12023 research outputs found
Sort by
I. Development of Nickel- and Palladium-Catalyzed Asymmetric Allylic Alkylation Reactions. II. Enantioselective Syntheses of Tetrahydroisoquinoline–Based Natural Products and Unnatural Analogs
Get PDF Described in this thesis are four projects related to the development of synthetic methodologies for the preparation of enantioenriched building blocks, and the total syntheses of complex tetrahydroisoquinoline natural products. In Chapter 1, the development of nickel-catalyzed asymmetric allylic alkylation of lactones and lactams with allylic alcohols is presented. In Chapter 2, the development of palladium-catalyzed enantioselective decarboxylative allylic alkylation of silicon-containing heterocycles is detailed. In these chapters, the utilization of prochiral enolates as nucleophiles has enabled access to enantioenriched all-carbon quaternary stereocenters.
Chapter 3 describes the total syntheses of bis-tetrahydroisoquinoline alkaloids,
(–)-jorumycin and (–)-jorunnamycin A. A general synthetic strategy, which exploits the tandem cross-coupling/hydrogenation approach, represents the first non-biomimetic synthetic route and allows for an efficient construction of the pentacyclic core in a highly modular fashion. Additional bis-tetrahydroisoquinoline analogs were prepared, and preliminary studies to probe their cytotoxicity against cancer cell lines were conducted. Finally, an extension of the enantioselective and diastereoselective hydrogenation technology to include simple 1,3-disubstituted isoquinolines is described in Chapter 4.</p
Safe and Interpretable Autonomous Systems Design: Behavioral Contracts and Semantic-Based Perception
Get PDFWe are on the verge of experiencing a new, integrated society where autonomous vehicles will become a fabric of our everyday lives. And yet, seamless integration of autonomous vehicles into our society will require vehicles to interface safely with humans in an incredibly complex, fast-paced, and dynamic environment. Premature deployment of these new autonomous systems — without safety guarantees or interpretability of algorithms, could prove catastrophic. How can algorithms governing vehicle behavior be designed in a way that guarantees safety, performance, interpretability and scalability? This is the question this thesis seeks to answer.
First, we present a framework for architecting the decision-making module of autonomous vehicles so that safety and progress of agents can be formally guaranteed. In particular, all agents are defined to act according to what is termed an assume-guarantee contract, which is broadly defined as a set of behavioral preferences. The first version of the assume-guarantee contract is a behavioral profile, which is a set of ordered rules that agents must use to select actions in a way that is interpretable. With all agents operating according to a behavioral profile, the interactions however, are not necessarily coordinated. We then constrain agent behavior with an additional set of interaction rules. The behavioral profile combined with these additional constraints, are what we term a behavioral protocol. With all agents operating according to a local, decentralized behavioral protocol, we can provide formal proofs of the correctness of agent behavior, i.e. all agents will never collide and agents will make it to their respective destinations. Not only does the protocol so defined allow us to make formal guarantees, but it is also designed in a way that scales well in the number of agents and provides interpretability of agent behaviors. Safety and progress guarantees are proven and verified in simulation.
Second, we focus on using information from object classifiers to enhance an autonomous vehicle's ability to localize where it is within its environment. The proposed approach for incorporating this semantic information is based on solving the maximum likelihood problem. With a hierarchical formulation, we are not only able to improve upon the accuracy of traditional localization techniques, but we are also able to improve our confidence in the accuracy of object detection classifications. The improvement in robustness and accuracy of these algorithms are shown in simulation.</p
Learned Feedback & Feedforward Perception & Control
Get PDFThe notions of feedback and feedforward information processing gained prominence under cybernetics, an early movement at the dawn of computer science and theoretical neuroscience. Negative feedback processing corrects errors, whereas feedforward processing makes predictions, thereby preemptively reducing errors. A key insight of cybernetics was that such processes can be applied to both perception, or state estimation, and control, or action selection. The remnants of this insight are found in many modern areas, including predictive coding in neuroscience and deep latent variable models in machine learning. This thesis draws on feedback and feedforward ideas developed within predictive coding, adapting them to improve machine learning techniques for perception (Part II) and control (Part III). Upon establishing these conceptual connections, in Part IV, we traverse this bridge, from machine learning back to neuroscience, arriving at new perspectives on the correspondences between these fields.</p
Artificial Neural Networks for Nonlinear System Identification of Neuronal Microcircuits
Get PDFThis thesis explores the application of artificial neural networks (ANNs) to nonlinear system identification. We use neuronal microcircuits in the retina as a testbed for our technique, which relies upon the marriage of partial anatomical information with large electrophysiological datasets. Rather than a typical application of machine learning, our primary goal is not to predict the output of retinal circuits, but rather to uncover their structure. We begin with a theoretical exploration in a toy problem and provide a proof of unique identifiability under a specific set of conditions. We then perform empirical simulations in a number of different circuit architectures and explore the space of constraints and regularizers to demonstrate that this technique is feasible in a hyperparametric regime that lends itself well to neuroscience datasets. We then apply the technique to mouse retinal datasets and show that we can both recover known biological information as well as discover new hypotheses for biological exploration. We end with an exploration of active stimulus design algorithms to distinguish between circuit hypotheses.</p
Deployment Dynamics of Thin-Shell Space Structures
Get PDFThin-shell structures provide a lightweight solution for deployable structure applications. Despite being only few tens of microns thick, these structures provide excellent bending stiffness, thanks to their curved cross-section. Their thinness also allows them to be elastically packaged into small volumes to fit into a launch vehicle; once in space, they can be self-deployed by releasing their stored elastic energy.
