Matching Items (28)
Filtering by
- All Subjects: artificial intelligence
- All Subjects: Atmospheric science
- Genre: Academic theses
- Creators: Huang, Huei-Ping
- Creators: Seo, Jae-Sun
- Creators: Turaga, Pavan
- Status: Published
Description
The numerical climate models have provided scientists, policy makers and the general public, crucial information for climate projections since mid-20th century. An international effort to compare and validate the simulations of all major climate models is organized by the Coupled Model Intercomparison Project (CMIP), which has gone through several phases since 1995 with CMIP5 being the state of the art. In parallel, an organized effort to consolidate all observational data in the past century culminates in the creation of several "reanalysis" datasets that are considered the closest representation of the true observation. This study compared the climate variability and trend in the climate model simulations and observations on the timescales ranging from interannual to centennial. The analysis focused on the dynamic climate quantity of zonal-mean zonal wind and global atmospheric angular momentum (AAM), and incorporated multiple datasets from reanalysis and the most recent CMIP3 and CMIP5 archives. For the observation, the validation of AAM by the length-of-day (LOD) and the intercomparison of AAM revealed a good agreement among reanalyses on the interannual and the decadal-to-interdecadal timescales, respectively. But the most significant discrepancies among them are in the long-term mean and long-term trend. For the simulations, the CMIP5 models produced a significantly smaller bias and a narrower ensemble spread of the climatology and trend in the 20th century for AAM compared to CMIP3, while CMIP3 and CMIP5 simulations consistently produced a positive trend for the 20th and 21st century. Both CMIP3 and CMIP5 models produced a wide range of the magnitudes of decadal and interdecadal variability of wind component of AAM (MR) compared to observation. The ensemble means of CMIP3 and CMIP5 are not statistically distinguishable for either the 20th- or 21st-century runs. The in-house atmospheric general circulation model (AGCM) simulations forced by the sea surface temperature (SST) taken from the CMIP5 simulations as lower boundary conditions were carried out. The zonal wind and MR in the CMIP5 simulations are well simulated in the AGCM simulations. This confirmed SST as an important mediator in regulating the global atmospheric changes due to GHG effect.
ContributorsPaek, Houk (Author) / Huang, Huei-Ping (Thesis advisor) / Adrian, Ronald (Committee member) / Wang, Zhihua (Committee member) / Anderson, James (Committee member) / Herrmann, Marcus (Committee member) / Arizona State University (Publisher)
Created2013
Description
The partitioning of available solar energy into different fluxes at the Earth's surface is important in determining different physical processes, such as turbulent transport, subsurface hydrology, land-atmospheric interactions, etc. Direct measurements of these turbulent fluxes were carried out using eddy-covariance (EC) towers. However, the distribution of EC towers is sparse due to relatively high cost and practical difficulties in logistics and deployment. As a result, data is temporally and spatially limited and is inadequate to be used for researches at large scales, such as regional and global climate modeling. Besides field measurements, an alternative way is to estimate turbulent fluxes based on the intrinsic relations between surface energy budget components, largely through thermodynamic equilibrium. These relations, referred as relative efficiency, have been included in several models to estimate the magnitude of turbulent fluxes in surface energy budgets such as latent heat and sensible heat. In this study, three theoretical models based on the lumped heat transfer model, the linear stability analysis and the maximum entropy principle respectively, were investigated. Model predictions of relative efficiencies were compared with turbulent flux data over different land covers, viz. lake, grassland and suburban surfaces. Similar results were observed over lake and suburban surface but significant deviation is found over vegetation surface. The relative efficiency of outgoing longwave radiation is found to be orders of magnitude deviated from theoretic predictions. Meanwhile, results show that energy partitioning process is influenced by the surface water availability to a great extent. The study provides insight into what property is determining energy partitioning process over different land covers and gives suggestion for future models.
