation. On the first problem, we come up with a novel notion of low rank representation of conditional probability tables for a Bayesian network and connect it to Fourier transformation of real valued set functions to recover the exact structure of the Bayesian networks. For the second problem, we propose a novel invex relaxation for the combinatorial version of sparse linear regression with fairness. For the final problem, we again use invex relaxation to learn a mixture of sparse linear regression models. We formally show correctness of our proposed methods and provide provable theoretical guarantees for efficient computational and sample complexity. We also develop efficient first order algorithms to solve invex problems. We provide convergence rate analysis for our proposed methods. Furthermore, we also discuss possible future research directions and the problems we want to tackle in future.</p>

<p>Determining the macroscopic properties such as diffusivity, concentration, and viscosity is of paramount importance to many engineering applications. The determination of macroscopic properties from experimental or numerical data is a challenging task due to the inverse nature of these problems. Data analytic techniques with recent advances in machine learning as well as optimization techniques have enabled tackling problems that were once considered impossible to solve. In the current proposal, we focus on using Bayesian and the state of the art machine learning techniques to solve three problems that involve calculations of the macroscopic transport properties. </p> <p><br></p> <p>i) We developed a Bayesian approach to estimate the diffusion coefficient of rhodamine 6G in breast cancer spheroids. Determination of the diffusivity values of drugs in tumors is crucial to understanding drug resistivity, particularly in breast cancer tumors. To this end, we invoked Bayesian inference to solve the problem of determining the light attenuation coefficient and diffusion coefficient in breast cancer spheroids for Rhodamine 6G (R6G) as a mock drug for the tyrosine kinase inhibitor, Neratinib. We noticed that the diffusion coefficient values do not noticeably vary across a HER2+ breast cancer cell line as a function of transglutaminase 2 levels, even in the presence of fibroblast cells. </p> <p><br></p> <p>ii) We developed a multi-fidelity model to predict the rheological properties of a suspension of fibers using neural networks and Gaussian processes. Determining the rheological properties of fiber suspensions is of indispensable to many industrial applications. To this end,  multi-fidelity Gaussian processes and neural networks were utilized to predict the apparent viscosity. Results indicated that with tuned hyperparameters, both the multi-fidelity Gaussian processes and neural networks lead to predictions with a high level of accuracy, where neural networks demonstrate marginally better performance.</p> <p><br></p> <p><br></p> <p>iii) We developed machine learning models to analyze measles,</p> <p>mumps, rubella, and varicella (MMRV) vaccines using Raman and absorption spectra. Monitoring the concentration of viral particles is indispensable to producing vaccines or anti-viral medications. To this end, we designed and optimized a convolutional neural network and random forest models to map spectroscopic signals to concentration values. Results indicated that when the joint Raman-absorption signals are used for training, prediction accuracies are higher, with the random forest model demonstrating marginally better performance.  </p> <p><br></p> <p>iv) We developed four machine learning models, including random forest, support vector machine, artificial neural networks, and convolutional neural networks to classify diseases using gut microbiota data. We distinguished between Parkinson’s disease, Crohn’s disease (CD), ulcerative colitis (UC), human immune deficiency virus (HIV), and healthy control (HC) subjects in the</p> <p>presence and absence of fiber treatments. Our analysis demonstrated that it would be possible to use machine learning to distinguish between healthy and non-healthy cases in addition to predicting four different types of diseases with very high accuracy. </p> <p>v</p>

<p>In this work, machine learning (ML) modeling is proposed for the detection and classification of global positioning system (GPS) spoofing in unmanned aerial vehicles (UAVs). Three testing scenarios are implemented in an outdoor yet controlled setup to investigate static and dynamic attacks. In these scenarios, authentic sets of GPS signal features are collected, followed by other sets obtained while the UAV is under spoofing attacks launched with a software-defined radio (SDR) transceiver module. All sets are standardized, analyzed for correlation, and reduced according to feature importance prior to their exploitation in training, validating, and testing different multiclass ML classifiers. Two schemes for the dataset are proposed, location-dependent and location-independent datasets. The location-dependent dataset keeps the location specific features which are latitude, longitude, and altitude. On the other hand, the location-independent dataset excludes these features. The resulting performance evaluation of these classifiers shows a detection rate (DR), misdetection rate (MDR), and false alarm rate (FAR) better than 92%, 13%, and 4%, respectively, together with a sub-millisecond detection time. Hence, the proposed modeling facilitates accurate real-time GPS spoofing detection and classification for UAV applications.</p> <p><br></p> <p>Then, a three-class ML model is implemented on a UAV with a Raspberry Pi processor for classifying the two GPS spoofing attacks (i.e., static, dynamic) in real-time. First, several models are developed and tested utilizing the prepared dataset. Models evaluation is carried out using the DR, F-score, FAR, and MDR, which all showed an acceptable performance. Then, the optimum model is loaded to the onboard processor and tested for real-time detection and classification. Location-dependent applications, such as fixed-route public transportation, are expected to benefit from the methodology presented herein as the longitude, latitude, and altitude features are characterized in the implemented model.</p>

