93453 research outputs found
Sort by
Exploring Factors Influencing Large-Scale Atmospheric Circulation and Precipitation over North America
No full textThe observed large-scale atmospheric circulation and precipitation pattern over North America consists of substantial zonal (east-west) asymmetries. Despite such a pattern's relevance to regional climate, a thorough explanation behind how such features arise remains somewhat lacking in existing literature. Here we further investigate the dynamical causes of zonal asymmetries in atmospheric circulation and precipitation over North America by performing and exploring a series of global climate model experiments with altered geography and land surface characteristics over North America and analyzing observations.
In Chapter 2, we explore the role of the Gulf of California in the North American monsoon by comparing a realistic climate simulation with a water-filled Gulf of California to one where the Gulf of California is instead replaced with flat land. We find that the Gulf of California enhances North American monsoon precipitation, and the results provide numerous additional insights on the nature of the North American monsoon.
In Chapter 3, we examine how North America's topography shapes the atmospheric circulation and precipitation pattern over North America and beyond during the cold season (December-March). Here, we perform and analyze a series of climate model simulations with altered topography over North America and show how the simulated impact of topography is consistent with observations. We find that North America's topography mechanically diverts the large-scale midlatitude westerly flow and subtropical low-level easterly flow meridionally, contributing substantially to the observed high-amplitude atmospheric wave pattern and zonal asymmetries in mean precipitation over North America.
In Chapter 4, we return to the warm season and investigate how topography and land surface heating shape the summer atmospheric circulation and precipitation pattern over North America. We use the topography experiments performed in Chapter 3 and perform additional land albedo perturbation experiments. We find that both topography and land surface heating are essential. Whereas the northern Rockies primarily mechanically divert the large-scale midlatitude westerly flow so as to contribute to the observed ridge-trough pattern over northern North America, we explain how the circulation over the southern Rockies is largely driven by strong surface sensible heating over the region's sloping terrain. As such, in contrast to some recent work, we argue that well-known features such as the Great Plains low-level jet and North American monsoon can be largely characterized as thermally-driven
Mechanisms of Cognitive Aging in Caenorhabditis elegans: Normal aging, insulin signaling, and sexual dimorphism
No full textCognitive aging has become a serious public health concern in our aging society. Molecular, cellular, and connectivity changes occur with age in the brain, leading to behavioral changes, including the decline in learning and memory abilities. Here, we used the model organism Caenorhabditis elegans to study the molecular, morphological, and behavioral changes that occur during neuronal aging. We first characterized the transcriptomic changes that occur in neurons during aging, and then we characterized how the daf-2 Insulin/IGF-1 receptor mutant exhibits a significant DAF-16-dependent extension of learning and memory span with age compared to wild-type worms. Using RNA sequencing, we discovered that aged IIS/FOXO pathway mutants exhibit distinct neuronal transcriptomic alterations in response to cognitive aging, including the upregulation of stress response genes whose homologs have been found to be neuroprotective. Second, in addition to assessing hermaphrodite aging, we also characterized the male neuronal aging process. Besides sex-shared neuronal aging genes, males differentially downregulate mitochondrial metabolic genes and upregulate GPCR genes with age, while the X chromosome exhibits increased gene expression in hermaphrodites and altered dosage compensation complex expression with age. Third, we identified neuron-type-specific targets of the daf-2 mutant that were overshadowed in bulk neuronal sequencing using single-nucleus sequencing. We found that chemosensory neurons exhibit the biggest changes between daf-2 mutants and wild-type neurons. We validated the neuron-type-specific expression changes in daf-2 using promoter-GFP constructs and identified AWC-specific gene expression changes in daf-2 neurons that control the learning and memory improvements in daf-2 worms through behavioral validation. Combining deep single-neuron transcriptomics, genetic manipulation, and behavioral analyses, this thesis enabled us to identify conserved genes that function in specific adult neurons to control behaviors such as learning and memory, elucidating possible pathways for further study, and shedding light on intervention development
Interpreting cancer genomes: Computational methods for uncovering the genomic basis of transcriptional dysregulation in cancer
No full textThough somatic mutations play a critical role in driving cancer initiation and progression, the functional impacts of these mutations—particularly, how they alter expression patterns across the genome and give rise to cancer hallmarks—are not yet well-understood, even for mutations in well-studied cancer driver genes like KRAS or PIK3CA. Given the rise in cancer therapies designed to target commonly mutated driver genes, such as mutant KRAS inhibitors, the need for a nuanced understanding of the molecular effects of these mutations has become urgent. This thesis describes a suite of computational methods that integrate multiomic data from patient tumors—including somatic mutation, copy number alteration, methylation, and germline variation—to discern the effects of mutated cancer driver genes on expression across the genome. This begins with a linear regression-based framework, Dyscovr, which we apply both pan-cancer and individually across 19 cancer types in the Cancer Genome Atlas (TCGA), obtaining thousands of broad and cancer type-specific links. To hone in on driver-target pairs with the most potential clinical relevance, we developed a pipeline to predict which of Dyscovr's significant pairings are most likely to exhibit negative genetic interactions, such as synthetic lethality (SL). In collaboration with the Rabinowitz lab at Princeton, we experimentally validated some of these putative SL pairs in cell lines, identifying novel pairings with clinical potential. The output of Dyscovr is available in a user-friendly website, dyscovr.princeton.edu, which we anticipate will prove a valuable tool to precipitate further experimental and clinical innovations in cancer. Finally, we put forth a complementary framework, DyscovrSNP, which systematically uncovers the role that germline variants play in modulating the transcriptional effects of mutant driver genes. DyscovrSNP identifies hundreds of novel somatic-germline interactions pan-cancer, many of which also relate to patient outcomes such as survival and drug response. Altogether, this work introduces integrative computational approaches to uncover the transcriptional impacts of somatically mutated cancer driver genes, germline SNPs, and their interactions, a critical step forward in developing personalized cancer therapies
Hardware-aware Training for In-memory Computing Systems
No full textIn-memory Computing (IMC) is an emerging approach to address compute and data-movement costs inherent in high-dimensional matrix-vector multiplies (MVM), which are crucial operations in modern deep learning models. However, IMC suffers from various noise sources, which can be classified as 1) analog noise, predominant in low-SNR IMC systems; and 2) quantization noise which dominants in high-SNR IMC systems. These unavoidable noise sources in practical hardware lead to a degradation in deep learning inference performance. This dissertation delves into approaches to enable high performance deep learning for IMC systems affected by various types of noise. The research includes both algorithmic techniques and co-design strategies with hardware modeling.
