150813 research outputs found
Sort by
Nanoparticle-induced lipid membrane deformation influences the design of biomedicine
No full textControlling the physicochemical properties of nanoparticles is important for their performance as drug carriers, pharmaceuticals, or imaging contrast agents in nanomedicine. Predictive models can accelerate experimental designs at reduced time and costs compared to a brute-force approach conventionally used. However, physical principles underlying particle-cell interactions are still poorly understood due to their large size contrast, hindering the model development. In this work, we describe a model that examines the interaction between multiple particles and the membrane of a mammalian cell or an artificial vesicle, thus influencing the outcomes of surface adsorption, detachment or uptake of particles. Compared to existing biophysical models on particle-membrane interactions accounting for membrane adhesion, stretching and bending energies, we make several important updates that are essential to reaching quantitative agreement with existing experimental data. Particle-induced membrane tension changes are crucial to the membrane deformation even at very low surface concentrations (0.1%); we explain this surprising finding using a new length scale previously neglected. Furthermore, a multi-step and non-equilibrium endocytosis mechanism is proposed in the absence of specific receptor-ligand interactions, inspired by recent experimental evidence on the dynamic regulation of membrane tension through the active transport of lipid molecules. We demonstrate the predictive power of our model in generating the adsorption isotherms and shear-induced particle detachment from cell surfaces and the size-dependent rate of particle uptake. Our research provides a framework to design tailor-made nanoparticles with controllable interaction outcomes with various cell types based on a quantitative and fundamental understanding
Belief revision revised
No full textI outline a novel counterexample to the principle ofbelief revision, Anticipation: if both learning andlearning not- would render belief in unjustified, youcannot now be justified in believing . If I am right,not only is the leading theory of belief revision false, soare various recently proposed weakenings. I develop anddefend a new theory that correctly predicts the failuresof Anticipation I argue for, predicated on the simpleidea that one is justified in ruling out possibility just incase that possibility is sufficiently improbable
Biophysical specializations supporting efficiency in neural networks
No full textNeuroscience and artificial intelligence (AI) research have long enjoyed a synergistic relationship. AI has drawn key inspiration from the organization and function of the brain, while our understanding of the biological processes underlying computation has been profoundly enriched by studying the behavior of artificial systems. As breakthroughs in generative AI continue to transform our world, and as the need for more sustainable artificial neural systems becomes more urgent, the neuro-AI feedback loop has never been more important. AI needs ever more powerful and efficient systems, and neuroscience needs further insight into how our brains work. The development of more brain-like AI promises solutions to both of these problems. Unfortunately, this has thus far been stymied by two critical challenges: 1) how do we identify the features that make a system brain-like and 2) how do we incorporate these features into artificial networks in a useful and interpretable way? To address the first of these challenges, I will use the remarkable structural and biophysical diversity of the brain as an introduction into what it means for a system to be “brain-like.” This will lead us to a discussion of dendrites, the tree-like structures implicated at virtually every length scale of neural computation. Dendrites will thereafter act as the focal point for our study of brain-like computation. Specifically, I will trace how relatively simple biophysical features defined at the subcellular level can transform the computational landscape of large networks of neurons. To address the second of these challenges, it is necessary to discuss several enduring problems in computational neuroscience, broken down as chapters in this thesis. In Chapter 2, I will present the development of a new model of single-neuron dynamics that is realistic enough to capture the rich dynamics of dendritic spiking but efficient enough for use in simulations of thousands of neurons, thereby filling a long unmet need in the field. In Chapter 3, I will describe a solution to the general problem of training neural networks with arbitrary differentiable dynamics, thus opening the door for the study of countless biophysical phenomena in the context of networks that can learn to perform computations. In Chapter 4, I will use these tools to test several longstanding hypotheses regarding the utility of different biophysical features in neurons, performing first-of-their-kind fair comparisons of the computational performance of spiking networks, rate-based networks, and networks with nonlinear and linear dendrites. Finally, in Chapter 5, I will use insights gained from studying dendrites at the network level to provide a new perspective as to how the structural and biophysical diversity of the brain could emerge from a complex interplay of functional pressures (e.g., task demands) and physical constraints (e.g., space and energy). Together, the chapters of this thesis outline a general quantitative framework for building more brain-like AI for use in both AI research and neuroscience. This framework illustrates how biophysical specializations arising at the level of single neurons shape the emergent dynamics of the brain.Ph.D
