DISSERTATION DEFENSE

Author : Iftakhar Ahmad
Advisors: Dr. Jason Bakos
Date: October 23, 2026
Time: 11:00 am
Location: Seminar room 2277, Storey Innovation Center
Link:  https://teams.microsoft.com/meet/276223938519678?p=lH74ckYG9mZET45s5D

Abstract

Cloud-based deep learning models are experiencing an era of unprecedented improvement in capability and adoption, driven in large part by corresponding growth in cloud-based infrastructure. However, models that are deployable on unconnected embedded systems are seeing a substantially slower pace of growth. This is a critical capability gap because embedded models, especially those that can operate in real-time, have the potential to revolutionize autonomous systems. One of the challenges inherent in improving capability in locally-executed models is their asymptotic tradeoff between accuracy and cost. Additionally, neuromorphic architectures also exhibit an asymptotic tradeoff between performance and scale. Co-optimizing model accuracy and the capabilities delivered by an underlying target platform is still an open problem, which requires exploration of both the hyperparameter and hardware parameter spaces. Specifically, Deep Neural Network (DNN) models like MobileNet, YOLO, and Vision Transformers (ViT) offer great promise in real-time applications. Such models must be tuned carefully to meet the constraints of the embedded platforms while also finding a practical compromise between performance and cost in terms of resource usage and energy expenditure. This dissertation will develop tools to perform constrained multi-objective optimization of large DNN models along with their corresponding target platforms. The proposed tools will develop objective functions that combine model accuracy and execution time while constraining memory and power requirements. The proposed tools will derive model accuracy using Bayesian optimization over the space of hyperparameters and will model platform performance using a roofline model. The proposed framework will also generate an optimal Pareto front of network and device configurations to meet real-time constraints. The outcome of this dissertation will be a generalized hardware-software co-optimization framework that can take in any large DNN model and optimize it for a given hardware.

DISSERTATION DEFENSE

Author : Fawad Kirmani
Advisors: Dr. John Rose
Date: October 22, 2026
Time: 10:00 am
Location: Virtual
Link: https://teams.microsoft.com/l/meetup-join/19%3ameeting_NzI3MDUwODctMzVi…

Abstract

In deep learning classification tasks, class imbalance is a frequent challenge, because a model trained on imbalanced data tends to favor the majority class and miss the rare cases that are often of greatest interest. In this work, we introduce our custom loss function, which adds a symmetric Kullback--Leibler divergence term to categorical cross-entropy. It is designed to increase recall of the minority class while limiting the loss of recall in the majority class. We evaluated our custom loss function against categorical cross-entropy and focal loss using the first two classes of the CIFAR-10 dataset. Over 30 seeds and 12 class imbalance ratios, it increased the recall of the minority class most under severe imbalance, by 5.15 percentage points at the most imbalanced ratio, while keeping the decrease in the recall of the majority class below 2 percentage points at every ratio, at the cost of a lower ROC-AUC. In addition, we also evaluated our custom loss function on the German Credit dataset. These additional results are presented in an appendix. Having demonstrated the utility of our custom loss function, we applied it to two binary classification tasks in proteomics and astronomy. We conclude with a multiclass task in astronomy that does not use it.

First, we developed a bidirectional long short-term memory (BiLSTM) autoencoder for identifying proteotypic peptides. The model combines peptide embeddings with physicochemical peptide features and is trained with our custom loss function. Under 5x2 cross-validation, it achieved a mean ROC-AUC of 95.08%, sensitivity of 92.39%, and specificity of 77.55%. Relative to categorical cross-entropy, our custom loss function significantly reduced the misclassification rate, and an ensemble with an XGBoost model reduced it further.

Second, we developed a deep convolutional neural network to detect polar ring galaxies in Sloan Digital Sky Survey images, trained with our custom loss function. Across ten folds, the model achieved a mean test ROC-AUC of approximately 99% and a mean recall of 96% for the polar ring galaxy class, and it maintained strong performance on a heavily imbalanced subset of 20 polar ring galaxies and 1,000 other galaxies.

