Thesis defence

Master’s Thesis Presentation • Artificial Intelligence • Demystifying Foreground-Background Memorization in Diffusion Models

Mayuri Punithan — Wed, 20 Aug 2025 19:41:27 +0000

Master’s Thesis Presentation • Artificial Intelligence • Demystifying Foreground-Background Memorization in Diffusion Models Mayuri Punithan Wed, 08/20/2025 - 15:41

Please note: This master’s thesis presentation will take place in DC 2310 and online.

Jimmy Di, Master’s candidate
David R. Cheriton School of Computer Science

Supervisors: Professor Gautam Kamath

Diffusion models (DMs) memorize training images and can reproduce near-duplicates during generation. Current detection methods identify verbatim memorization but fail to capture two critical aspects: quantifying partial memorization occurring in small image regions, and memorization patterns beyond specific prompt-image pairs.

To address these limitations, we propose Foreground Background Memorization (FB-Mem), a novel segmentation-based metric that classifies and quantifies memorized
regions within generated images. Our method reveals that memorization is more pervasive than previously understood:

Individual generations from single prompts may be linked to clusters of similar training images, revealing complex memorization patterns that extend beyond one-to-one correspondences
Existing model-level mitigation methods, such as neuron deactivation and pruning, fail to eliminate local memorization, which persists particularly in foreground regions.

Our work establishes an effective framework for measuring memorization in diffusion models, demonstrates the inadequacy of current mitigation approaches, and proposes a stronger mitigation method using a clustering approach.

Join by:

In-person: Go to DC 2310
Online: Zoom

Master’s Thesis Presentation • Data Systems • Conjunctive Queries with Negations: Bridging Theory and Practice

Mayuri Punithan — Tue, 19 Aug 2025 14:46:26 +0000

Master’s Thesis Presentation • Data Systems • Conjunctive Queries with Negations: Bridging Theory and Practice Mayuri Punithan Tue, 08/19/2025 - 10:46

Please note: This master’s thesis presentation will take place in DC 2314.

Boyi Li, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Xiao Hu

Antijoin, given its great expressive power, sees many applications in relational data analytics. Notwithstanding its importance, there remain great research potentials in antijoin processing. In practical database systems, existing techniques to process antijoins are still considered rudimentary, building upon heuristics and cost-based optimization strategies that offer no theoretical guarantees. Meanwhile, the database theory community has proposed algorithms for antijoins with strong theoretical guarantees, yet these algorithms build upon specialized, complicated data structures and have not made their way to practice. In light of such gap between theory and practice, we propose new algorithms for antijoin processing in this thesis. Not only do our new algorithms provide the same theoretical guarantees as the state-of-the-art algorithm, but they also use only basic relational operations. The latter property enables our new algorithms to be rewritten in basic SQL statements, allowing an easy, system-agnostic integration into any SQL-based database system. We then empirically evaluate one of our new algorithms, rewritten in SQL, over real-life graph datasets with a variety of SQL database systems. Experimental results show order-of-magnitude improvements of our new algorithm over vanilla SQL queries.

Master’s Research Paper Presentation • Cryptography, Security, and Privacy (CrySP) • Replicating AceDroid: A Path-Sensitive Normalization Approach for Android Security Check Modeling

Mayuri Punithan — Tue, 19 Aug 2025 14:19:32 +0000

Master’s Research Paper Presentation • Cryptography, Security, and Privacy (CrySP) • Replicating AceDroid: A Path-Sensitive Normalization Approach for Android Security Check Modeling Mayuri Punithan Tue, 08/19/2025 - 10:19

Please note: This master’s research paper presentation will take place in DC 2314.

Adeola Tijani, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Yousra Aafer

The Android framework enforces access control through various conditional checks. However, these checks are inconsistently applied, especially across customized vendor frameworks. AceDroid was developed to detect and normalize these diverse access control checks into a canonical form to identify security inconsistencies. In this work, we replicate AceDroid’s key findings, implement path-sensitive analysis based on interprocedural control flow graphs (ICFGs), and enhance the normalization logic to capture both explicit and implicit security checks. Our experiments confirm AceDroid’s insights while proposing improvements in data dependency modeling and compound condition detection. We successfully replicate canonical security conditions across several Android system methods and offer a critical assessment of equivalence check modeling. This paper serves as a reproducibility and enhancement effort of AceDroid, contributing new findings on check modeling fidelity and performance tuning.

