DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Investigation of errors and uncertainty in low-frequency impedance tube measurements
No full textAugust2025School of ArchitectureThe accuracy of impedance tube measurements can be improved by determining the acoustic conditions within the impedance tube prior to introducing a sample for test. By measuring the empty impedance tube, the influence of environmental changes on critical parameters can be quantified. Low-frequency dissipation is estimated using Bayesian inference to incorporate into a transfer function model, reducing error when measuring samples. Previous work has focused on a frequency range of 1.5 kHz - 5 kHz; the subject of this study is a lower frequency range (sub-1.5 kHz) using larger sized impedance tubes. The empty tube model is then used to validate the acoustic performance of jute and thermoplastic starch-based microslit panel absorbers. By determining conformity of alternative-material acoustic elements to modeled behavior, and including acoustic testing in the design process, the viability of sustainable material substitutions can be determined.M
Speculative ecoacoustic composition systems: listening to insects through music
No full textAugust2025School of Humanities, Arts, and Social SciencesOngoing insect decline is of critical concern for all species, since insect diversity and resilience are critical to ecosystem function. If we are to rely on anthropocentric rationale alone, this should be ample evidence to pay more attention to insects. However, humans in many western societies have become disconnected from insects, exacerbated by a socialized narrative of fear, avoidance, and exclusion. Parallel to our deteriorating relations with insects there has been a decline in our practices of listening to our environments. This research considers how listening to insects might change the way we perceive and value them – which in turn informs how we interact with and affect them. Composer David Dunn has posited that the production and reception of music can facilitate a connection with beyond-humans through the ‘fabric of mind’ that connects living beings through sound. In this dissertation, I present methods for listening to insects through percussion, electronic, and spatialized music. My focus is on ants and their associates, whose complex societies have tremendous ecological impacts globally. Ants sonically communicate through the architectures in and on which they live, yet there has been very little scientific research on the topic. I posit that we can listen to ants’ mindedness through their aural architectures: the material infrastructure through which they communicate, and by which their sonic agency and social cohesion is realized. Building on Dunn’s artistic works, I present methods for the realization of speculative ecoacoustic composition systems that interact with the sound of insects and our shared environments through recording, playback, and processing. I experiment with composition methods for electronic music, percussion and spatial sound to challenge conceptions of what we can hear, how we listen, and our relations to insects. Percussion and the sounds of our environments are both pushed aside and often labeled “noise” or “experimental” in traditional Western music institutions, yet they are at the same time familiar and embodied, and prompt expansion of what we listen to. My research experiments in bringing these sounds to a broader public through the development and implementation of formats such as performance, public sound installation, in-situ experiences in the field, and web-based media. Through these formats, the work intends to prompt auralization in listeners – composer Pauline Oliveros’ concept of ‘hearing or sounding in the mind,’ through an awareness of cryptic insect sounds in our environments. Building on historical precedents of active listening that empowered social change, and Donna Haraway’s concept of committed sympoietic relationships with beyond-human, I posit that listening through music to insects can foster sonic sympoiesis where we live, sustained by auralization of insect sound in the mind. It is through this affective listening and auralization that we can shift our relations with and challenge our assumptions about insects.Ph
Curation and completion of knowledge graphs
No full textAugust2025School of ScienceKnowledge graphs (KGs) store facts expressed in relationships between entities. Each fact is represented as an ordered triple of \textit{(h,r,t)} or \textit{(head, relation, tail)}. The incompleteness, incorrectness, and sparseness of real-world KGs motivate the task of link prediction or KG completion, where a machine learning model predicts missing entities or relations by learning patterns from a training graph. In this thesis, our contribution focuses on the curation and completion of knowledge graphs, specifically addressing the following challenges. 1. In Chapter 2, we design a novel ontology to capture malware threat intelligence from unstructured text data. The ontology is combined with a framework we developed, TINKER, to instantiate a malware knowledge graph. We show the graph's competency by assessing it using prominent KG completion models.2. In Chapter 3, we consider the inductive KG completion task, where not all entities were seen during training. The most performant models in the inductive setting have employed path encoding modules in addition to subgraph encoding modules. We propose a scalable and efficient alternative to the explicit use of paths. Experimental evaluations on existing inductive KG completion benchmark datasets demonstrate the efficacy of our model.