Giant, non-perturbative tuning of light-matter interaction of embedded quantum dots in semiconducting matrices
No full textEmbedding quantum dots (QDs) in a solid-state matrix represents a promising hybrid platform that offers great flexibility and tunability. However, the lack of clear underlying designing principle and presence of large design space make the design process heavily relies on trial-and-error methods. Here we present a new principle that can drastically tailor the light-matter interaction of matrix by matrix-mediated QD interactions. We show that conducting matrices like P3HT can mediate a non-perturbative inter-QD interactions that lead to qualitatively distinct properties, including the enhanced carrier lifetime and enhanced binding energies with increased QD densities, which cannot be explained by conventional perturbative scattering theories and in sharp contrast to independent embedded QDs in an insulating matrix like PMMA. An effective quantum-field-theory is developed, showing qualitative agreement with experiments. Our study serves as a foundation for the predictive design of advanced hybrid materials aimed at optimizing functionalities
Formal Verification of Relational Algebra Transformations in Fiat2 Using Coq
No full textData-intensive applications often involve operations over structured datasets, such as filtering, joining, and projecting records. Relational database systems generally use query planners to optimize high-level SQL queries into efficient execution plans. While these systems apply well-established query transformations, they typically assume the correctness of these transformations rather than formally proving them. The absence of formal guarantees can be a significant limitation for systems with strict correctness requirements. This thesis contributes to Fiat2, a Python-like high-level programming language for data-intensive workloads that integrates formal verification via the Coq proof assistant. We focus on proving the correctness of several rewrite-based query optimizations commonly used in database engines. Specifically, we formalize and prove the correctness of algebraic rewrites involving combinations of filters, joins, and projections, as well as join-reordering rewrites. All rewrites are proven in Coq to preserve the semantics of the original program under list semantics, meaning that the output lists are fully equivalent (or permutations, in the case of join reordering). These verified rewrites serve as a foundation for future optimization in Fiat2, enabling significant optimizations while preserving the semantics of the original queries with correctness guarantees. The results demonstrate the feasibility of integrating formally verified query optimizations into a practical high-level programming language.M.Eng
Towards Fully Automated Volumetric Analysis of Lung Nodules in Computed Tomography
No full textEarly detection of lung cancer significantly improves patient outcomes, and tracking the growth of lung nodules over time is key to understanding their progression and informing future treatment decisions. However, calculating nodule growth in computed tomography (CT) scans remains a highly manual and time-consuming task. In this work, we develop an automated end-to-end pipeline to compute lung nodule growth using state-of-the-art computer vision techniques. While modern advances in deep learning have all but solved many learning tasks in the domain of natural images, biomedical imaging presents unique challenges due to limited data availability, inconsistent annotations, and deployment constraints. We address these challenges by training robust detection and segmentation models using the LUNA16 and LNDb datasets. On the held-out UniToChest dataset, our methods generalize well, attaining a nodule recall of 77.49%, reducing false positives per scan by a factor of 11.3 compared to existing techniques, and achieving a mean nodule-wise Dice score of 0.6453. We then apply our methods to analyze nodule growth in 1,378 patients from the National Lung Screening Trial; we estimate a median nodule volume-doubling time of 791.23 days across all nodules from the patients that do not receive a cancer diagnosis and a median nodule volume-doubling time of 637.38 days across all nodules from the patients that do receive a cancer diagnosis. We also recall 82.20% of radiologist-annotated nodules that are directly associated with a cancer diagnosis and estimate a shorter median nodule volume-doubling time of 370.11 days for these nodules. By automating lung nodule growth quantification, this work lays the foundation for improved screening protocols, personalized treatment planning, and the development of novel imaging biomarkers. To encourage further work in this area, we release our full software pipeline at https://github.com/evanrubel/nodule_volumes.M.Eng
Semiclassical Measures for Complex Hyperbolic Quotients
No full textWe study semiclassical measures for Laplacian eigenfunctions on compact complex hyperbolic quotients. Geodesic flows on these quotients are a model case of hyperbolic dynamical systems with different expansion/contraction rates in different directions. We show that the support of any semiclassical measure is either equal to the entire cosphere bundle or contains the cosphere bundle of a compact immersed totally geodesic complex submanifold. The proof uses the one-dimensional fractal uncertainty principle of Bourgain–Dyatlov (Ann. Math. (2) 187(3):825–867, 2018) along the fast expanding/contracting directions, in a way similar to the work of Dyatlov–Jézéquel (Ann. Henri Poincaré, 2023) in the toy model of quantum cat maps, together with a description of the closures of fast unstable/stable trajectories relying on Ratner theory