Third, we developed a deep learning model to detect gravitationally lensed supernovae (SNe), which are extremely rare and fade quickly. Instead of relying on magnification, our model identifies doubly-imaged SNe with small separations (<0.6 arcsec) from the shape of the transient in a single difference image, so it can give an early alert before a light curve is available. We trained a VGG19 network with a Convolutional Block Attention Module on simulated SNe planted into archival Hubble Space Telescope images. Across ten test folds, the model achieved a mean ROC-AUC of 99.29% and mean recall scores of 96.6%, 91.0%, and 97.1% for images with zero, one, and two simulated transients. Unlike the Source-Extractor baseline, which could not identify single transients, it separated single from doubly-imaged SNe. Further work is needed to assess its performance on real survey data.

Together, these studies show that our custom loss function can improve minority-class performance in imbalanced binary classification while making the trade-offs explicit, and they demonstrate deep learning applications in proteomics and astronomy, where identifying rare objects or events is an important scientific problem.
 

DISSERTATION DEFENSE

Author : Ivan Panchenko
Advisors: Dr. Jason Bakos
Date: October 21, 2026
Time: 02:30 pm
Location: Meeting room 2267, Storey Innovation Center
Link:  https://teams.microsoft.com/meet/2315667690889461?p=FqdCxjM455jN32NJOF

Abstract

This dissertation develops a routing scheme that guarantees delivery and real-time performance constraints for distributed controller networks having static, periodic traffic patterns and suffer unexpected faults. The proposed scheme targets an electrical distribution system consisting of a set of power electronic converters with associated local controller sharing a common global controller.  Each converter and local controller uses FPGA-based interfaces to a torus network topology supporting static traffic patterns needed to achieve real-time control. The proposed routing scheme will maintain delivery and performance guarantees for unpredictable node and channel failures through balancing channel load and limiting routing path distance. To achieve these goals, we will investigate two approaches.  The first will precompute optimal routing tables for failure state contingencies.  The second is to use unused channel capacity for redundant packets that follow maximally divergent paths

DISSERTATION DEFENSE

Author : Andrew Kenneth Smith 
Advisors: Dr. Homayoun Valafar
Date: October 21, 2026
Time: 02:00 pm
Location: AI Seminar Room 529
Link:  https://teams.microsoft.com/meet/226969496196462?p=FsccssHidNdPu1F1S5

Abstract

Physiological signals such as EEG, accelerometry, and inertial motion carry rich information about health and behavior, yet building reliable classifiers that work outside the laboratory remains difficult. This dissertation advances a single thesis: machine learning classifiers for physiological time series generalize well across health behaviors, clinical contexts, and species when sufficient population data is available, but population models do not generalize uniformly to the individual. Closing the gap to the individual requires adapting the model itself: recalibrating a per-person decision threshold is a cheap, label-light first step, but it recovers only part of what fine-tuning achieves. How much it recovers depends on the person as well as the behavior, so the regimes that decide the right intervention (calibration-limited, representation-limited, and individual-invariant) belong to a person and behavior pair rather than to a behavior, and the central contribution is a framework that identifies which regime a person and behavior pair falls in and matches the personalization intervention accordingly.

The argument is developed through five studies that progress from population models toward personalization. A feature-engineering baseline establishes classical machine learning for rodent sleep staging, after which a deep convolutional recurrent model reproduces expert-level vigilance-state scoring at a macro F1 of 0.878. The same architectural principles transfer to human activity recognition, where a single model discriminates multiple health behaviors from a single wrist sensor at a macro F1 of 0.817, and to medication-adherence monitoring in opioid use disorder treatment, where neural gesture recognition detects therapeutic self-administration. A capstone study on smoking detection exposes the core obstacle directly: population models cluster gestures cleanly yet classify individuals only moderately. Recalibrating each participant’s decision threshold, with zero gradient updates, raises mean smoking-class F1 from 0.702 to 0.767, but this recovers only 43.1% of the gain that fine-tuning achieves (F1 0.853), and fine-tuning beats calibration significantly. The convolutional layers can stay frozen, but the recurrent layers must adapt, and for one participant calibration recovers almost nothing and only fine-tuning closes the gap. In the medication-adherence setting of opioid use disorder, distinguishing methadone from buprenorphine is representation-limited at the behavior level: threshold recalibration does not help in any of three runs, while fine-tuning does. Personalization therefore needs model adaptation, with calibration as its cheap first step, and whether that first step suffices must be judged for each person and behavior pair, not per behavior alone.