3. In Chapter 4, we identify gaps in the benchmark datasets and experimental setup of inductive KG completion. We establish a new research direction of realistic and temporally aware inductive knowledge graph completion by providing appropriate datasets and baseline models. Specifically, we propose an algorithm and utilize it to generate datasets that consider temporal progression with varying degrees of entity overlap and independent validation context graphs. We suggest an alternative negative sampling strategy for improved representation learning. Furthermore, we propose two baseline models to address the temporal aspect and entity overlap in the datasets.Ph
Calibration of a hydromechanical discrete-continuum model and liquefaction analysis of a staurated soil system
No full textDecember 2024School of EngineeringThis dissertation presents a calibration of a hydromechanical discrete-continuum model of granular media. The approach emphasizes the critical role of particle-scale characteristics, such as size, shape, and surface texture, in influencing the macroscopic behavior of a granular system. While challenges exist in quantifying these micro-level properties, basing model calibration on such parameters demonstrably yields realistic system responses across diverse loading conditions. This approach facilitates the development of complex and robust numerical models that are easy-to-calibrate. The research highlights the significant influence of shape parameters, like sphericity and roundness, on the shear behavior of granular materials. Traditional contact models, like the Hertz-Mindlin model, can sometimes fall short due to their assumption of perfectly spherical contact surfaces. This limitation is addressed through the implementation of of general models like the J\"{a}ger model, which account for local variations in particle angularity. This allows for a more accurate representation of a wider range of particle conditions. Once calibrated, the model is then employed in constant volume simple shear tests, demonstrating an ability to reproduce liquefaction curves that closely match experimental data. This performance surpasses the accuracy achieved with models employing spherical particles and conventional contact models. Furthermore, the calibrated model is utilized to simulate a sheet pile supporting a dry soil backfill under dynamic loading. The resulting simulations produce realistic outcomes, that accurately capture the results obtained experimentally in centrifuge tests, including lateral displacement and settlement of the backfill soil and rotation of the retaining wall. Finally, this dissertation presents an integration of the discrete model of the particles with an average Navier Stokes model of the pore fluid, resulting in a coupled CFD-DEM model for a saturated soil-sheet pile system. This coupled framework exhibits reasonable accuracy, effectively replicating both the sheet pile response and the dynamics of pore pressure within the soil. Notably, the model maintains consistency across crucial parameters, including coordination number, acceleration, and the stress-strain response. The findings from this model aligns well with established theoretical predictions and existing experimental data.Ph
Stochastic multi-armed bandits: optimality, complexity, robustness, and risk-sensitivity
No full textDecember 2024School of EngineeringMulti-armed bandit problems serve as fundamental models in sequential experimental design, where the goal is to balance exploration and exploitation to maximize cumulative rewards. These problems are pivotal in various real-world applications, including clinical trials, online advertising, and recommendation systems. In the realm of multi-armed bandits, two primary paradigms have emerged: regret minimization, where the objective is to minimize the difference between the rewards obtained and those that would have been obtained by always choosing the best arm, and best arm identification (BAI), aiming to accurately identify the arm with the highest reward mean. This dissertation delves into both paradigms, addressing some unresolved questions spanning across {\em optimal} and {\em efficient} experiment design. Furthermore, the viewpoint of {\em decision safety} and {\em risk sensitivity} is adopted, addressing some of the practical challenges in sequential experimental design. In the context of BAI, existing optimal algorithms have been computationally expensive. These algorithms aim to compute an optimal allocation over the arms by solving a minimax optimization problem in each round, introducing significant