Master’s Thesis Presentation • Systems and Networking • Combining High-Level-Synthesis and Register-Transfer Level Design for Programmable Hardware

Mayuri Punithan — Thu, 14 Aug 2025 14:41:18 +0000

Master’s Thesis Presentation • Systems and Networking • Combining High-Level-Synthesis and Register-Transfer Level Design for Programmable Hardware Mayuri Punithan Thu, 08/14/2025 - 10:41

Please note: This master’s thesis presentation will take place in DC 2314.

Kimiya Mohammadtaheri, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Mina Tahmasbi Arashloo

As network speeds continue to grow to hundreds of Gbps and beyond, offloading transport-layer functionality to programmable Network Interface Cards (NICs) has gained traction for improving performance and enabling higher-level offloads. FPGAs on NICs offer low latency and high throughput but are notoriously difficult to program. To reduce developer effort for hardware transport, prior works propose hardware architectures with reusable fixed-function modules for common transport data structures and operations, and programmable modules for protocol-specific operations. However, they either still require intricate Verilog programming for their programmable modules or are tied to specific protocols or pipeline abstractions.

In this work, we explore the design of a programmable hardware architecture for transport, called HTraP, that:

offers a simple, intuitive programming interface without requiring detailed understanding of the rest of the architecture for effective programming
supports a broad range of transport protocols with varying levels of complexity

HTrap has one main programmable module that can be programmed to capture the core protocol logic in simple C code that is amenable to automated, efficient hardware generation with HLS. The generated hardware can then plug into the rest of HTraP which implements complex transport operations through a protocol-agnostic interface.

PhD Defence • Computer Graphics • Toward General-Purpose Monte Carlo PDE Solvers for Graphics Applications

Mayuri Punithan — Tue, 12 Aug 2025 15:08:43 +0000

PhD Defence • Computer Graphics • Toward General-Purpose Monte Carlo PDE Solvers for Graphics Applications Mayuri Punithan Tue, 08/12/2025 - 11:08

Please note: This PhD defence presentation will take place in DC 2314 and online.

Ryusuke Sugimoto, PhD candidate
David R. Cheriton School of Computer Science

Supervisors: Professors Toshiya Hachisuka and Christopher Batty

This thesis develops novel Monte Carlo methods for solving a wide range of partial differential equations (PDEs) relevant to computer graphics. While traditional discretization-based approaches efficiently compute global solutions, they often require expensive global solves even when only local evaluations are needed, and can struggle with complex or fine-scale geometries. Monte Carlo methods based on the classical Walk on Spheres (WoS) approach [Muller 1956] offer pointwise evaluation with strong geometric robustness, but in practice, their application has been largely limited to interior Dirichlet problems in volumetric domains. We significantly broaden this scope by designing versatile Monte Carlo solvers that handle a diverse set of PDEs and boundary conditions, validated through comprehensive experimental results.

First, we introduce the Walk on Boundary (WoB) method [Sabelfeld 1982, 1991] to graphics. While retaining WoS’s advantages, WoB applies to a broader range of second-order linear elliptic and parabolic PDE problems: various boundary conditions (Dirichlet, Neumann, Robin, and mixed) in both interior and exterior domains. Because WoB is based on boundary integral formulations, its structure more closely parallels Monte Carlo rendering than WoS, enabling the application of advanced variance reduction techniques. We present WoB formulations for elliptic Laplace and Poisson equations, time-dependent diffusion problems, and develop a WoB solver for vector-valued Stokes equations. Throughout, we discuss how sampling and variance reduction methods from rendering can be adapted to WoB.

Next, we address the nonlinear Navier--Stokes equations for fluid simulation, whose complexity challenges Monte Carlo techniques. Employing operator splitting, we separate nonlinear terms and solve the remaining linear terms with pointwise Monte Carlo solvers. Recursively applying these solvers with timestepping yields a spatial-discretization-free method. To deal with the resulting exponential computational cost, we also propose cache-based alternatives. Both vorticity- and velocity-based formulations are explored, retaining the advantages of Monte Carlo methods, including geometric robustness and variance reduction, while integrating traditional fluid simulation techniques.

We then propose Projected Walk on Spheres (PWoS), a novel solver for surface PDEs, inspired by the Closest Point Method. PWoS modifies WoS by projecting random walks onto the surface manifold at each step, preserving geometric flexibility and discretization-free, pointwise evaluation. We also adapt a noise filtering technique for WoS to improve PWoS.

Finally, we outline promising future research directions for Monte Carlo PDE solvers in graphics, including concrete proposals to enhance WoB.