Most space applications use thin-shell structures to deploy and tension thin membranes, such as solar sails, drag sails, and solar arrays. Recently, a novel space solar power architecture has been developed at Caltech, and it relies on distributed thin-shell components, connected in a space frame, to create large-area deployable structures. Thanks to the unique properties of thin shells, these structure provide superior stiffness-to-mass ratio and self-deployment capabilities. However, to demonstrate their reliability and enable their use on space missions, their deployment dynamics must be understood and predicted.
Ground testing is the established approach to verify a structure throughout its design and qualification process. However, replicating the space environment in a laboratory setting is generally not possible, especially for lightweight structures, which are very sensitive to the effects of gravity and air. Numerical models are therefore the only tool to predict the behavior of a structure in space. However, validation with ground experiments is necessary to build confidence in the models, which must be able to capture the complexity of the interaction with air, gravity, and the suspension system that supports the weight of the structure.
The goal of this thesis is to develop high-fidelity models for large space structures, where multiple thin-shell components are folded together and deploy by releasing their strain energy. This overall objective is achieved in 3 steps. First, a ladder-type rectangular strip is introduced, as a building block for more complex architectures. The strip is composed by two thin-shell longerons, symmetrically folded at two locations. The deployment dynamics of this structure is investigated through experiments on 1 m-scale prototypes, both in air and in vacuum. A detailed analysis of its elastic folds is performed using full-field displacement measurements from Digital Image Correlation. A finite element model of this strip is presented, and it is shown to accurately capture the dynamics of the strip for all tested conditions. Then, the implementation of the packaging and deployment scheme of a space solar power spacecraft, composed of multiple strips, is discussed. A kinematic model of the structure is proposed as a design tool to achieve systematic folding. A novel concept of a deployment mechanism to coil the structure in a robust and reliable way is proposed. Also, a staged deployment scheme is demonstrated, to reduce the uncertainty of strain-energy deployment for large space structures. Finally, the deployment dynamics of a 2 m-scale space structural prototype, based on the space solar power architecture, is investigated. A full-scale finite element model of the structure is implemented to replicate its complex folding scheme and capture the deployment process, including the interaction with the deployment mechanism and the suspension system. The simulations predict well the behavior of the structure observed in experiments through motion capture techniques.
The work presented in this thesis advances previous studies on the deployment dynamics of simple thin-shell components, and demonstrates that even complex thin-shell architectures can be packaged and deployed in a controlled and predictable way. The solutions proposed in this thesis have guided the packaging process and the design of the deployment mechanism for DOLCE, an upcoming flight demonstration of the space solar power architecture described in this work. However, this research has much broader implications, as the experimental and numerical framework presented herein can be generalized to different shell-based architectures, and contributes to enabling a new generation of lightweight deployable structures for future space applications.</p
A Combined High Temperature Tape-Casting and 2D Freeze-Casting Method for Polymer Membranes with Vertically Oriented Pores
No full textAlternatives to lithium-ion batteries often require novel electrolytes, such as solvent-in-salt electrolytes and ionic liquids, for which commercial battery separators are unsuited. The Faber Lab at Caltech is researching processing methods for separators made of polymer/ceramic composite materials to ensure high wettability in novel liquid electrolytes. To this end, a flexible polymer separator is needed which provides a short working-ion transportation path for high ionic conductance. Here, a method was developed for the fabrication of poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) membranes with vertically aligned, nontortuous pores with pore diameters of a few microns by directional crystallization of dimethyl sulfone (DMSO₂). The method combines high-temperature tape casting
and bidirectional freeze casting into a single step. Important considerations for the improvement of the method (particularly, avenues for reducing the pore size to below 1 μm) are discussed. In addition, the role of PVDF-HFP crystallization in bidirectional freeze-casting, a solvent crystallization method, is discussed
Convergent Synthetic Strategies toward Heterodimeric Bisindole Alkaloids and Polyoxygenated Diterpenoids
Get PDFDescribed herein are two projects in the field of natural product synthesis unified by their use of convergent strategies. An introduction into a relevant subclass of natural products, the bis(monoterpenoid) indole alkaloids, precedes our synthetic efforts. The molecules in this class are comprised of two monoterpenoid indole alkaloids conjoined by at least one carbon–carbon bond, and we review efforts to construct these dimers using semi-, partial, and total synthesis.