computational challenges. Conversely, existing efficient algorithms for BAI have not been optimal. These algorithms allocate a pre-chosen fraction of samples to the best arm, denoted by a tunable parameter , and their optimality depends on selecting an appropriate value of , creating a gap between efficiency and optimality. This dissertation addresses this gap by introducing algorithms that are both optimal and computationally efficient. The efficiency is attributed to {\em implicitly} estimating the optimal value of , without having to solve any optimization problem. Furthermore, while existing literature predominantly concentrates on the exponential distribution family, this approach extends its scope to encompass a broader range of distributions parameterized by their mean values, subject to mild regularity conditions. As the second focus, this dissertation delves into the aspect of {\em decision safety} in stochastic bandits. In applications such as recommender systems, user feedback is often malicious, and in clinical trials, experiment results are susceptible to human errors. These fluctuations in the reward is modeled through an oblivious adversary which is capable of replacing the reward by adversarial samples for a fraction of rounds, also known as the Huber's contamination model. Viewing the problem from a robustness perspective, this dissertation introduces two algorithms: a gap-based algorithm and a successive elimination-based algorithm, Performance guarantees of these algorithms show that these are robust to adversarial contamination, and may achieve optimal sample complexity (up to constant factors). Hence, this dissertation takes a step towards decision safety in practical applications involving experimental design. As the third focus, this dissertation proposes to investigate human risk-taking behavior in decision-making contexts. While most existing algorithms aim to maximize the average reward, this is not always the objective in many human-centric applications. For instance, in high-frequency trading, decision-making revolves around the principle of high-risk and high-reward, in contrast to maximizing the average reward over a period of time. Designing experiments in such risk-sensitive settings involve significantly distinct approaches compared to its canonical risk-neutral counterpart. This dissertation proposes a framework for risk-sensitive bandits and lays down theoretical context and real-world applications in which decision safety is critical to the algorithm design.Ph
Development of peg-mimetic peptides by a combined experimental and computational approach
No full textDecember 2024School of EngineeringPEG is widely used in drug delivery as a “stealth” polymer, where it acts as a protective coating for biologics and drug nanocarriers to improve stability and blood circulation time. It is typically considered to have high biocompatibility and virtually no immunogenicity. However, studies in the past decade have shown that administering PEG in vivo can cause anti-PEG immunity which can lead to hypersensitivity and anaphylactic shock. Another consequence of this immune response is accelerated blood clearance of PEGylated drugs with repeated use, which actually decreases the efficacy of PEGylation in prolonging circulation time. With the increased use of complex biologics and mRNA vaccines, the problem of anti-PEG immunity is expected to grow. Therefore, our main goal is to develop a new class of biocompatible, biodegradable peptide-based materials that can be used as an alternative to PEG. Recent work in our lab has demonstrated that eADF4(C16), a recombinant protein inspired by spider dragline spidroin, exhibits hydrogel-like properties that shield the surface against bacterial attachment. Thus, we hypothesize that spidroin-inspired sequences can be the intellectual foundation for PEG-mimetic peptides. However, the role of specific sequence motifs and the mechanisms of PEG-like behavior in spidroins is currently unknown, hindering rational design of PEG-mimetic peptides. This project will utilize solid phase peptide synthesis, high throughput assays, and machine learning in an integrated, iterative approach. However, this document will focus on the creation of the peptide library and the development of a high-throughput antifouling assay. This is critical to the overall goal, as antifouling will be used as the metric for PEG-mimetic behavior. In addition, due to the large data set needed to train the machine learning algorithm, high-throughput method needs to be developed to evaluate the peptides for antifouling performance.