Alongside the empirical studies, the dissertation contributes a reproducible annotation and experimentation platform that supports the full pipeline from raw signal import to labeled datasets and trained models. Taken together, the studies show that the path from accurate population-level classification to deployable personalized health monitoring runs through diagnosing which regime an individual’s behavior falls in and matching the personalization intervention to that regime, not through ever-larger models, and they chart a practical route for that transition across diverse physiological sensing domains.
 

DISSERTATION DEFENSE

Author : Yuxin Zi
Advisors: Dr. Amit Sheth
Date: October 20, 2026
Time: 09:00 am
Location: AI Seminar Room 529
Link:  https://sc-edu.zoom.us/j/84342587325
Meeting ID: 843 4258 7325

Abstract

While Large Language Models (LLMs) possess a remarkable capacity to generate fluent, step-by-step explanations, their underlying reasoning processes remain epistemologically fragile. The prevailing paradigm of Chain-of-Thought (CoT) prompting forces formal, multi-dimensional logic into a one-dimensional textual sequence. This fundamental misalignment between the topology of deductive reasoning and the linearity of autoregressive generation conceals causal dependencies, leading to hallucination snowballs, context dilution, and catastrophic error cascading. Furthermore, current oversight mechanisms, such as linear Process Reward Models (PRMs), evaluate reasoning steps based on chronologically flat histories and symmetric distance metrics, thereby conflating mere topical relevance with rigorous logical entailment.

This dissertation introduces a paradigm shift in how artificial reasoning is represented, verified, grounded, and repaired. It argues that robust reasoning requires the explicit scaffolding of a machine-checkable Directed Acyclic Graph (DAG), rigorously governed by asymmetric geometry, neuro-symbolic boundaries, and continuous latent-space optimization. 

DISSERTATION DEFENSE

Author :Nawras Alkassab
Advisors: Dr. Chin-Tser Huang
Date: October 19, 2026
Time: 03:15 pm
Location: Room 2267, Storey Innovation building

Abstract
In psychology, humans who are intrinsically motivated tend to outperform those whose self-validation is more dependent on extrinsic rewards. In reinforcement learning, curiosity-driven agents suffer from the noisy TV problem, since calculating the error in predicting the future observations leaves the agents perplexed in extremely randomized environments. In this work, we tackle the video prefetching problem at edge networks, an NP-Hard problem, using intrinsically motivated reinforcement learning agents. The extrinsic rewards in the video prefetching problem are sparse and delayed, since they are received from access networks by edge networks infrequently. First, we explore the benefits and schemes of video prefetching from cloud networks to edge networks. Next, we explore the design and implementation of deep reinforcement learning for video prefetching at edge networks. By formulating the prefetching problem at edge networks as Partially Observable Markov Decision Process, we propose an intrinsically motivated reinforcement learning agent, Techie, to maximize both prefetching accuracy and prefetching coverage. Techie self-tunes its aggressiveness to manage the trade-offs between prefetching accuracy and prefetching coverage, given a long-term trajectory of video requests. Finally, we conclude that the prefetching problem at edge networks is susceptible to large action space and observation space, which exponentially increases as the size of the edge network's storage increases linearly. In addition, we believe that intrinsically motivated reinforcement learning agents for video prefetching at edge networks offer multiple benefits such as their sensitivity to the popularity of video items issued by end-users at different heterogeneous access networks and their ability to make intelligent prefetching decisions online without relying on access networks' metrics such as Quality of Experience metrics and bandwidth wastage.
 

