Abstract: In the field of mathematics, the systole of hyperbolic surfaces, represented by the shortest simple closed geodesic, has been a subject of investigation. This research focuses on the genus 2 surfaces, where the presence of 12 systole relise by the Bolza surface has been established. However, for surfaces of genus greater than or equal to 3, the question remains unanswered. To address this inquiry, our study aims to derive a general solution by initially exploring the systole within a specialized subset of the Moduli space. By narrowing our focus, we endeavor to contribute insights that could potentially lead to a broader understanding of systole behavior across hyperbolic surfaces.

Abstract: The Abstract-based verification, as the name suggests, is based on the approach of abstract interpretation, a well-studied framework for defining program analyses. We are interested in applying abstract interpretation to over-approximate the semantics of a neural network, and therefore produce proofs of correctness. Suppose there is an image and we pick another image that is "like" the original image, but is "slightly" brighter or darker by at most 0.1 per pixel, assuming each pixel is some real number encoding its brightness. Now, execute the network on the distorted image. The correctness property says that the network must predict that the distorted image should belong to the same class as that of the original image. The issue in checking this is that there are infinitely many possible images and we cannot run all those images through the network and ensure that each and every one of them is assigned the same class. The trick is to define infinite sets of inputs using data structures that we can manipulate, called Abstract domains.

Abstract: AI on the edge has emerged as an important research area in the last decade to deploy different applications in the domains of computer vision and natural language processing on tiny devices. These devices have limited on-chip memory and are battery-powered. On the other hand, neural network models require large memory to store model parameters and intermediate activation values. Thus, it is critical to make the models smaller so that their on-chip memory requirements are reduced. Various existing techniques like quantization and weight sharing reduce model sizes at the expense of some loss in accuracy. We propose a lossless technique of model size reduction by focusing on sharing of exponents in weights, which is different from sharing of weights. We present results based on hardware implementations of Generalized Matrix Multiplication (GEMM) in neural network models. Our method achieves at least a 20% reduction in memory when using Bfloat16 and around 10% reduction when using IEEE single-precision floating-point, for models in general, with a very small impact (up to 10% on the processor and less than 1% on FPGA) on the execution time with no loss in accuracy. We extend this work by weight approximation on few state of the art convolutional neural networks. To bridge the loss in accuracy caused by this approximation we introduce a novel method of exponent share aware training. This presentation will cover the ideas, implementations, experimental set ups and findings for above all mentioned approaches.

Abstract: Modern multiprocessor systems adopt optimization techniques to boost the speed of execution. These optimizations create vulnerabilities that can be exploited by attackers, thus causing security breaches. The hierarchical structure of cache memory where the Last Level Cache is a super set of previous levels and is shared between multiple cores of the processors creates an attack vector for cache side-channel attacks (SCA). In such attacks, the attacker is able to trace the pattern of victim process execution and correspondingly retrieve secret information by monitoring the shared cache. Mitigation techniques against such attacks trade off security against overall system performance . Hence, mitigation only when an attack is detected is needed. We propose an architecture-agnostic approach that uses hardware performance counters at run time and at thread level instead of current state of the art which use counters at system level to detect cache SCA. The proposed approach reduces the false positives by 48% when compared with system level approaches. Thus, the trade off with performance is also reduced and hence, the proposed approach is especially significant for embedded systems where processor cycle time is a limited resource.

Abstract: In this presentation, I will talk about different deep learning based models, probabilistic approaches and a mixture of two approaches for identifying the relevant features. In the healthcare domain, the gene expression dataset generally consists of high dimensional genes and low sample size. Hence, feature selection is an important step but a challenging step due to the need of human involvement with a domain expertise. I will conclude the talk with the preliminary results on different models.