The account of our synthetic work begins with a detailed approach to the bis(monoterpenoid) indole alkaloid leucophyllidine. An enantioselective Pd-catalyzed decarboxylative allylic alkylation generates an α-quaternary-substituted lactam, which serves as a building block for both monomeric subunits. The northern fragment, eburnamonine, is constructed through a five-step sequence comprised of Fischer indole synthesis, Bischler–Napieralski cyclization, and diastereoselective hydrogenation. The southern fragment, eucophylline, is constructed through a ten-step formal synthesis comprised of a Friedländer quinoline synthesis, followed by two orthogonal C–H functionalizations that each displayed unexpected reactivity.
We then describe the evolution of a convergent coupling strategy to unify the two polycyclic fragments. While the "biomimetic" Friedel–Crafts and "bio-inspired" organometallic addition approaches failed, a Pd-catalyzed cross-coupling was ultimately successful in forging the key C–C bond. Extensive efforts to install the final stereogenic center with a variety of reducing agents were unsuccessful, and DFT modeling was utilized to probe the recalcitrant nature of the trisubstituted alkene. Preliminary investigations of a directed hydrogenation are then discussed.
Finally, we report an approach to the first total synthesis of the polyoxygenated diterpenoid (–)-scabrolide A. The route begins with the synthesis of an enantioenriched cyclopentendiol building block and an acyclic diyne from (R)-linalool and (R)-carvone, respectively. A Stieglich esterification and thermal [4+2] cycloaddition affords a tricylic intermediate bearing all 19 carbons observed in the natural product. The cycloheptenoid motif is installed through a photochemical [2+2]/fragmentation sequence, exploiting an unusual alkene protecting group strategy to counteract unexpected reactivity.</p
Controlling and Calibrating Interferometric Gravitational Wave Detectors
Get PDFIn September 2015, the Advanced LIGO detectors made the first direct detection of gravitational waves from a binary black hole merger [1]. Since then, around fifty total gravitational wave detections have been reported by Advanced LIGO and Advanced Virgo over three dedicated gravitational wave observation times, known as observing runs.
Observing run three (O3) ran from April 2019 to March 2020, with higher sensitivity and more stable operation of the Advanced LIGO detectors [2]. In the first half of O3, thirty-nine gravitational wave events were detected [3], as opposed to eleven in all of observing runs one (O1) and two (O2) [4]. The higher rate of detections is due primarily to the increased detector sensitivity to gravitational waves.
Although the Advanced LIGO detectors are more sensitive to gravitational waves than any detector in history, they have not yet achieved design sensitivity. Work continues to push the detectors to their fundamental limit of sensitivity. The work in this thesis partially covers the effort to improve the sensitivity of the LIGO Hanford detector prior to O3.
Calibration of the Advanced LIGO interferometer is the conversion of raw detector data into gravitational wave strain data. This process is crucial to an accurate and precise understanding of astrophysical sources of gravitational waves. The calibration uncertainty pipeline for characterizing the strain uncertainty during O1 and O2 is discussed in detail [5].