To attend this PhD Defence in person, please go to DC 2314. You can also attend virtually on Zoom.

Master’s Thesis Presentation • Artificial Intelligence • Instance Segmentation with Occlusion Order Supervision: Two Problems

Mayuri Punithan — Mon, 11 Aug 2025 19:59:36 +0000

Master’s Thesis Presentation • Artificial Intelligence • Instance Segmentation with Occlusion Order Supervision: Two Problems Mayuri Punithan Mon, 08/11/2025 - 15:59

Please note: This master’s thesis presentation will take place online.

Cheuk-To (Jeremy) Yu, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Olga Veksler

Joint architectures for predicting instance masks with additional depth-related information have shown improvements in performance on both the original segmentation tasks and their corresponding depth-related tasks. Closely related to depth, occlusion can also provide strong cues for segmentation. Although occlusion ordering between instances provides less supervision than full pixel-wise depth, occlusions can be easily annotated for existing datasets. Motivated by the above, we propose two problems that incorporate occlusion information into standard segmentation tasks with their corresponding methods. For both methods, we explore the level of supervision occlusion provides in terms of segmentation accuracy by comparing our results with baseline segmentation approaches trained without occlusion supervision.

Firstly, we develop an end-to-end framework to perform instance segmentation and per-instance global occlusion order prediction simultaneously by appending a global occlusion order head to standard instance segmentation architectures. Our approach to occlusions differs from most prior work. Prior work performs occlusion estimation in a local pairwise manner: given two instances, classify one of them as the occluder. Due to locality, for scenes known not to have occlusion cycles, such approaches can (and do) produce occlusion cycles, which are errors. Unlike most prior work, we directly label instances with their occlusion-order labels, where an instance with a larger label occludes any neighboring instances with smaller labels. This approach is cycle-free by design. Using cross-entropy with occlusion-order labels fails as occlusion-order labels do not have a fixed semantic meaning. Therefore, we develop a novel regularized loss function for successful training. Our framework achieves high occlusion-order accuracy with improved performance in instance segmentation, likely due to the added supervision.

Secondly, we develop a new direction for occlusion-supervised amodal instance segmentation (AIS). AIS is an emerging task that segments the complete object instance, both the visible and occluded parts. All prior work for AIS can be divided into two groups: (i) methods constructing a synthetic amodal dataset; (ii) methods using human-annotated amodal masks. The drawback of methods in the first group is that constructing a realistic synthetic dataset is difficult. The drawback of methods in the second group is that human annotation is prone to error, as humans must reason about invisible regions. Our method requires neither a synthetic dataset nor human-annotated amodal masks, but instead uses ground-truth modal masks and the pairwise occlusion order between instances. By using the occlusion order to reason about the visible and invisible parts of the amodal mask, we develop effective loss functions for the visible and occluded parts of the predicted mask. We achieve comparable accuracies to those of the same architecture trained on human-annotated amodal masks, despite not using amodal masks for training.

Attend this master’s thesis presentation virtually on Zoom.

Master’s Thesis Presentation • Algorithms and Complexity • Efficient Algorithm with No-Regret Bound for Sleeping Expert Problem

Mayuri Punithan — Mon, 11 Aug 2025 19:38:44 +0000

Master’s Thesis Presentation • Algorithms and Complexity • Efficient Algorithm with No-Regret Bound for Sleeping Expert Problem Mayuri Punithan Mon, 08/11/2025 - 15:38

Please note: This master’s thesis presentation will take place in DC 2314.

Junhao Lin, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Ian Munro

The sleeping experts problem is a variant of decision-theoretic online learning (DTOL) where the set of available experts may change over time. In this thesis, we study a special case of the sleeping experts problem with constraints on how the set of available experts can change. The benchmark we use is ranking regret, which is a common benchmark used in sleeping experts problem. Previous research shows that achieving sublinear ranking regret bound in the general sleeping experts problem is NP-hard, so we relax the sleeping experts problem by imposing constraints on how the set of available experts may change. Under those constraints, we present an efficient algorithm which achieves a sublinear ranking regret bound.

Master’s Research Paper Presentation • Quantum Computing • On the Complexity of Quantum One-shot Compression

Joe Petrik — Wed, 06 Aug 2025 21:42:55 +0000

Master’s Research Paper Presentation • Quantum Computing • On the Complexity of Quantum One-shot Compression Joe Petrik Wed, 08/06/2025 - 17:42

Please note: This master’s research paper presentation will take place online.