Abstract: Several challenges are encountered while integrating microgrids (MGs) with an existing utility grid. These are low inertia, intermittent nature of renewable energy resources (RESs), sensor and actuator faults, unbalanced and nonlinear loads, supply-demand mismatch, and uncertain switching functions of the power electronic converter. MG control is fully dependent on the communication network, which is also susceptible to different types of failures such as noise in communication links, communication delay, limited bandwidth, packet dropout, and various malicious cyber attacks. Very few papers are available in the literature that focus on the review of control schemes for MG. Consequently, we present a comprehensive review of different robust and adaptive control schemes that address the challenges encountered due to communication constraints, uncertainties, and disturbances for different MG topologies such as AC, DC, and hybrid AC/DC MGs, as well as the current research trends in the field of robust and adaptive control. It can be concluded that to achieve control objectives of MG and overcome the existing challenges, robust and adaptive controllers show significantly improved performance in terms of transient and steady-state behavior and robustness as compared to traditional controllers.

Abstract: The main challenge in using the Modular Multilevel Converter-based constant-torque variable-speed motor drives is increased sub-module capacitor voltage ripples (SM-CVR) at low fundamental frequency operation due to the inverse relationship between SM-CVR and operating frequency. A variable slope trapezoidal circulating current (CC) is injected with square wave common-mode voltage (CMV) to address the increase in the SMCVR issue. Compared to sinusoidal CC and sinusoidal CMV injection, the proposed injection technique can reduce the peak of the CC in the range of 0% to 50%, resulting in lesser device stress and improved efficiency. Simulation results of the proposed technique are presented, and they are further compared with the existing injection techniques to show the superiority.

Abstract: A high-Curie temperature based hybrid ferrite Al-Ni-Co (FAF) permanent magnet assisted synchronous reluctance motor (PMa-SyRM) for an air-conditioner compressor is proposed in this paper. In these applications, the temperature around the motor shaft is high, typically around 120 to 150 degree Celsius. Hence, hybrid ferrite and Al-Ni-Co PMs are used. However, Al-Ni-Co PMs have steep-II quadrant B-H curve and hence the demagnetization issues need to be addressed. In the proposed hybrid PMa-SyRM the Al-Ni-Co PMs are suitably placed in the center flux barrier to address the demagnetization issues. An analytical magnetic equivalent circuit (MEC) approach is used to evaluate the d- and q- axes inductances by considering the reluctance elements, including leakage and non-linear reluctances. Further, 2-D finite element (FE) analysis of the proposed PMa-SyRM, demagnetization analysis of permanent magnets (PM) and stress analysis of the rotor are presented. Finally, a prototype is fabricated, experimental validation is carried-out and the results are compared.

Abstract: Electric vehicles (EVs) integrated with a microgrid (MG) and renewable energy sources (RES) like PV systems enable the envisioned future of a pollution-free environment. An efficient design of EV charging infrastructure is essential to the automotive industry. These charging infrastructures connected to MG are expected to provide flexible and uninterrupted service. However, designing an energy management strategy (EMS) for a charging station connected to a MG is a challenging task due to the intermittency of PV power generation and power transfer scenarios. A smart bidirectional charging station reduces transmission losses with better power flow control. However, the uncoordinated charging of EV-MG results in ineffective utilisation of the RES connected to the charging station. There is a need to develop charging stations that include multiport charging facilities, which prohibits overloading of the grid. This work analyses the technical issues, e.g., the EMS and converter control of EV charging from a MG-PV and its effective utilisation, and the maintenance of the DC-bus voltage irrespective of the utility-grid overloading caused by either local load or the inadequate use of PV power. Further, the charge controller provides closed-loop charging through constant current and voltage to reduce the charging time.