This thesis covers topics in long-baseline interferometric gravitational wave detector technology, including an overview of the performance of the detector in O3, commissioning tasks done to increase the sensitivity of the detector for O3, overall calibration uncertainty in the gravitational wave data, and methods for robust estimation of spectral quantities from LIGO data.</p
New Tensor Network Methods and Studies of Criticality in Low-Dimensional Quantum Systems
Get PDFSeveral investigations are presented around the general topic of the ground state and low-energy behavior of models for many-body quantum physics in one dimension (1d). We develop a novel numerical method for the ground and low-energy sectors of local Hamiltonians in 1d which is based on proofs from quantum information theory. This method, the rigorous renormalization group (RRG), enjoys the benefits of explicit global information from the Hamiltonian in its local step, allowing it to avoid spurious convergence in systems with challenging energy landscapes. We apply RRG to the random XYZ spin chain in an unbiased numerical study evaluating infinite-randomness fixed point physics and continuously varying critical exponents in the ground state, finding evidence for both. In a related effective model with correlations preventing the exact solution of the strong-disorder renormalization group equations, we use the framework of random walks to rigorously establish continuously varying critical exponents. We also perform detailed studies of deconfined quantum critical points (DQCP) in 1d, providing strong evidence for phase transitions which display similar phenomenology to the canonical examples in 2d. A family of DQCP phase transitions in 1d is exhibited which appears to controlled by complex fixed points corresponding to a walking scenario for renormalization group flows
Structured Signal Recovery from Nonlinear Measurements with Applications in Phase Retrieval and Linear Classification
Get PDFNonlinear models are widely used in signal processing, statistics, and machine learning to model real-world applications. A popular class of such models is the single-index model where the response variable is related to a linear combination of dependent variables through a link function. In other words, if x ∈ Rp denotes the input signal, the posterior mean of the generated output y has the form, E[y|x] = ρ(xTw), where ρ :R → R is a known function (referred to as the link function), and w ∈ Rp is the vector of unknown parameters. When ρ(•) is invertible, this class of models is called generalized linear models (GLMs). GLMs are commonly used in statistics and are often viewed as flexible generalizations of linear regression. Given n measurements (samples) from this model, D = {(xi, yi) | 1 ≤q i ≤ n}, the goal is to estimate the parameter vector w. While the model parameters are assumed to be unknown, in many applications these parameters follow certain structures (sparse, low-rank, group-sparse, etc.) The knowledge on this structure can be used to form more accurate estimators.
The main contribution of this thesis is to provide a precise performance analysis for convex optimization programs that are used for parameter estimation in two important classes of single-index models. These classes are: (1) phase retrieval in signal processing, and (2) binary classification in statistical learning.
The first class of models studied in this thesis is the phase retrieval problem, where the goal is to recover a discrete complex-valued signal from amplitudes of its linear combinations. Methods based on convex optimization have recently gained significant attentions in the literature. The conventional convex-optimization-based methods resort to the idea of lifting which makes them computationally inefficient. In addition to providing an analysis of the recovery threshold for the semidefinite-programming-based methods, this thesis studies the performance of a new convex relaxation for the phase retrieval problem, known as phasemax, which is computationally more efficient as it does not lift the signal to higher dimensions. Furthermore, to address the case of structured signals, regularized phasemax is introduced along with a precise characterization of the conditions for its perfect recovery in the asymptotic regime.
The next important application studied in this thesis is the binary classification in statistical learning. While classification models have been studied in the literature since 1950's, the understanding of their performance has been incomplete until very recently. Inspired by the maximum likelihood (ML) estimator in logistic models, we analyze a class of optimization programs that attempts to find the model parameters by minimizing an objective that consists of a loss function (which is often inspired by the ML estimator) and an additive regularization term that enforces our knowledge on the structure. There are two operating regimes for this problem depending on the separability of the training data set D. In the asymptotic regime, where the number of samples and the number of parameters grow to infinity, a phase transition phenomenon is demonstrated that happens at a certain over-parameterization ratio. We compute this phase transition for the setting where the underlying data is drawn from a Gaussian distribution.
In the case where the data is non-separable, the ML estimator is well-defined, and its attributes have been studied in the classical statistics. However, these classical results fail to provide reasonable estimate in the regime where the number of data points is proportional to the number of samples. One contribution of this thesis is to provide an exact analysis on the performance of the regularized logistic regression when the number of training data is proportional to the number of samples. When the data is separable (a.k.a. the interpolating regime), there exist multiple linear classifiers that perfectly fit the training data. In this regime, we introduce and analyze the performance of "extended margin maximizers" (EMMs). Inspired by the max-margin classifier, EMM classifiers simultaneously consider maximizing the margin and the structure of the parameter. Lastly, we discuss another generalization to the max-margin classifier, referred to as the robust max-margin classifier, that takes into account the perturbations by an adversary. It is shown that for a broad class of loss functions, gradient descent iterates (with proper step sizes) converge to the robust max-margin classifier.</p