Matching Items (24)
Filtering by
- All Subjects: engineering
- Creators: Liu, Yongming
- Member of: Theses and Dissertations
Description
Machine learning has demonstrated great potential across a wide range of applications such as computer vision, robotics, speech recognition, drug discovery, material science, and physics simulation. Despite its current success, however, there are still two major challenges for machine learning algorithms: limited robustness and generalizability.
The robustness of a neural network is defined as the stability of the network output under small input perturbations. It has been shown that neural networks are very sensitive to input perturbations, and the prediction from convolutional neural networks can be totally different for input images that are visually indistinguishable to human eyes. Based on such property, hackers can reversely engineer the input to trick machine learning systems in targeted ways. These adversarial attacks have shown to be surprisingly effective, which has raised serious concerns over safety-critical applications like autonomous driving. In the meantime, many established defense mechanisms have shown to be vulnerable under more advanced attacks proposed later, and how to improve the robustness of neural networks is still an open question.
The generalizability of neural networks refers to the ability of networks to perform well on unseen data rather than just the data that they were trained on. Neural networks often fail to carry out reliable generalizations when the testing data is of different distribution compared with the training one, which will make autonomous driving systems risky under new environment. The generalizability of neural networks can also be limited whenever there is a scarcity of training data, while it can be expensive to acquire large datasets either experimentally or numerically for engineering applications, such as material and chemical design.
In this dissertation, we are thus motivated to improve the robustness and generalizability of neural networks. Firstly, unlike traditional bottom-up classifiers, we use a pre-trained generative model to perform top-down reasoning and infer the label information. The proposed generative classifier has shown to be promising in handling input distribution shifts. Secondly, we focus on improving the network robustness and propose an extension to adversarial training by considering the transformation invariance. Proposed method improves the robustness over state-of-the-art methods by 2.5% on MNIST and 3.7% on CIFAR-10. Thirdly, we focus on designing networks that generalize well at predicting physics response. Our physics prior knowledge is used to guide the designing of the network architecture, which enables efficient learning and inference. Proposed network is able to generalize well even when it is trained with a single image pair.
The robustness of a neural network is defined as the stability of the network output under small input perturbations. It has been shown that neural networks are very sensitive to input perturbations, and the prediction from convolutional neural networks can be totally different for input images that are visually indistinguishable to human eyes. Based on such property, hackers can reversely engineer the input to trick machine learning systems in targeted ways. These adversarial attacks have shown to be surprisingly effective, which has raised serious concerns over safety-critical applications like autonomous driving. In the meantime, many established defense mechanisms have shown to be vulnerable under more advanced attacks proposed later, and how to improve the robustness of neural networks is still an open question.
The generalizability of neural networks refers to the ability of networks to perform well on unseen data rather than just the data that they were trained on. Neural networks often fail to carry out reliable generalizations when the testing data is of different distribution compared with the training one, which will make autonomous driving systems risky under new environment. The generalizability of neural networks can also be limited whenever there is a scarcity of training data, while it can be expensive to acquire large datasets either experimentally or numerically for engineering applications, such as material and chemical design.
In this dissertation, we are thus motivated to improve the robustness and generalizability of neural networks. Firstly, unlike traditional bottom-up classifiers, we use a pre-trained generative model to perform top-down reasoning and infer the label information. The proposed generative classifier has shown to be promising in handling input distribution shifts. Secondly, we focus on improving the network robustness and propose an extension to adversarial training by considering the transformation invariance. Proposed method improves the robustness over state-of-the-art methods by 2.5% on MNIST and 3.7% on CIFAR-10. Thirdly, we focus on designing networks that generalize well at predicting physics response. Our physics prior knowledge is used to guide the designing of the network architecture, which enables efficient learning and inference. Proposed network is able to generalize well even when it is trained with a single image pair.
ContributorsYao, Houpu (Author) / Ren, Yi (Thesis advisor) / Liu, Yongming (Committee member) / Li, Baoxin (Committee member) / Yang, Yezhou (Committee member) / Marvi, Hamidreza (Committee member) / Arizona State University (Publisher)
Created2019
Description
Traditionally nanoporous gold is created by selective dissolution of silver or copper from a binary silver-gold or copper-gold alloy. These alloys serve as prototypical model systems for a phenomenon referred to as stress-corrosion cracking. Stress-corrosion cracking is the brittle failure of a normally ductile material occurring in a corrosive environment under a tensile stress. Silver-gold can experience this type of brittle fracture for a range of compositions. The corrosion process in this alloy results in a bicontinuous nanoscale morphology composed of gold-rich ligaments and voids often referred to as nanoporous gold. Experiments have shown that monolithic nanoporous gold can sustain high speed cracks which can then be injected into parent-phase alloy. This work compares nanoporous gold created from ordered and disordered copper-gold using digital image analysis and electron backscatter diffraction. Nanoporous gold from both disordered copper-gold and silver-gold, and ordered copper-gold show that grain orientation and shape remain largely unchanged by the dealloying process. Comparing the morphology of the nanoporous gold from ordered and disordered copper-gold with digital image analysis, minimal differences are found between the two and it is concluded that they are not statistically significant. This reveals the robust nature of nanoporous gold morphology against small variations in surface diffusion and parent-phase crystal structure.