Nicholas Allen, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Ashwin Nayak

Many important tasks in quantum information processing involve synthesizing quantum states or implementing unitary transformations. Traditional complexity theory, which typically addresses problems with classical inputs and outputs, does not readily capture the computational hardness of these inherently quantum tasks. One such task is the quantum one-shot compression problem, where the goal is to compress a single, unknown quantum state drawn from a known ensemble using as few qubits as possible.

In this work, we investigate the computational hardness of the quantum one-shot compression problem using various tools from complexity theory and information theory. First, we begin by upper bounding the unitary complexity of the problem via a reduction to the Uhlmann transformation problem, before discussing lower bounds based on pseudorandom state generators and other quantum cryptographic primitives. Next, we describe algorithms for the task under different error criteria based on semidefinite programming, assuming complete classical descriptions of the source. Afterwards, we attempt to derive structural characterizations of compression protocols, with connections to partially entanglement breaking channels and the Knill-Laflamme conditions for quantum error correction. Finally, we explore generalizations to related tasks such as state redistribution before showing that the problem of testing “closeness” to product states is strongly NP-hard. Our results provide new insight into the complexity and structure of quantum compression in the one-shot setting.

Attend this master’s research paper presentation virtually on Zoom.

Master’s Thesis Presentation • Human-Computer Interaction • DeTAILS: Deep Thematic Analysis with Iterative LLM Support

Joe Petrik — Thu, 31 Jul 2025 16:37:59 +0000

Master’s Thesis Presentation • Human-Computer Interaction • DeTAILS: Deep Thematic Analysis with Iterative LLM Support Joe Petrik Thu, 07/31/2025 - 12:37

Please note: This master’s thesis presentation will take place in DC 2314 and online.

Ansh Sharma, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Jim Wallace

Reflexive thematic analysis (TA) yields rich insights but is challenging to scale to large datasets due to the intensive, iterative interpretation it requires. We present DeTAILS: Deep Thematic Analysis with Iterative LLM Support, a researcher-centered toolkit that integrates large language model (LLM) assistance into each phase of Braun & Clarke’s six-phase reflexive TA process through iterative human-in-the-loop workflows. DeTAILS introduces key features such as “memory snapshots” to incorporate the analyst’s insights, “redo-with-feedback” loops for iterative refinement of LLM suggestions, and editable LLM-generated codes and themes, enabling analysts to accelerate coding and theme development while preserving researcher control and interpretive depth.

In a user study with 18 qualitative researchers (novice to expert) analyzing a large, heterogeneous dataset, DeTAILS demonstrated high usability. The study also showed that chaining LLM assistance across analytic phases enabled scalable yet robust qualitative analysis. This work advances Human-LLM collaboration in qualitative research by demonstrating how LLMs can augment reflexive thematic analysis without compromising researcher agency or trust.

To attend this master’s thesis presentation in person, please go to DC 2314. You can also attend virtually on MS Teams.

Master’s Thesis Presentation • Data Systems • Distributed Eventual Durability

Joe Petrik — Tue, 29 Jul 2025 14:00:41 +0000

Master’s Thesis Presentation • Data Systems • Distributed Eventual Durability Joe Petrik Tue, 07/29/2025 - 10:00

Please note: This master’s thesis presentation will take place online.

Joseph Boulis, Master’s candidate
David R. Cheriton School of Computer Science

Supervisor: Professor Sujaya Maiyya

We present the first design and implementation of a Distributed Eventually Durable transactional system. Durability latency often dominates transaction commit time in distributed databases, limiting performance for latency-sensitive applications. The Eventual Durability (ED) model (VLDB’24) addresses this by decoupling durability from commit, allowing transactions to commit quickly and persist in the background. So far, the ED model has been applied only to centralized databases. Extending ED to distributed, multi-shard settings introduces new challenges, as each shard persists changes independently potentially making partial changes of a transaction durable while other changes may be lost due to failures.

This work addresses these challenges effectively by first defining a new correctness criterion, ED Snapshot Isolation (ED-SI), which ensures that transactions only observe the effects of non-failed committed transactions. To enforce ED-SI, we develop a distributed commit protocol that builds on Percolator and adds validation to prevent reads from failed transactions. Our prototype of this system, called ED Percolator, deployed on AWS, supports both fast (speculative) and safe (durable) transactions to allow applications a latency vs. durability trade off while preserving formal correctness. Experiments show that ED fast transactions offer up to 7.6x speed-up over classical Percolator transactions, while ED safe transactions incur similar latency but abort significantly fewer transactions.

Attend this master’s thesis presentation virtually on Zoom.