Towards an Augmented Reality-based Cyber-Physical Production System Planner
No full textInvestment in automation by small and medium-sized enterprise (SME) manufacturers in the United States has lagged behind their larger counterparts for decades, despite comprising a majority of the nation’s manufacturing industry. The cyber-physical production systems (CPPSs) introduced by Industry 4.0 promise to bolster productivity and efficiency, but only for those enterprises which invest in constituent technologies. These technologies are not easily integrated in existing factories, typically requiring installation of invasive infrastructure and continuous technical support. Robotic integration is typically performed by specialized third-party firms or by in-house staff with extensive technical training, such as engineers. SME manufacturers are particularly sensitive to the complexities of robot integration due to limited access to technologists, and their need for frequent reconfiguration under economies of scope. This thesis introduces Marve: the Mobile Augmented Reality Visual Editor. Marve is a proof-of-concept Android application that enables line workers to directly configure and control an autonomous mobile robot (AMR)-backed hybrid intralogistics system using lowcost consumer hardware. Workers can use Marve’s augmented reality (AR)-based interface to define and visualize the essential geometry and components of such a system. Once configured, workers are able to simulate how the system would respond to their requests to move material throughout the factory. The use of AR enables extensive work to be done at the planning stage of CPPS integration by line workers themselves, bypassing the need for modeling by engineers. Marve relies exclusively on fiducials and visual-inertial odometry (VIO) for localization, and fiducial tags for object tracking, thus eliminating the need for supporting infrastructure. Taken together, these features make Marve an easy on-ramp for SMEs seeking to transition legacy production lines into the CPPSs of Industry 4.0.M.Eng
Solid-State Quantum Memories for Near-Term Quantum Repeaters
No full textOver the past decade, quantum computers have emerged as a promising technology to enable transformational advances in information processing and communication and solve problems that are intractable to classical computers. While there is great promise in linking quantum computers together over long distances via quantum channels, these technologies are still under development. Solid-state emitters with coherent spin-photon interfaces, long spin lifetimes, and narrow optical transitions are a leading platform for use as quantum memories in networked quantum repeaters. However, while such emitters have already enabled advanced quantum networking demonstrations in laboratory settings, applying these devices as useful memory devices at scale is a key outstanding challenge. In this thesis, we experimentally investigate solid-state quantum memories for quantum information applications. First, we develop experimental techniques to characterize solid-state emitters with high throughput, enabling both better understanding of the distribution of emitter properties and improved feedback on material preparation and device fabrication. Next, we implement quantum frequency conversion to create a coherent spin-photon interface between silicon-vacancy centers in diamond and optical photons in the low-loss telecom band. Finally, we investigate color centers in other engineering materials, including silicon and silicon carbide, to better understand the fundamental trade space of requirements for solid-state hosts. Together, these efforts represent a significant advance in creating, controlling, and deploying telecom-compatible spin interfaces, paving the way for memory-enabled quantum repeaters.Ph.D
Generative Latent Motion Planning and Reinforcement Learning for Legged Locomotion
No full textIn recent years, reinforcement learning has demonstrated its promise as a powerful tool for developing innovative and advanced control systems for legged robots. The method’s robustness, versatility, and generality have made it a prime candidate for future robotic systems deployed in the real world. Through the development of more advanced machine learning algorithms and more reliable and efficient physics simulators, reinforcement learning continues to improve and enable new, dynamic, and agile capabilities. While the results are often impressive and the tools relatively beginner-friendly, there remain impediments to scalable and reliable progress. Poor reward function scaling, challenges balancing exploration versus exploitation, and misalignment from the engineer’s intent are roadblocks to better performance. To get beyond these limitations, new tools and frameworks are necessary. In this work, I present novel methods to address these challenges and extend the capabilities of reinforcement learning on robot hardware. Through the quantification of the distributional sim-to-real gap, simulation model optimization for hardware matching, latent space motion sequence planning, and latent style training, I demonstrate never-before-seen performance on legged hardware.Ph.D