Peptides were designed based on 10 amino acids (G, A, L, Q, Y, S, P, R, F, N) using 5 amino acids motifs resulting a sequence space of 105 unique peptides. After filtering sequences, 1000 sequences were selected to be synthesized. Further analysis of GRAVY Index, isoelectric point and amino acid positional preferences confirmed there were no bias between the selected list and filtered lists. For the antifouling assay, several plate-based assays were tested using both CelluSpot Membranes and peptide synthesis resins; however, we were unable to measure fouling using those methods. The imaging-based methods were able demonstrate fouling in both the CelluSpot microarrays and the resin-based systems, however, because the base substrates seem to be anti-fouling, further optimizations need to be conducted for this assay, including finding a higher fouling substrate. The use of kosmotropic and neutral salts had a minimal effect on the fouling, in both plate-reader and image-based systems. Additionally, imaging-based methods demonstrated that sequence charge can significantly impact fouling and thus, this will need further consideration in the assay development and sequence design.M
Principled and practical static analysis for python
No full textDecember 2024School of SciencePython is a popular dynamic language for data science and machine learning, and is increasingly adopted for large and complex programming tasks; however, static analysis tools to support software quality and productivity are largely lacking.In programming languages type systems are the most fundamental form of static analysis. Therefore, we first embark on a study to understand the current state of Python type systems. We perform a large-scale empirical study of Python 3 types, usage of types by developers, and two popular static type checking tools, MyPy and PyType. The study shows that Python 3 types are not widely used, and that existing type systems tend to disagree with each other and with user annotations. It demonstrates the need to further study and develop types and type inference analyses as well as deeper and automatic static analyses to support Python development.
The result of the study on Python 3 types motivates the remainder of this thesis. We create a principled approach for static analysis for Python with the initial goal of improving type inference analysis and later applying the concept to other client static analyses.
Our approach to static analysis for Python requires analysis formalization into two phases: (1) syntax and (2) interpretation semantics. Given a specific static analysis, analysis designers first define a subset of Python syntax called PetPy that specifies explicitly what Python constructs are interpreted precisely and what constructs are interpreted approximately. The designers then define interpretation over these syntactic constructs as needed by the specific analysis. The novelty of our approach is the separation of precisely-interpreted constructs from approximately-interpreted ones, thus making analysis approximation, unavoidable in Python, explicit.
Following the principled and practical static analysis approach, we build three novel static analyses for Python: (1) a reference immutability analysis, (2) an interprocedural weakest precondition analysis, and (3) an Andersen-style points-to analysis. Each analysis defines a version of PetPy and interpretation of syntax suited to its needs. The analyses are mostly static, with a novel integration of concrete evaluation.
The analyses scale to production-strength real-world Python codes and largely outperform state-of-the-art techniques and tools. The weakest precondition analysis uncovers bugs in both code and documentation of popular machine learning libraries. We submitted issues and pull requests, which developers promptly merged, to sklearn, TensorFlow, numpy, and IBM’s Lale framework.
In conclusion, our techniques advance static analysis for Python. In future work, we will continue to refine the formalization and build new static analyses for Python.Ph
Enhancing treatment effect estimation using machine learning and large language models
No full textMay2025School of ScienceIn modern healthcare, researchers and clinicians face numerous challenges when determining how well treatments work. Randomized Clinical Trials (RCTs) and Observational Studies are the most popular methods for evaluating treatment effectiveness, but each has limitations. Observational studies are easier to conduct than RCTs, but contain inherent biases that affect their reliability in measuring treatment effects. RCTs, while considered the best scientific approach, are hindered by high costs, long timelines, and difficulties recruiting enough participants. These problems are made worse by two key issues: the growing complexity of selecting the right baseline features needed for valid results, and the frequent lack of diversity among trial participants, which limits how widely the findings can apply to different populations. Recent advances in Large Language Models show potential for automating aspects of clinical trial design, especially baseline feature selection, though careful testing is needed to prevent false information generation. Beyond these research challenges, healthcare systems struggle to identify eligible patients for specialized programs, particularly