137091 research outputs found
Sort by
Elucidating The Phase Morphology-Property Relationship in Polymeric Donor: Acceptor Blends
No full textOrganic photoelectrochemical devices are gaining attraction in the renewable energy community for their printable, flexible alternative to existing hydrogen fuel generation technologies. In particular, novel polymer:polymer blends hold promise as efficient and scalable photo-hydrolysis materials. However, the active-layer phase morphology in polymer:polymer blends is difficult to control, and little research has been done to understand the impact phase morphology has on exciton generation and separation in the electrode layers. By utilizing PBTTT:PCBM as a model system, we demonstrate various methods for establishing control over the heterojunction phase morphology using processing parameters and asymmetrical processing additives. We subsequently extend these learnings to the more complex P3HT:N2200 system, with an eye towards next-generation polymer:polymer blends.M.S.Materials Science and Engineerin
Benchmarking Test-Time DNN Adaptation at Edge with Compute-In-Memory
No full textThe prediction accuracy of deep neural networks (DNNs) deployed at the edge can deteriorate over time due to shifts in the data distribution. For heightened robustness, it’s crucial for DNNs to continually refine and improve their predictive capabilities. However, adaptation in resource-limited edge environments is fraught with challenges: (i) new labeled data might be unavailable; (ii) on-device adaptation is a necessity as cloud connections may be inaccessible; and (iii) the adaptation procedure should prioritize speed, memory efficiency, and energy conservation. Compute-In-Memory (CIM) has recently garnered attention for its computational efficacy and superior operational bandwidth. Additionally, emerging lightweight unsupervised DNN adaptation techniques during test-time have showcased promising results in enhancing model accuracy
for data with noise. This paper pioneers a holistic benchmarking exploration of these methods, assessing their performance and energy efficacy across diverse CIM architectures in edge and autonomous systems. Our findings reveal that the proposed adaptation strategies can adapt to both environment shifts and inherent hardware noise. Engaging in a thorough cross-layer algorithm-hardware-technology co-design space exploration, we highlight pivotal trade-offs among accuracy, performance, and energy for various DNN adaptation
techniques and CIM configurations.M.S.Electrical and Computer Engineerin
A Novel Approach to Describe Transshipment Processes with Consolidation within Nodes: The Column-First Data Model
No full textThis paper proposes a column-first data model as a novel approach to describe transshipment with consolidation for implementing the Physical Internet. The strength of this approach lies in its ability to represent complex existing logistics structures, such as variable container hierarchies and multiple transshipments, and in its flexibility to support the future addition of new data elements. These attributes are essential for constructing a logistics digital twin. As the system infrastructure and applications underpinning the Physical Internet are expected to evolve incrementally, the approach proposed in this study is well positioned to make a meaningful contribution to that progressive development
Investigating the statistical relationship between radar-based measures of subglacial hydrology and model-inferred basal friction at Thwaites Glacier, West Antarctica
No full textIce sheet models use observations to infer basal shear stress, but the variety of methods and datasets available has resulted in a wide range of estimates. Radar-based metrics such as reflectivity and specularity content have been used to characterize subglacial hydrologic conditions that are linked to spatial variations in basal shear stress. We explore whether radar metrics can be used to inform models about basal shear stress. At Thwaites Glacier, West Antarctica, we sample basal shear stress inversions across a wide range of ice sheet models to see how the basal shear stress distribution changes in regions of varying relative reflectivity and specularity content. Our results reveal three key findings: (1) Regions of high specularity content exhibit lower mean basal shear stresses (2) Wet and bumpy regions, as characterized by high relative reflectivity and low specularity content, exhibit higher mean basal shear stresses (3) Models disagree about what basal shear stress should be at the onset of rapid ice flow and high basal melt where relative reflectivity and specularity content are low.UndergraduateEarth and Atmospheric Science
Improving MST Weight Approximation in Sublinear Time
No full textMinimum spanning trees have a variety of applications in computer science as they often form useful sub-structures which aid in computation. Unfortunately, in the modern world, many graph datasets are so intractably large (some even infinite in size) that running a standard MST construction algorithm is impossible. We present an algorithm which, for average degree d, edge weight ratio ω, and error constant ϵ can compute an approximation for the weight of the minimum spanning tree in an general graph in time O(dwϵ−2 log(ωϵ−1)). Additionally as a sub result, we present a deterministic algorithm which maps MST edges to unique graph vertices in expected time O(d log(n)). This result pushes the known upper bound time complexity of the MST weight approximation problem closer to Chazelle, Rubinfeld, and Trevisan’s lower bound of Ω(dωϵ−2).UndergraduateComputer Scienc