<p dir="ltr">Lithium-ion batteries are being used increasingly more often to provide power for systems that range all the way from common cell-phones and laptops to advanced electric automotive and aircraft vehicles. However, as is the case for all battery types, lithium-ion batteries are prone to naturally occurring degradation phenomenon that limit their effective use in these systems to a finite amount of time. This degradation is caused by a plethora of variables and conditions including things like environmental conditions, physical stress/strain on the body of the battery cell, and charge/discharge parameters and cycling. Accurately and reliably being able to predict this degradation behavior in battery systems is crucial for any party looking to implement and use battery powered systems. However, due to the complicated non-linear multivariable processes that affect battery degradation, this can be difficult to achieve. Compared to traditional methods of battery degradation prediction and modeling like equivalent circuit models and physics-based electrochemical models, data-driven machine learning tools have been shown to be able to handle predicting and classifying the complex nature of battery degradation without requiring any prior knowledge of the physical systems they are describing.</p><p dir="ltr">One of the most critical steps in developing these data-driven neural network algorithms is data procurement and preprocessing. Without large amounts of high-quality data, no matter how advanced and accurate the architecture is designed, the neural network prediction tool will not be as effective as one trained on high quality, vast quantities of data. This work aims to gather battery degradation data from a wide variety of sources and studies, examine how the data was produced, test the effectiveness of the data in the Interfacial Multiphysics Laboratory’s autoencoder based neural network tool CD-Net, and analyze the results to determine factors that make battery degradation datasets perform better for use in machine learning/deep learning tools. This work also aims to relate this work to other data-driven models by comparing the CD-Net model’s performance with the publicly available BEEP’s (Battery Evaluation and Early Prediction) ElasticNet model. The reported accuracy and prediction models from the CD-Net and ElasticNet tools demonstrate that larger datasets with actively selected training/testing designations and less errors in the data produce much higher quality neural networks that are much more reliable in estimating the state-of-health of lithium-ion battery systems. The results also demonstrate that data-driven models are much less effective when trained using data from multiple different cell chemistries, form factors, and cycling conditions compared to more congruent datasets when attempting to create a generalized prediction model applicable to multiple forms of battery cells and applications.</p>

<p>While instance-dependent contextual bandits have been previously studied, their analysis<br> has been exclusively limited to pure disagreement-based learning. This approach lacks a<br> nuanced understanding of disagreement and treats it in a binary and absolute manner.<br> In our work, we aim to broaden the analysis of instance-dependent contextual bandits by<br> studying them under the framework of disagreement-based learning in sub-regions. This<br> framework allows for a more comprehensive examination of disagreement by considering its<br> varying degrees across different sub-regions.<br> To lay the foundation for our analysis, we introduce key ideas and measures widely<br> studied in the contextual bandit and disagreement-based active learning literature. We<br> then propose a novel, instance-dependent contextual bandit algorithm for the realizable<br> case in a transductive setting. Leveraging the ability to observe contexts in advance, our<br> algorithm employs a sophisticated Linear Programming subroutine to identify and exploit<br> sub-regions effectively. Next, we provide a series of results tying previously introduced<br> complexity measures and offer some insightful discussion on them. Finally, we enhance the<br> existing regret bounds for contextual bandits by integrating the sub-region disagreement<br> coefficient, thereby showcasing significant improvement in performance against the pure<br> disagreement-based approach.<br> In the concluding section of this thesis, we do a brief recap of the work done and suggest<br> potential future directions for further improving contextual bandit algorithms within the<br> framework of disagreement-based learning in sub-regions. These directions offer opportuni-<br> ties for further research and development, aiming to refine and enhance the effectiveness of<br> contextual bandit algorithms in practical applications.<br> <br> </p>