DISSERTATION DEFENSE

Author : Chathurangi Jayakody Kankanamalage
Advisors: Dr. Amit Sheth
Date: October 13, 2026
Time: 09:30 am
Location: 529 Seminar Room, AI Institute
Link:   https://sc-edu.zoom.us/j/84342587325

Abstract
Modern manufacturing systems are complex cyber–physical environments in which rare failures, changing operating conditions, and interdependent processes make reliable decision-making challenging. Although neural models achieve strong predictive performance, they often function as isolated components with limited support for causal reasoning, interpretable explanations, and process constraints. This dissertation addresses these limitations through the design and operationalization of neurosymbolic multi-agent copilots for industrial systems.

The central thesis is that integrating data-driven learning with symbolic process knowledge, causal reasoning, and coordinated agents can improve the robustness, intelligence, and operational trustworthiness of industrial decision support. The dissertation develops this thesis through five interconnected contributions.

First, it characterizes rare events in industrial data and develops structured data enrichment methods to support learning under severe class imbalance. Second, it develops specialized predictive models for anomaly detection, anomaly prediction, and forecasting, demonstrating the effectiveness of compact, task-aware architectures relative to larger foundation models on the evaluated industrial tasks. Third, it introduces knowledge-guided inference methods that combine neural predictions with process ontologies, knowledge graphs, and causal models to support logically consistent predictions, root-cause analysis, and counterfactual reasoning. Fourth, it develops multi-agent copilot architectures that coordinate specialized agents through shared knowledge and adaptive workflows. Fifth, it advances operationalization through evidence-grounded industrial question answering, agent benchmarking, and staged deployment frameworks.

Evaluations using real industrial datasets and laboratory manufacturing environments demonstrate improvements over task-specific baselines in predictive performance, reasoning consistency, and agent coordination. They also reveal remaining limitations in root-cause identification and agent evaluation. Across these contributions, process constraints, traceable evidence, and human oversight are integrated into system design to support operational trustworthiness. The dissertation provides methods and practical frameworks for advancing neurosymbolic multi-agent systems toward reliable use in industrial decision-making.
 

CSCE Colloquium

Friday, September 4, 11:00 a.m.
Innovation Center, Room 2277
550 Assembly St

Abstract
A particularly challenging problem in AI safety is providing guarantees on the behavior of high-dimensional autonomous systems. Verification approaches centered around reachability analysis fail to scale, and purely statistical approaches are constrained by the distributional assumptions about the sampling process. Instead, we pose a distributionally robust version of the statistical verification problem for black-box systems, where our performance guarantees hold over a large family of distributions. This talk describes an approach based on uncertainty quantification using concepts from imprecise probabilities. A central piece of our approach is an ensemble technique called Imprecise Neural Networks, which provides the uncertainty quantification. Additionally, we solve the allied problem of exploring the input set using active learning. The active learning uses an exhaustive neural-network verification tool Sherlock to collect samples. An evaluation on multiple physical simulators in the openAI gym Mujoco environments with reinforcement-learned controllers demonstrates that our approach can provide useful and scalable guarantees for high-dimensional systems.
 

Bio:

Dr. Ivan Ruchkin is a Malachowsky Family Endowed Rising Star Assistant Professor in the Department of Electrical & Computer Engineering at the University of Florida.  His research improves the safety, reliability, interpretability, and trustworthiness of autonomous systems by advancing their modeling, analysis, verification, monitoring, and prediction.

Previously, Ivan was a postdoctoral researcher at the PRECISE Center at the University of Pennsylvania. He received his PhD in Software Engineering from Carnegie Mellon University (CMU) and a Specialist Degree in Applied Mathematics and Computer Science (with honors) from the Lomonosov Moscow State University. Ivan also held several positions in the industry and government, including at the Air Force Research Lab (AFRL) and NASA Jet Propulsion Lab (JPL).