ContributorsYang, Jiachuan (Author) / Wang, Zhihua (Thesis advisor) / Huang, Huei-Ping (Committee member) / Vivoni, Enrique (Committee member) / Mays, Larry (Committee member) / Arizona State University (Publisher)
Created2012
Description
This thesis outlines the development of a vector retrieval technique, based on data assimilation, for a coherent Doppler LIDAR (Light Detection and Ranging). A detailed analysis of the Optimal Interpolation (OI) technique for vector retrieval is presented. Through several modifications to the OI technique, it is shown that the modified technique results in significant improvement in velocity retrieval accuracy. These modifications include changes to innovation covariance portioning, covariance binning, and analysis increment calculation. It is observed that the modified technique is able to make retrievals with better accuracy, preserves local information better, and compares well with tower measurements. In order to study the error of representativeness and vector retrieval error, a lidar simulator was constructed. Using the lidar simulator a thorough sensitivity analysis of the lidar measurement process and vector retrieval is carried out. The error of representativeness as a function of scales of motion and sensitivity of vector retrieval to look angle is quantified. Using the modified OI technique, study of nocturnal flow in Owens' Valley, CA was carried out to identify and understand uncharacteristic events on the night of March 27th 2006. Observations from 1030 UTC to 1230 UTC (0230 hr local time to 0430 hr local time) on March 27 2006 are presented. Lidar observations show complex and uncharacteristic flows such as sudden bursts of westerly cross-valley wind mixing with the dominant up-valley wind. Model results from Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS®) and other in-situ instrumentations are used to corroborate and complement these observations. The modified OI technique is used to identify uncharacteristic and extreme flow events at a wind development site. Estimates of turbulence and shear from this technique are compared to tower measurements. A formulation for equivalent wind speed in the presence of variations in wind speed and direction, combined with shear is developed and used to determine wind energy content in presence of turbulence.
ContributorsChoukulkar, Aditya (Author) / Calhoun, Ronald (Thesis advisor) / Mahalov, Alex (Committee member) / Kostelich, Eric (Committee member) / Huang, Huei-Ping (Committee member) / Phelan, Patrick (Committee member) / Arizona State University (Publisher)
Created2013
Description
The implications of a changing climate have a profound impact on human life, society, and policy making. The need for accurate climate prediction becomes increasingly important as we better understand these implications. Currently, the most widely used climate prediction relies on the synthesis of climate model simulations organized by the Coupled Model Intercomparison Project (CMIP); these simulations are ensemble-averaged to construct projections for the 21st century climate. However, a significant degree of bias and variability in the model simulations for the 20th century climate is well-known at both global and regional scales. Based on that insight, this study provides an alternative approach for constructing climate projections that incorporates knowledge of model bias. This approach is demonstrated to be a viable alternative which can be easily implemented by water resource managers for potentially more accurate projections. Tests of the new approach are provided on a global scale with an emphasis on semiarid regional studies for their particular vulnerability to water resource changes, using both the former CMIP Phase 3 (CMIP3) and current Phase 5 (CMIP5) model archives. This investigation is accompanied by a detailed analysis of the dynamical processes and water budget to understand the behaviors and sources of model biases. Sensitivity studies of selected CMIP5 models are also performed with an atmospheric component model by testing the relationship between climate change forcings and model simulated response. The information derived from each study is used to determine the progressive quality of coupled climate models in simulating the global water cycle by rigorously investigating sources of model bias related to the moisture budget. As such, the conclusions of this project are highly relevant to model development and potentially may be used to further improve climate projections.
ContributorsBaker, Noel C (Author) / Huang, Huei-Ping (Thesis advisor) / Trimble, Steve (Committee member) / Anderson, James (Committee member) / Clarke, Amanda (Committee member) / Calhoun, Ronald (Committee member) / Arizona State University (Publisher)
Created2013
Description
This study considered the impact of grid resolution on wind velocity simulated by the Weather Research and Forecasting (WRF) model. The period simulated spanned November 2009 through January 2010, for which, multi-resolution nested domains were examined. Basic analysis was performed utilizing the data assimilation tools of NCEP/NCAR (National Center for Environmental Prediction/National Center for Atmospheric Research) to determine the ideal location to examine during the simulation was the Pacific Northwest portion of the United States, specifically the border between California and Oregon. The simulated mutli-resolution nested domains in this region indicated an increase in apparent wind speed as the resolution for the domain was increased. These findings were confirmed by statistical analysis which identified a positive bias for wind speed with respect to increased resolution as well as a correlation coefficient indicating the existence of a positive change in wind speed with increased resolution. An analysis of temperature change was performed in order to test the validity of the findings of the WRF simulation model. The statistical analysis performed on temperature change throughout the increased grid resolution did not indicate any change in temperature. In fact the correlation coefficient values between the domains were found in the 0.90 range, indicating the non-sensitivity of temperature across the increased resolutions. These results validate the findings of the WRF simulation: increased wind velocity can be observed at higher grid resolution. The study then considered the difference between wind velocity observed over the entire domains and the wind velocity observed solely over offshore locations. Wind velocity was observed to be significantly higher (an increase of 68.4%) in the offshore locations. The findings of this study suggest simulation tools should be utilized to examine domains at a higher resolution in order to identify potential locations for wind farms. The results go further to suggest the ideal location for these potential wind farms will be at offshore locations.