Then the corrosion penetration down the grain boundary is compared to the depth of crack injections in polycrystal silver-gold. Based on statistical comparison, the crack-injections penetrate into the parent-phase grain boundary beyond the corrosion-induced porosity. To compare crack injections to stress-corrosion cracking, single crystal silver-gold samples are employed. Due to the cleavage-like nature of the fracture surfaces, electron backscatter diffraction is possible and employed to compare the crystallography of stress-corrosion crack surfaces and crack-injection surfaces. From the crystallographic similarities of these fracture surfaces, it is concluded that stress-corrosion can occur via a series of crack-injection events. This relationship between crack injections and stress corrosion cracking is further examined using electrochemical data from polycrystal silver-gold samples during stress-corrosion cracking. The results support the idea that crack injection is a mechanism for stress-corrosion cracking.
Then the corrosion penetration down the grain boundary is compared to the depth of crack injections in polycrystal silver-gold. Based on statistical comparison, the crack-injections penetrate into the parent-phase grain boundary beyond the corrosion-induced porosity. To compare crack injections to stress-corrosion cracking, single crystal silver-gold samples are employed. Due to the cleavage-like nature of the fracture surfaces, electron backscatter diffraction is possible and employed to compare the crystallography of stress-corrosion crack surfaces and crack-injection surfaces. From the crystallographic similarities of these fracture surfaces, it is concluded that stress-corrosion can occur via a series of crack-injection events. This relationship between crack injections and stress corrosion cracking is further examined using electrochemical data from polycrystal silver-gold samples during stress-corrosion cracking. The results support the idea that crack injection is a mechanism for stress-corrosion cracking.
ContributorsKarasz, Erin (Author) / Sieradzki, Karl (Thesis advisor) / Chawla, Nikhilesh (Committee member) / Peralta, Pedro (Committee member) / Rajagopalan, Jagannathan (Committee member) / Arizona State University (Publisher)
Created2020
Description
Two fatigue life prediction methods using the energy-based approach have been proposed. A number of approaches have been developed in the past five decades. This study reviews some common models and discusses the model that is most suitable for each different condition, no matter whether the model is designed to solve uniaxial, multiaxial, or biaxial loading paths in fatigue prediction. In addition, different loading cases such as various loading and constant loading are also discussed. These models are suitable for one or two conditions in fatigue prediction. While most of the existing models can only solve single cases, the proposed new energy-based approach not only can deal with different loading paths but is applicable for various loading cases. The first energy-based model using the linear cumulative rule is developed to calculate random loading cases. The method is developed by combining Miner’s rule and the rainflow-counting algorithm. For the second energy-based method, I propose an alternative method and develop an approach to avert the rainflow-counting algorithm. Specifically, I propose to use an energy-based model by directly using the time integration concept. In this study, first, the equivalent energy concept that can transform three-dimensional loading into an equivalent loading will be discussed. Second, the new damage propagation method modified by fatigue crack growth will be introduced to deal with cycle-based fatigue prediction. Third, the time-based concept will be implemented to determine fatigue damage under every cycle in the random loading case. The formulation will also be explained in detail. Through this new model, the fatigue life can be calculated properly in different loading cases. In addition, the proposed model is verified with experimental datasets from several published studies. The data include both uniaxial and multiaxial loading paths under constant loading and random loading cases. Finally, the discussion and conclusion based on the results, are included. Additional loading cases such as the spectrum including both elastic and plastic regions will be explored in future research.
ContributorsTien, Shih-Chuan (Author) / Liu, Yongming (Thesis advisor) / Nian, Qiong (Committee member) / Jiao, Yang (Committee member) / Arizona State University (Publisher)
Created2021
Description
Collision-free path planning is also a major challenge in managing unmanned aerial vehicles (UAVs) fleets, especially in uncertain environments. The design of UAV routing policies using multi-agent reinforcement learning has been considered, and propose a Multi-resolution, Multi-agent, Mean-field reinforcement learning algorithm, named 3M-RL, for flight planning, where multiple vehicles need to avoid collisions with each other while moving towards their destinations. In this system, each UAV makes decisions based on local observations, and does not communicate with other UAVs. The algorithm trains a routing policy using an Actor-Critic neural network with multi-resolution observations, including detailed local information and aggregated global information based on mean-field. The algorithm tackles the curse-of-dimensionality problem in multi-agent reinforcement learning and provides a scalable solution. The proposed algorithm is tested in different complex scenarios in both 2D and 3D space and the simulation results show that 3M-RL result in good routing policies. Also as a compliment, dynamic data communications between UAVs and a control center has also been studied, where the control center needs to monitor the safety state of each UAV in the system in real time, where the transition of risk level is simply considered as a Markov process. Given limited communication bandwidth, it is impossible for the control center to communicate with all UAVs at the same time. A dynamic learning problem with limited communication bandwidth is also discussed in this paper where the objective is to minimize the total information entropy in real-time risk level tracking. The simulations also demonstrate that the algorithm outperforms policies such as a Round & Robin policy.
ContributorsWang, Weichang (Author) / Ying, Lei (Thesis advisor) / Liu, Yongming (Thesis advisor) / Zhang, Junshan (Committee member) / Zhang, Yanchao (Committee member) / Arizona State University (Publisher)
Created2021