Investigation of Hydrogen and Ammonia Auto-ignition in A Shock Tube through Optical Imaging
No full textThis thesis presents a comprehensive study on the auto-ignition characteristics of hydrogen and ammonia mixtures in a high-pressure shock tube using advanced optical imaging diagnostics. Motivated by the discrepancies observed between experimental and theoretical ignition delay times (IDTs), this work addresses the limitations of conventional diagnostic methods by implementing both endwall and sidewall imaging systems. Utilizing high-speed OH* chemiluminescence imaging and pressure-based diagnostics, this study reveals significant insights into ignition homogeneity and axial ignition locations. For the first time at Georgia Tech, endwall and sidewall imaging was employed in a circular cross-section shock tube, enhancing the spatial resolution and diagnostic accuracy. Experimental IDTs for various H₂/O₂/Ar and NH₃/O₂/Ar mixtures were compared against predictions from detailed kinetic models. The findings underscore the critical need for spatially resolved diagnostics in shock tube combustion research and contribute to the development of more accurate chemical kinetic mechanisms for clean energy fuels.M.S.Aerospace Engineerin
Monopoly in the Machines: How Antitrust Can Spur AI Innovation
No full textThis paper examines the evolution of antitrust enforcement in the United States, focusing on the shift from structuralist approaches of enforcement to the new consumer welfare standard. The paper will examine the implications of this transformation for innovation and competition in emerging industries like artificial intelligence (AI). Through an analysis of historical case studies—including the breakups of Standard Oil and AT&T—this paper demonstrates how dismantling monopolies through a structuralist approach can spur innovation in the telecommunication and energy sectors. However, the adoption of the consumer welfare standard, championed by the Chicago School, has narrowed antitrust focus to price effects, overlooking structural harms in concentrated markets, and failing to provide similar outcomes in innovation. Drawing parallels to past antitrust actions, the paper argues for a return to structuralist principles to address monopolistic control in the AI sector. By fostering competitive markets, regulators can unlock technological advancements and ensure long-term industry dynamism
Integrating Machine Learning Techniques for Streamlined Predictive Modeling in Cosmological Applications
No full textThis dissertation advances the integration of machine learning (ML) into computational astrophysics, demonstrating its capacity to streamline cosmological simulations and enhance predictive modeling. The research addresses three interconnected challenges: accelerating radiative transfer (RT) calculations, inferring galaxy properties from early observational data, and emulating stellar feedback in hydrodynamic simulations.
First, we develop ANNgelina, an artificial neural network trained on IllustrisTNG50 and FIREbox simulations, to emulate computationally intensive RT calculations and predict spectral energy distributions (SEDs) of galaxies. By learning the nonlinear relationships between galaxy properties and their SEDs, ANNgelina achieves a median absolute error of 0.06 dex (15%) across UV-to-millimeter wavelengths, bypassing traditional Monte Carlo RT methods while accelerating post-processing by orders of magnitude. Secondly, leveraging early James Webb Space Telescope (JWST) photometric data, we infer stellar masses and star formation histories of high-redshift galaxies. Our models incorporate diverse stellar populations, AGN contributions, and dust attenuation, revealing that stellar-dominated spectra—even with non-standard initial mass functions (IMFs)—best align with observations, yielding stellar masses consistent with initial JWST estimates. Finally, we introduce CosmoConv, a convolutional recurrent network trained on high-resolution SG256 simulation data, to emulate stellar feedback effects in cosmological simulations. By predicting gas density, temperature, and metal enrichment patterns, CosmoConv achieves ~62% accuracy in capturing feedback dynamics, offering a scalable, resolution-independent alternative to subgrid prescriptions.
Collectively, these projects highlight ML’s transformative role in astrophysics, enabling faster exploration of parameter spaces, improved synergy between simulations and observations, and novel insights into galaxy formation and cosmic reionization. The methodologies developed here provide a foundation for future applications in large-scale cosmological modeling, observational data interpretation, and next-generation simulation frameworks.Ph.D.Physic
An Information-geometric Approach to Fluid Dynamical Problems
No full textFor this project, we will derive a regularization for the barotropic Euler equation using information geometry.
The proposed regularization avoids shocks while preserving the long-term behavior of the vanishing viscosity solution. Furthermore, we provide numerical examples to validate our approach.UndergraduateComputer Scienc