<p dir="ltr">The parallel growth of contemporary machine learning (ML) technologies alongside edge/-fog networking has necessitated the development of novel paradigms to effectively manage their intersection. Specifically, the proliferation of edge devices equipped with data generation and ML model training capabilities has given rise to an alternative paradigm called federated learning (FL), moving away from traditional centralized ML common in cloud-based networks. FL involves training ML models directly on edge devices where data are generated.</p><p dir="ltr">A fundamental challenge of FL lies in the extensive heterogeneity inherent to edge/fog networks, which manifests in various forms such as (i) statistical heterogeneity: edge devices have distinct underlying data distributions, (ii) structural heterogeneity: edge devices have diverse physical hardware, (iii) data quality heterogeneity: edge devices have varying ratios of labeled and unlabeled data, and (iv) adversarial compromise: some edge devices may be compromised by adversarial attacks. This dissertation endeavors to capture and model these intricate relationships at the intersection of FL and highly heterogeneous edge/fog networks. To do so, this dissertation will initially develop closed-form expressions for the trade-offs between ML performance and resource cost considerations within edge/fog networks. Subsequently, it optimizes the fundamental processes of FL, encompassing aspects such as batch size control for stochastic gradient descent (SGD) and sampling for global aggregations. This optimization is jointly formulated with networking considerations, which include communication resource consumption and device-to-device (D2D) cooperation.</p><p dir="ltr">In the former half of the dissertation, the emphasis is first on optimizing device sampling for global aggregations in FL, and then on developing a self-sufficient hierarchical meta-learning approach for FL. These methodologies maximize expected ML model performance while addressing common challenges associated with statistical and system heterogeneity. Novel techniques, such as management of D2D data offloading, adaptive CPU clock cycle control, integration of meta-learning, and much more, enable these methodologies. In particular, the proposed hierarchical meta-learning approach enables rapid integration of new devices in large-scale edge/fog networks.</p><p dir="ltr">The latter half of the dissertation directs its ocus towards emerging forms of heterogeneity in FL scenarios, namely (i) heterogeneity in quantity and quality of local labeled and unlabeled data at edge devices and (ii) heterogeneity in terms of adversarially comprised edge devices. To deal with heterogeneous labeled/unlabeled data across edge networks, this dissertation proposes a novel methodology that enables multi-source to multi-target federated domain adaptation. This proposed methodology views edge devices as sources – devices with mostly labeled data that perform ML model training, or targets - devices with mostly unlabeled data that rely on sources’ ML models, and subsequently optimizes the network relationships. In the final chapter, a novel methodology to improve FL robustness is developed in part by viewing adversarial attacks on FL as a form of heterogeneity.</p>

<p>  </p> <p>Semantic segmentation is a machine learning task that is seeing increased utilization in multiples fields, from medical imagery, to land demarcation, and autonomous vehicles. Semantic segmentation performs the pixel-wise classification of images, creating a new, segmented representation of the input that can be useful for detected various terrain and objects within and image. Recently, convolutional neural networks have been heavily utilized when creating neural networks tackling the semantic segmentation task. This is particularly true in the field of autonomous driving systems.</p> <p>The requirements of automated driver assistance systems (ADAS) drive semantic segmentation models targeted for deployment on ADAS to be lightweight while maintaining accuracy. A commonly used method to increase accuracy in the autonomous vehicle field is to fuse multiple sensory modalities. This research focuses on leveraging the fusion of long wave infrared (LWIR) imagery with visual spectrum imagery to fill in the inherent performance gaps when using visual imagery alone. This comes with a host of benefits, such as increase performance in various lighting conditions and adverse environmental conditions. Utilizing this fusion technique is an effective method of increasing the accuracy of a semantic segmentation model. Being a lightweight architecture is key for successful deployment on ADAS, as these systems often have resource constraints and need to operate in real-time. Multi-Spectral Fusion Network (MFNet) [ 1 ] accomplishes these parameters by leveraging a sensory fusion approach, and as such was selected as the baseline architecture for this research.</p> <p>Many improvements were made upon the baseline architecture by leveraging a variety of techniques. Such improvements include the proposal of a novel loss function categorical cross-entropy dice loss, introduction of squeeze and excitation (SE) blocks, addition of pyramid pooling, a new fusion technique, and drop input data augmentation. These improvements culminated in the creation of the Fast Thermal Fusion Network (FTFNet). Further improvements were made by introducing depthwise separable convolutional layers leading to lightweight FTFNet variants, FTFNet Lite 1 & 2.</p>

<p>This thesis aims to design and develop the spatial adaptation approach through spatial transformers to improve the accuracy of human keypoint recognition models. We have studied different model types and design choices to gain an accuracy increase over models without spatial transformers and analyzed how spatial transformers increase the accuracy of predictions. A neural network called Widenet has been leveraged as a specialized network for providing the parameters for the spatial transformer. Further, we have evaluated methods to reduce the model parameters, as well as the strategy to enhance the learning supervision for further improving the performance of the model. Our experiments and results have shown that the proposed deep learning framework can effectively detect the human key points, compared with the baseline methods. Also, we have reduced the model size without significantly impacting the performance, and the enhanced supervision has improved the performance. This study is expected to greatly advance the deep learning of human key points and movement dynamics. </p>