ContributorsBouey, Michael (Author) / Huang, Huei-Ping (Thesis advisor) / Trimble, Steve (Committee member) / Ronald, Ronald (Committee member) / Arizona State University (Publisher)
Created2012
Description
Adversarial threats of deep learning are increasingly becoming a concern due to the ubiquitous deployment of deep neural networks(DNNs) in many security-sensitive domains. Among the existing threats, adversarial weight perturbation is an emerging class of threats that attempts to perturb the weight parameters of DNNs to breach security and privacy.In this thesis, the first weight perturbation attack introduced is called Bit-Flip Attack (BFA), which can maliciously flip a small number of bits within a computer’s main memory system storing the DNN weight parameter to achieve malicious objectives. Our developed algorithm can achieve three specific attack objectives: I) Un-targeted accuracy degradation attack, ii) Targeted attack, & iii) Trojan attack. Moreover, BFA utilizes the rowhammer technique to demonstrate the bit-flip attack in an actual computer prototype. While the bit-flip attack is conducted in a white-box setting, the subsequent contribution of this thesis is to develop another novel weight perturbation attack in a black-box setting. Consequently, this thesis discusses a new study of DNN model vulnerabilities in a multi-tenant Field Programmable Gate Array (FPGA) cloud under a strict black-box framework. This newly developed attack framework injects faults in the malicious tenant by duplicating specific DNN weight packages during data transmission between off-chip memory and on-chip buffer of a victim FPGA. The proposed attack is also experimentally validated in a multi-tenant cloud FPGA prototype. In the final part, the focus shifts toward deep learning model privacy, popularly known as model extraction, that can steal partial DNN weight parameters remotely with the aid of a memory side-channel attack. In addition, a novel training algorithm is designed to utilize the partially leaked DNN weight bit information, making the model extraction attack more effective. The algorithm effectively leverages the partial leaked bit information and generates a substitute prototype of the victim model with almost identical performance to the victim.
ContributorsRakin, Adnan Siraj (Author) / Fan, Deliang (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Seo, Jae-Sun (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2022
Description
Generative models are deep neural network-based models trained to learn the underlying distribution of a dataset. Once trained, these models can be used to sample novel data points from this distribution. Their impressive capabilities have been manifested in various generative tasks, encompassing areas like image-to-image translation, style transfer, image editing, and more. One notable application of generative models is data augmentation, aimed at expanding and diversifying the training dataset to augment the performance of deep learning models for a downstream task. Generative models can be used to create new samples similar to the original data but with different variations and properties that are difficult to capture with traditional data augmentation techniques. However, the quality, diversity, and controllability of the shape and structure of the generated samples from these models are often directly proportional to the size and diversity of the training dataset. A more extensive and diverse training dataset allows the generative model to capture overall structures present in the data and generate more diverse and realistic-looking samples.
In this dissertation, I present innovative methods designed to enhance the robustness and controllability of generative models, drawing upon physics-based, probabilistic, and geometric techniques. These methods help improve the generalization and controllability of the generative model without necessarily relying on large training datasets. I enhance the robustness of generative models by integrating classical geometric moments for shape awareness and minimizing trainable parameters. Additionally, I employ non-parametric priors for the generative model's latent space through basic probability and optimization methods to improve the fidelity of interpolated images. I adopt a hybrid approach to address domain-specific challenges with limited data and controllability, combining physics-based rendering with generative models for more realistic results.
These approaches are particularly relevant in industrial settings, where the training datasets are small and class imbalance is common. Through extensive experiments on various datasets, I demonstrate the effectiveness of the proposed methods over conventional approaches.
ContributorsSingh, Rajhans (Author) / Turaga, Pavan (Thesis advisor) / Jayasuriya, Suren (Committee member) / Berisha, Visar (Committee member) / Fazli, Pooyan (Committee member) / Arizona State University (Publisher)
Created2023
Description
The rapid growth of Internet-of-things (IoT) and artificial intelligence applications have called forth a new computing paradigm--edge computing. Edge computing applications, such as video surveillance, autonomous driving, and augmented reality, are highly computationally intensive and require real-time processing. Current edge systems are typically based on commodity general-purpose hardware such as Central Processing Units (CPUs) and Graphical Processing Units (GPUs) , which are mainly designed for large, non-time-sensitive jobs in the cloud and do not match the needs of the edge workloads. Also, these systems are usually power hungry and are not suitable for resource-constrained edge deployments. Such application-hardware mismatch calls forth a new computing backbone to support the high-bandwidth, low-latency, and energy-efficient requirements. Also, the new system should be able to support a variety of edge applications with different characteristics. This thesis addresses the above challenges by studying the use of Field Programmable Gate Array (FPGA) -based computing systems for accelerating the edge workloads, from three critical angles. First, it investigates the feasibility of FPGAs for edge computing, in comparison to conventional CPUs and GPUs. Second, it studies the acceleration of common algorithmic characteristics, identified as loop patterns, using FPGAs, and develops a benchmark tool for analyzing the performance of these patterns on different accelerators. Third, it designs a new edge computing platform using multiple clustered FPGAs to provide high-bandwidth and low-latency acceleration of convolutional neural networks (CNNs) widely used in edge applications. Finally, it studies the acceleration of the emerging neural networks, randomly-wired neural networks, on the multi-FPGA platform. The experimental results from this work show that the new generation of workloads requires rethinking the current edge-computing architecture. First, through the acceleration of common loops, it demonstrates that FPGAs can outperform GPUs in specific loops types up to 14 times. Second, it shows the linear scalability of multi-FPGA platforms in accelerating neural networks. Third, it demonstrates the superiority of the new scheduler to optimally place randomly-wired neural networks on multi-FPGA platforms with 81.1 times better throughput than the available scheduling mechanisms.