in complex care management (CCM). The current system relies heavily on doctor referrals, creating bottlenecks that prevent many qualified patients from receiving helpful services. This thesis presents AI and Machine Learning based novel methods that improve treatment effect measurement in both observational studies and clinical trials, while also improving access essential healthcare programs for eligible patients. We begin by tackling the challenge of bias in observational studies, where confounding variables typically distort treatment effect estimates (Chapter 2). We develop a novel hybrid matching algorithm that combines multiple matching techniques to address this issue. Using a type-2 diabetes (T2D) health management program as our test case, we apply causal inference methods to observational data from a regional health insurance provider. Through hybrid matching and survival analysis, we evaluate both T2D onset timing and acute care utilization (emergency and inpatient visits). Our results reveal the program's dual impact: it significantly accelerates early T2D detection while substantially reducing participants' need for acute care services, though it shows no significant effect on T2D onset after the initial two-month period. While our hybrid matching approach proved effective for observational studies, we encountered significant challenges in identifying optimal baseline features for matching treated and control groups to minimize bias. This challenge becomes even more critical in clinical trials, where the stakes are higher and mistakes in feature selection can bias treatment effect estimation and contaminate trial outcomes. In Chapter 3, we address this by exploring the potential of Large Language Models (LLMs) to assist in baseline feature selection for clinical trial design. We create two unique datasets covering nearly 1,700 clinical trials to benchmark how state-of-the-art models like GPT-4o and LLaMa3 can assist researchers in identifying crucial basiline features. Our evaluation process, combining LLM-as-a-Judge and human expert validation, reveals that while state-of-the-art general-purpose LLMs offer substantial benefits in this domain, their performance remains mediocre with significant room for improvement. In Chapter 4, we extend our research to address a critical challenge in clinical trials: achieving both accurate treatment effect measurement and demographic representativeness simultaneously. We introduce ``Framework for Research In Synthetic Control Arms (FRESCA)" to tackle inequity in hybrid clinical trials—studies that combine randomized controlled trial (RCT) participants with historical synthetic controls borrowed from existing trial data or real-world evidence. In this context, equity refers to ensuring trial populations adequately represent the target population for which the treatment is intended, particularly regarding protected attributes like age, gender, and race/ethnicity. Synthetic controls are non-randomized patients selected from historical data to supplement or replace standard control group recruits, potentially reducing trial costs and recruitment challenges. FRESCA implements a dual-stage approach that first employs propensity score matching (PSM) to recommend appropriate synthetic control patients, then applies Iterative Proportional Fitting (IPF) to adjust the population distribution to match target demographic characteristics. We do initial validation of FRESCA using the SPRINT (Systolic Blood Pressure Intervention Trial) data with NHANES (National Health and Nutrition Examination Survey) serving as our target population reference. Initial results demonstrate FRESCA's ability to create more representative hybrid trial populations while maintaining statistical validity of treatment effect estimates. Chapter 5 expands our evaluation of FRESCA across multiple clinical trials, including SPRINT and ALLHAT clinical trials, with NHANES serving as the target population. We compare two new methods that integrate PSM and IPF against other baseline and state-of-the-art methods, confirming the effectiveness of combining propensity and equity adjustments in achieving both accurate and representative treatment effect estimates. Notably, our findings show that even with reduced RCT recruitment supplemented by synthetic controls, these methods maintain accuracy and equity across various trials and outcomes. This chapter also highlights the critical influence of treatment and control group sizes on estimation precision and the importance of balanced synthetic control usage for accurate treatment effect estimations. Finally, in Chapter 6, we address the challenge of healthcare access by examining how patients can be better connected with appropriate healthcare programs. We present a novel end-to-end machine learning approach that simulates the physician's decision-making process for Complex Care Management (CCM) program referrals. Using proprietary Electronic Health Record (EHR) and Claims data from a regional health provider, we tackle challenges such as non-stationarity, dataset imbalance, and partial labeling in time-series data. Our approach successfully identifies additional eligible patients for CCM program referrals who might otherwise be overlooked, providing clear justifications for their selection and potentially bridging the gap between available healthcare resources and patient needs.Ph