Abstract: In recent times, the study of Iridates has gained major attention due to the delicate interplay amongst various energy scales which include Coulomb correlation U, Hund’s coupling JH, spin-orbit coupling (SOC), and crystal-field splitting. The local environment of the Iridium atom is one of the crucial factors in dictating the properties of such systems. While the major focus has been on Iridates in the octahedral or tetrahedral environment, there are very few references in literature that report the rare occurrence of Iridium in a square planar environment such as in Cs₂Na₂IrO₄. The structure consists of isolated IrO₄ square planes which are oriented orthogonal to each other in each consecutive layer. In our work we study the electronic and magnetic properties of Cs₂Na₂IrO₄. Microscopic exchange interactions and Wannier function analysis reveals the quasi two dimensional AFM ground state, in spite of the absence of long-range structural connectivity. The estimated magneto-crystalline anisotropy is significantly large. The phonon modes analysis confirms very weak spin-phonon coupling and reveals the possible mechanism for the evolution of orthogonally oriented IrO₄ moieties. In the absence of any experimental study on Cs₂Na₂IrO₄, our work plays a crucial role in determining the electronic and magnetic properties of material.

Abstract: Wire arc additive manufacturing is a method where materials are layered using a molten filler wire to create desired products. The resulting manufactured products typically have similar properties to the filler wire in their initial state. In this study, a ferritic low carbon steel ER70S6 was used, and each layer deposited using the wire arc additive manufacturing technique was modified by adding a mixture of titanium and copper micro powders between the layers. The objective of this paper was to manufacture a multi-phase steel using ferritic low carbon steel as parent filler material and then investigate the metallurgical and mechanical properties of the resulting product. The manufactured wall was carefully examined using non-destructive testing equipment based on phased array ultrasonic testing, and no defects were found. To understand the metallurgical characteristics of the steel produced through wire arc additive manufacturing, various techniques such as optical microscopy, scanning electron microscopy, energy dispersive spectroscopy, and X-ray diffraction were employed to identify the different phases present in the steel. It was observed that the microhardness and the tensile strength of the manufactured steel significantly increased due to the formation of a specific microstructures such as ferritic-bainitic, with retained austenite as well as due to precipitation strengthening. Additionally, the toughness properties of the manufactured steel remained same at the room temperature while the low temperature impact strength of the CuTi added ER70S6 steel was found 125% higher compared to the as deposited ER70S6 steel.

Abstract: Globally, the utilization of anthropogenic carbon dioxide (CO2) is a pressing concern. Consequently, there is a need for a CO2 fixation technique that utilizes renewable energy sources. Molecular catalysts that use electrical energy to convert CO2 to CO, formic acid and, to some extent, methanol and methane are being widely studied. Molecular catalysts containing bipyridine ligand systems are widely known, but the terpyridine derivatives are lesser explored. The main aim of this project is to study the role of this terpyridine ligand on the electrochemical catalytic efficiency of the metal complex. The electrochemical study of rhenium complexes containing the terpyridine ligand framework shows promising CO2 reduction activity. Initial studies by replacing rhenium with earth-abundant metals also show CO2 reduction, which will be studied in detail.

Abstract: The development of sustainable organic transformations and the use of sustainable catalysis in producing valuable compounds, ranging from medicines to materials, have become significant goals in chemistry. Heterocycles are an important family of organic compounds, and indole, furan, and their derivatives are electron-rich heterocycles with interesting properties. They are found in pharmaceuticals and natural compounds, and they also have enormous importance in materials sciences. Given how important indole and furan are, the study of biheteroaryl compounds with indoles directly linked to furan through a C-C bond has gotten a lot of attention in organic chemistry, but there is still not much known about them. The primary aim of this research was to discover sustainable methods for synthesizing indole-furan-based heterocycles and to evaluate their usefulness. To achieve this objective, we developed a cross-dehydrogenative coupling-based approach for synthesizing these unsymmetric biheteroaryls. This methodology provides a more sustainable and environmentally friendly approach to the synthesis of these biheteroaryls than traditional methods. Cheap and abundant catalysts make this approach attractive for large-scale production, and it can be applied in various fields. We found that the highly conjugated indole-furan biheteroaryl motif has impressive photophysical properties, indicating the usefulness of this class of compounds.