The initial attempt to tackle analog noise involves a co-design with algorithmic techniques, proposing an approach named Stochastic Data-Driven Hardware Resilience (S-DDHR). This approach integrates the stochastic nature of hardware into the training process, achieved by formulating a statistical model capturing the hardware variations. The focus lies particularly on process variations, as they constitute the primary source of analog noise in advanced silicon technologies. An MRAM-based IMC architecture is introduced to evaluate the S-DDHR method, where the variation parameters are extracted and modeled based on foundry data. The evaluation spans various bit-precisions and datasets.
In order to fully recover the performance degradation due to analog noise on practical energy/throughput-aggressive IMC systems, particularly on IMC with emerging memory technologies that exhibit a low SNR at the compute output, an enhanced statistical framework is developed. It consists of a contrastive and progressive training algorithm aimed to enhance the model robustness against hardware noise, along with a macro-level modeling approach of analog circuit noise. The noise parameters for this modeling can be derived from a limited number of hardware measurements and calibrations. The effectiveness of this framework is tested on practical MRAM-based IMC prototype chips in 22nm FD-SOI, successfully demonstrating on-chip deep learning inference across multiple tasks and bit-precisions.
Finally, the challenge of quantization noise for high-SNR IMC systems is addressed by exploring purely algorithmic methods. The primary challenge of quantization in IMC systems lies in the additional ADC quantization introduced to each compute output. This is particularly critical in SRAM-based IMC, which is becoming promising due to its robustness and scalability. An approach named Reshape and Adapt for Output Quantization (RAOQ), is proposed, including a reshaping technique for neural network weights and a shift approach for activations, to reform their statistics and improve the degraded SQNR due to ADC quantization; and a bit augmentation method to aid the optimization process of model parameters, as well as an ADC-LoRA technique to reduce the training overhead. RAOQ is evaluated across large-scale models for a wide range of AI tasks and bit-precisions
TOPOLOGICAL PHASES OF MATTER IN CRYSTALS AND SUPERCONDUCTING QUANTUM CIRCUITS
No full textTopological states and topologically ordered phases are a cornerstone of physics, with applications ranging from quantum materials to quantum error correction. The robust quantum properties of topological materials may be harnessed for technology. On the other hand, topological order is a powerful tool for the storage and manipulation of quantum information with high fidelity. In this dissertation, I describe the experimental observation of many topological phases of matter using photoemission spectroscopy and quantum simulation. To start, I focus on magnets with a kagome crystal structure, where quantum interference generically leads to Dirac crossings and flat bands. In quasi-two dimensional TbMnSn, a Chern gapped state is evidenced by photoemission band structure maps, a Landau fan in tunneling spectroscopy data, and in-gap edge states. Then in the three-dimensional kagome ferromagnet CoSnS, the annihilation of Weyl points is demonstrated through careful photoemission measurements with varying temperature.
Next, I explore an extension of Weyl semimetals: higher-fold chiral semimetals. These materials host fermionic excitations that cannot be described by the standard model, where bands with large Chern number give rise to long helicoid Fermi arcs. In the topological chiral crystal NiRhSi, I fully characterize a higher-fold fermion by imaging all relevant bulk bands and extracting the Chern number of each band gap through the bulk-boundary correspondence. Then in stoichiometric RhSi and CoSi, I demonstrate a generic behavior of Fermi arcs to generate van Hove singularities. These van Hove points may be important for generating correlated states, such as the charge order recently observed in CoSi. The large topological nontrivial energy window in these compounds is also advantageous to search for quantized optical response.
Finally, I present results from a Google superconducting quantum processor demonstrating the quantum simulation of lattice gauge theory, which is equivalent to the toric code in the zero-field limit. The quantum dynamics in topological and trivial phases show deconfined and confined behavior, respectively. In addition, because the simulation is in two bona fide spatial dimensions, our protocol can be leveraged to visualized the dynamics of a Wegner-Wilson string. String breaking is also observed.
With the field of topological quantum materials headed toward ever more correlated platforms, the need for precise many-body quantum simulation techniques is paramount. At the same time, breakthroughs in quantum materials may put forward superior platforms for quantum computing. This dissertation promotes the budding synergetic relationship between quantum matter and quantum computers