Ivan’s research on assuring cyber-physical systems has been widely recognized. He was fortunate to receive the NSF CAREER Award, the TCCPS Early Career Award for rigorous assurance of learning-enabled cyber-physical systems, and the Frank Anger Memorial Award for the crossover of ideas between the SIGSOFT and SIGBED communities. Ivan's work was recognized with multiple Best Paper/Poster/Demo Awards as well as a Gold Medal in the ACM SRC student competition at the MODELS conference.

DISSERTATION DEFENSE
 

Author :  Ziyu Zhao
Advisors: Dr. Song Wang
Date: Aug 17, 2026
Time: 01:00 pm
Location:  2277 (M. Bert Storey Engineering and Innovation Center)
Link:  https://teams.microsoft.com/meet/240422703068254?p=EsBfT7Q4WmKzfnWa8K

Abstract
While the accuracy of semantic segmentation has improved rapidly, extending a trained model to new categories remains expensive: each addition still requires collecting
dense annotations and training again, which limits the use of segmentation where the categories of interest are not known in advance. This dissertation studies how external prior information can assist semantic segmentation in open-vocabulary and few-shot settings, and refers to this as Prior-Conditioned Dense \mbox{Prediction}. In these settings the target category is specified by an input, consumed the same way in training and at test time, rather than learned into the model's parameters, so new categories require no retraining. This is where assistance is needed: the inputs themselves are difficult to use directly, for two different reasons. A category name identifies what to segment but provides no appearance evidence with which to localize the category in a particular image. A labeled example identifies what to segment by showing it, but in 3D segmentation such examples conventionally take the form of point clouds in which the annotation is given point by point, and producing even a few of them is laborious. The dissertation therefore constructs auxiliary prior evidence, generated from the names, transferred from a cheaper modality, or synthesized from what is given, and converts it into a form directly comparable with the scene. Dense matching between the constructed evidence and the scene then yields the labels at every pixel or point. Where the constructed evidence is unreliable, because it crosses modalities or contains generated content, the model additionally adapts it to the scene or weights it by estimated quality.

Three studies instantiate this view, ordered by how much of the evidence must be constructed. In the first study, for open-vocabulary 2D segmentation, category names are paired with diffusion-generated exemplars that supply the appearance the names lack, combined into a dual-prompt cost volume refined at inference. In the second study, for label-efficient 3D segmentation, the point-wise labeled 3D examples are replaced by labeled 2D images, lifted into pseudo RGB-D point clouds by monocular depth estimation and matched to the scene through prototypes in a shared embedding space. In the third study, where a single labeled view leaves the geometry incomplete, further views are synthesized by RGB-D inpainting; since the newly revealed regions are generated rather than observed, learned view and point weights discount them when prototypes are formed. On standard 2D and 3D benchmarks, each study improves over methods that consume the supplied inputs in their original form.

Together, the studies show that auxiliary prior evidence can strengthen dense prediction well beyond what the task's own inputs support, once it is constructed into a comparable
form and consumed through localized matching. They further indicate that adaptation and weighted aggregation are best introduced according to the semantic, cross-modal, or visibility mismatch present in each setting, rather than adopted uniformly

DISSERTATION DEFENSE

Author :  Matthew Sharp
Advisors: Dr. Chin-Tser Huang
Date: Aug 14, 2026
Time: 04:00 pm
Location: Online
Link:  https://teams.microsoft.com/meet/211658530586897?p=AAPp2DUFHhla2JclD4

Abstract

Traditional blockchain systems typically process transactions using uniform ordering policies and fixed validation procedures, regardless of urgency, workload type, or contextual importance. While this design supports consistency and security, it limits the effectiveness of blockchain systems in applications where time-critical or trust-sensitive operations must be handled quickly without compromising fairness, accountability, or auditability. This dissertation addresses these limitations by introducing adaptive blockchain mechanisms that combine reputation signals, ledgernative structural controls, and priority-aware scheduling policies to support more responsive transaction processing.