ContributorsBiookaghazadeh, Saman (Author) / Zhao, Ming (Thesis advisor) / Ren, Fengbo (Thesis advisor) / Li, Baoxin (Committee member) / Seo, Jae-Sun (Committee member) / Arizona State University (Publisher)
Created2021
Description
This dissertation presents novel solutions for improving the generalization capabilities of deep learning based computer vision models. Neural networks are known to suffer a large drop in performance when tested on samples from a different distribution than the one on which they were trained. The proposed solutions, based on latent space geometry and meta-learning, address this issue by improving the robustness of these models to distribution shifts. Through the use of geometrical alignment, state-of-the-art domain adaptation and source-free test-time adaptation strategies are developed. Additionally, geometrical alignment can allow classifiers to be progressively adapted to new, unseen test domains without requiring retraining of the feature extractors. The dissertation also presents algorithms for enabling in-the-wild generalization without needing access to any samples from the target domain. Other causes of poor generalization, such as data scarcity in critical applications and training data with high levels of noise and variance, are also explored. To address data scarcity in fine-grained computer vision tasks such as object detection, novel context-aware augmentations are suggested. While the first four chapters focus on general-purpose computer vision models, strategies are also developed to improve robustness in specific applications. The efficiency of training autonomous agents for visual navigation is improved by incorporating semantic knowledge, and the integration of domain experts' knowledge allows for the realization of a low-cost, minimally invasive generalizable automated rehabilitation system. Lastly, new tools for explainability and model introspection using counter-factual explainers trained through interval-based uncertainty calibration objectives are presented.
ContributorsThopalli, Kowshik (Author) / Turaga, Pavan (Thesis advisor) / Thiagarajan, Jayaraman J (Committee member) / Li, Baoxin (Committee member) / Yang, Yezhou (Committee member) / Arizona State University (Publisher)
Created2023
Description
Millimeter-wave (mmWave) and sub-terahertz (sub-THz) systems aim to utilize the large bandwidth available at these frequencies. This has the potential to enable several future applications that require high data rates, such as autonomous vehicles and digital twins. These systems, however, have several challenges that need to be addressed to realize their gains in practice. First, they need to deploy large antenna arrays and use narrow beams to guarantee sufficient receive power. Adjusting the narrow beams of the large antenna arrays incurs massive beam training overhead. Second, the sensitivity to blockages is a key challenge for mmWave and THz networks. Since these networks mainly rely on line-of-sight (LOS) links, sudden link blockages highly threaten the reliability of the networks. Further, when the LOS link is blocked, the network typically needs to hand off the user to another LOS basestation, which may incur critical time latency, especially if a search over a large codebook of narrow beams is needed. A promising way to tackle both these challenges lies in leveraging additional side information such as visual, LiDAR, radar, and position data. These sensors provide rich information about the wireless environment, which can be utilized for fast beam and blockage prediction. This dissertation presents a machine-learning framework for sensing-aided beam and blockage prediction. In particular, for beam prediction, this work proposes to utilize visual and positional data to predict the optimal beam indices. For the first time, this work investigates the sensing-aided beam prediction task in a real-world vehicle-to-infrastructure and drone communication scenario. Similarly, for blockage prediction, this dissertation proposes a multi-modal wireless communication solution that utilizes bimodal machine learning to perform proactive blockage prediction and user hand-off. Evaluations on both real-world and synthetic datasets illustrate the promising performance of the proposed solutions and highlight their potential for next-generation communication and sensing systems.
ContributorsCharan, Gouranga (Author) / Alkhateeb, Ahmed (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Turaga, Pavan (Committee member) / Michelusi, Nicolò (Committee member) / Arizona State University (Publisher)
Created2024