Regulation of microtubule bundle mechanics by prc1 in metaphase and anaphase
No full textMay2025School of ScienceThe mitotic spindle is composed of distinct networks of microtubules, including interpolar bundles that can bridge sister kinetochore fibers and bundles that organize the spindle midzone in anaphase. The crosslinking protein PRC1 can mediate such interactions between antiparallel microtubules. PRC1 is a substrate of mitotic kinases including CDK/cyclin-B, suggesting that it can be phosphorylated in metaphase and dephosphorylated in anaphase. How these biochemical changes to specific residues regulate its function and ability to organize bundles is not known. Here, we perform biophysical analyses on microtubule networks crosslinked by two PRC1 constructs, one a wild-type reflecting a dephosphorylated state, and one phosphomimetic construct with two threonine to glutamic acid substitutions near PRC1’s microtubule binding domain. We find that the wild-type construct builds longer and larger bundles that form more rapidly and are much more resistant to mechanical disruption than the phosphomimetic PRC1. Interestingly, microtubule pairs organized by both constructs behave similarly within the same assays. Our results suggest that phosphorylation of PRC1 in metaphase would tune the protein to stabilize smaller and more flexible bundles, while removal of these PTMs in anaphase would favor the assembly of larger more mechanically robust bundles to resist chromosome and pole separation forces at the spindle midzone.In addition to these findings on PRC1’s biochemical regulation during mitosis via phosphorylation, we have begun characterizing the biophysical properties of PRC1 binding using a combination of in vitro experiments and computational simulations to theoretically model protein-protein interactions in the spindle. To achieve this, we are collaborating with mathematicians and physicists to focus on creating models that predict the formation of the mitotic spindle via relevant motor and non-motor crosslinking proteins. A computational model that reflects braking and coasting behaviors exhibited by crosslinked microtubule pairs based on previously published data from our lab has been developed. We find that braking occurs with smaller microtubule separation compared to coasting; the reduced separation between microtubule pairs results in increased resistive forces exerted by PRC1 and thus a reduced sliding speed. The model also shows that higher initial sliding speeds lead to a transition to braking. The results give insight on the relationship between microtubule separation and forces in the spindle exerted by crosslinkers and other MAPs. Furthermore, our collaborative project is currently exploring the possibility of PRC1 cooperativity. Thus far, data on the rate of PRC1 recruitment to single microtubules and overlaps from experiments suggest that a simplified binding model does not sufficiently explain PRC1’s binding behavior, as occupancy effects do not account for the experimental results. We plan on pursuing these findings further, as they may give insight into why PRC1 preferentially binds to antiparallel overlaps compared to single microtubules. The works presented here characterize the behaviors and regulatory mechanisms of the essential human mitotic crosslinker PRC1 via biochemical and biophysical approaches, as well as with structural and computational modeling.Ph
The pursuit of broad-spectrum antivirals: leveraging structural insights into viral proteases
No full textMay2025School of ScienceEmerging viral pathogens, such as SARS-CoV-2, the causative agent of COVID-19, present a global health threat, necessitating preparedness to mitigate future outbreaks. While public health strategies help curb transmission, there is a pressing need for effective broad-spectrum antiviral therapies deployable during the early stages of future outbreaks. Structural bioinformatic analyses reveal significant similarities between the substrate-binding sites of SARS-CoV-2 3C-like protease (3CLpro) and other viral proteases within the PA Clan enzyme superfamily, including key antiviral drug development targets like MERS 3CL protease, Dengue Virus NS2B-NS3 protease, and Hepatitis C Virus NS2/3A protease. Guided by this structural similarity, prior studies from our laboratory have shown that several HCV protease inhibitors can bind and inhibit SARS-CoV-2 3CLpro. In this study, we present results on investigations of cross-inhibition among structurally similar PA Clan proteases, including initial findings from a comprehensive all vs. all screening. We evaluated the cross-inhibitory activities of 18 commercially available inhibitors, together with molecules obtained from collaborators, targeting specific PA Clan proteases against the proteases of SARS-CoV-1, SARS-CoV-2, MERS, Polio, Dengue, West Nile and Zika viruses. Proteases from these viruses were cloned, expressed, and purified, and fluorescence- and NMR-based enzyme assays were developed. Using this approach, we have successfully identified several novel viral protease inhibitors of MERS, Polio, Dengue, West Nile and Zika virus proteases. Our studies provide novel directions for antiviral drug development targeting a wide range of pathogenic viruses with pandemic potential.Ph