In this dissertation, we propose priority-aware and adaptive blockchain mechanisms for time-sensitive and trust-sensitive environments. The motivation is to improve the responsiveness and fairness of blockchain systems when processing heterogeneous workloads that may differ in urgency, trust level, service requirements, or application context. The goal is to provide flexible blockchain mechanisms that can dynamically distinguish between urgent and ordinary work while preserving the trust
guarantees expected from distributed ledger systems. To achieve this goal, the dissertation builds on a smart-marker-based blockchain design that encodes control semantics directly within the ledger. These markers support operations such as branching, merging, referencing, and priority-aware processing while maintaining auditability and consistency. The proposed approach is grounded in distributed consensus, trustbased reasoning, and fairness-aware scheduling. Rather than treating blockchain
processing as a uniform sequence of transactions, this work frames blockchain opii eration as a decision-making problem shaped by dynamic workload characteristics, reputation signals, and policy constraints. By combining reputation-aware evaluation, ledger-native control structures, and bounded scheduling policies, the proposed system balances responsiveness, fairness, and security. This enables structured adaptation without abandoning the accountability and verification properties that make blockchain systems trustworthy. As a result, the dissertation supports the practical deployment of adaptive blockchain designs across distributed applications where timing, trust, and workload differentiation are critical.

The main research questions in this dissertation are as follows:

How can blockchain mechanisms support multiple concurrent computational outcomes without sacrificing ledger integrity?
Our first contribution proposes a smart-marker-based reputational probabilistic blockchain architecture that enables branching and merging within the ledger.
This design allows multiple agents or algorithms to produce concurrent outcomes, which are preserved as branchchains rather than discarded through premature consensus. Each branchchain is associated with a probabilistic reputation score that reflects historical performance and reliability. This approach enables the system to maintain alternative results while guiding decision-making through trust-aware evaluation. Experimental results demonstrate that the architecture supports multi-agent collaboration and improves interpretability in decision-oriented applications.

How can adaptive blockchain mechanisms be applied in real-world, resource-constrained environments?
Our second contribution evaluates the proposed mechanism through an application study in blockchain-based electronic voting. A smart-marker-based branching architecture is used to execute multiple vote-counting algorithms in parallel, improving robustness and decision reliability. The system is implemented on a resource-constrained platform to assess feasibility under limited computational resources. Experimental results demonstrate that the architecture supports scalable processing, low power consumption, and transparent auditability in practical deployment scenarios.

3. How can blockchain systems process heterogeneous workloads in a fair and efficient manner?
Our third contribution introduces a priority-aware scheduling mechanism for blockchain systems. Transactions are classified based on urgency, service tier, and workload characteristics. A scheduling mechanism is developed to allocate processing resources while enforcing fairness constraints such as bounded delay and anti-starvation guarantees. The system balances faster service for highpriority transactions with continued progress for lower-priority workloads. Evaluation metrics include latency, throughput, and contract satisfaction, demonstrating that the proposed scheduling approach improves responsiveness while maintaining equitable access.

4. How can blockchain systems reduce operational latency for timesensitive applications without weakening consensus guarantees?
Our fourth contribution proposes an early admission mechanism that allows privileged blocks to become partially validated and visible before full validation is complete. Admission decisions are governed by reputation thresholds and partial validation criteria. Smart markers are used to record provisional states and ensure auditability. The mechanism reduces time-to-inclusion while preserving the security guarantees of eventual consensus. A game-theoretic model is planned be developed to analyze adversarial behavior and determine safe operating parameters for early admission policies.

This dissertation advances blockchain design from static, uniform transaction processing toward adaptive, fairness-aware, and application-aware operation. By integrating reputation, structural control, and scheduling mechanisms, the proposed
mechanisms enable blockchain systems to support both ordinary and time-sensitiveworkloads within a unified and trustworthy infrastructure.