Doctor of Philosophy, The Ohio State University, 2022, Computer Science and Engineering

ML applications are driving the next computing revolution. In this context both performance and security are crucial. We propose hardware/software co-design solutions for addressing both. First, we propose RNNFast, an accelerator for Recurrent Neural Networks (RNNs). RNNs are particularly well suited for machine learning problems in which context is important, such as language translation. RNNFast leverages an emerging class of non-volatile memory called domain-wall memory (DWM). We show that DWM is very well suited for RNN acceleration due to its very high density and low read/write energy. RNNFast is very efficient and highly scalable, with a flexible mapping of logical neurons to RNN hardware blocks. The accelerator is designed to minimize data movement by closely interleaving DWM storage and computation. We compare our design with a state-of-the-art GPGPU and find 21.8X higher performance with 70X lower energy. Second, we brought ML security into ML accelerator design for more efficiency and robustness. Deep Neural Networks (DNNs) are employed in an increasing number of applications, some of which are safety-critical. Unfortunately, DNNs are known to be vulnerable to so-called adversarial attacks. In general, the proposed defenses have high overhead, some require attack-specific re-training of the model or careful tuning to adapt to different attacks. We show that these approaches, while successful for a range of inputs, are insufficient to address stronger, high-confidence adversarial attacks. To address this, we propose HASI and DNNShield, two hardware-accelerated defenses that adapt the strength of the response to the confidence of the adversarial input. Both techniques rely on approximation or random noise deliberately introduced into the model. HASI uses direct noise injection into the model at inference. DNNShield uses approximation that relies on dynamic and random sparsification of the DNN model to achieve inference approximation efficiently and wi (open full item for complete abstract)

Committee: Radu Teoderescu (Advisor); Yang Wang (Committee Member); Wei-Lun Chao (Committee Member) Subjects: Computer Engineering; Computer Science

Doctor of Philosophy, Case Western Reserve University, 2016, Physics

The work discussed in this thesis represents the accumulation of research I performed throughout my doctoral studies. My studies were focused towards two-dimensional electronic transport in semiconductor and perovskite oxide interfaces. Electronic materials with low dimensionality provides experimentalists and theorists with incredible systems to probe physics at non-intuitive levels. Once considered “toy problems,” low-dimensional systems, particularly in two dimensions, are now treated as highly relevant, modern electronic materials on the verge of being used in next-generation technology. This thesis entails three main parts, each contributing new knowledge to the field of two-dimensional electronics and condensed matter physics in general. The first part, found in Chapter 3, analyzes short-range scattering effects in two-dimensional GaAs/AlGaAs quantum wells. The effect of aluminum concentration in the material is correlated to the non-monotonic resistance behavior at low temperatures through the short-range disorder potential. By accounting for different electronic scattering mechanisms, temperature-dependent resistance is shown to have a universal behavior, independent of short-range scattering. Chapters 4 transitions from two-dimensional electron gasses in GaAs to quasi-two-dimensional electron gasses in perovskite oxides, specifically gamma-Al2O3/SrTiO3 heterointerfaces. For the first time in that system, a metal-to-insulator transition is measured by backgating the strontium titanate. By measuring the carrier density, it is shown that immobile charge carriers are induced through backgating. Chapter 5 discusses my research on the cubic-to-tetragonal structural phase transition in LaAlO3/SrTiO3 heterointerfaces. By engineering micron-scale devices, I was able to measure the electronic transport properties of tetragonal domain walls below the structural transition temperature. Domain walls are shown to cause anisotropic resistance, which is measurable o (open full item for complete abstract)

Committee: Xuan Gao (Advisor); Harsh Mathur (Committee Member); Kathleen Kash (Committee Member); Alp Sehirlioglu (Committee Member) Subjects: Condensed Matter Physics; Physics; Solid State Physics

PHD, Kent State University, 2015, College of Arts and Sciences / Chemical Physics

Topological defects plays an important role in many physical processes ranging from morphogenesis of phase transitions in condensed matter system to the response to surface confinement and application of external fields. In this dissertation, we investigate the topological defects both in lyotropic and thermotropic nematics in order to characterize the studied materials.

Committee: Oleg Lavrentovich (Advisor); Hiroshi Yokoyama (Committee Member); Liang-Chy Chien (Committee Member); Samuel Sprunt (Committee Member); Elizabeth Mann (Committee Member) Subjects: Engineering; Experiments; Materials Science; Optics; Physical Chemistry; Physics

Doctor of Philosophy, The Ohio State University, 2006, Physics

In this thesis, the physics underlying the magnetic behavior of metallic microstructures, including their responses to magnetic fields and electric currents is explored. The dynamic and static components of the magnetization are respectively probed through Brillouin light scattering and Kerr imaging method. The design, growth and fabrication of various structures are presented, while the experimental findings are analyzed by theoretical modeling and calculations. The highlights include (a) Brillouin light scattering studies of spin precession under tunable magnetic field imbalance, (b) Kerr imaging of layer-by-layer magnetic reversal in cobalt-platinum multilayer, (c) Observation of spin-polarized current induced domain wall motion in magnetic microwires. All of these results demonstrate that light scattering as an excellent tool for probing novel functionality of metallic magnetic microstructures. Future prospects along the direction of research involved in this thesis are also presented.

Committee: Ratnasingham Sooryakumar (Advisor) Subjects: