Indian Institute of Science Bangalore
etd@IISc Electronic Theses and Dissertations at Indian Institute of ScienceNot a member yet
6204 research outputs found
Sort by
Aspects of Holography and Quantum complexity
No full textWe begin with the Principle of holography and the AdS/CFT correspondence. We begin our investigations by considering the Principle of Holography, at the level of codimension-1 screens. We argue that the mechanism of encoding the holographic data in terms of sources, condensates, and correlators is fairly general and has generalizations to a large class of spacetimes. We formulate the holographic correspondence in terms of bulk sources localized on a screen, instead of boundary values of bulk fields. We discuss the extension of the familiar notion of normalizable and non-normalizable modes to moderately general settings beyond AdS. We discuss a simple map between this prescription and the usual AdS/CFT correspondence, as well as work out explicit correlators via this prescription in flat space. Finally, we discuss the general utility of this approach of using sources to describe dynamics.
We then focus on the widely successful AdS/CFT correspondence. Specifically, we study the Bulk Reconstruction program in AdS/CFT. We discuss various aspects of the HKLL bulk reconstruction in AdS. We construct the space-like Kernel for the non-normalizable mode as a mode sum and via a Green's function approach (in even dimensions). This puts the normalizable and non-normalizable modes on equal footing. In Poincaré AdS, we delve into the technical details of this construction. We propose a spatial complexification and discuss an antipodal identification as crucial steps in obtaining a space-like Kernel in terms of the chordal distance, for certain values of scaling dimension. We also note some interesting features of this construction inside the Brietienlohner-Freedman bound, where both the normalizable and non-normalizable modes have equivalent interpretations.
At this stage, we shift gears and start addressing some problems in quantum chaos via quantum complexity. In view of the Operator Growth Hypothesis for chaotic and intergable systems, we study classically integrable systems with unstable saddle points. We find that Krylov complexity (of operators) is a hypersensitive probe of chaos. We discuss some features of Krylov complexity and autocorrelation functions. We supplement our study by numerical calculations using the Lipkin-Meshkov-Glick and Feigngold-Peres models.
Using Krylov complexity, we also study quantum many-body scars. We utilize the Lanczos mechanism to study special eigenstates for the (chaotic) PXP model which behave as states with ``low chaos'' as compared to the other eigenstates. The presence of these states indicate the emergence of some underlying symmetry, which we characterize by deformed SU(2) algebra. Finally, we study the nature Krylov spread complexity for the scar states and compare them to the generic states. We conclude by discussing a ``tight-binding'' interpretation of Krylov spread complexity
Data Driven Stabilization Schemes for Singularly Perturbed Differential Equations
No full textThis thesis presents a novel way of leveraging Artificial Neural Network (ANN) to aid
conventional numerical techniques for solving Singularly Perturbed Differential Equation (SPDE). SPDEs are challenging to solve with conventional numerical techniques
such as Finite Element Methods (FEM) due to the presence of boundary and interior
layers. Often the standard numerical solution shows spurious oscillations in the vicinity
of these layers. Stabilization techniques are often employed to eliminate these spurious
oscillations in the numerical solution. The accuracy of the stabilization technique depends on a user-chosen stabilization parameter whose optimal value is challenging to
find. A few formulas for the stabilization parameter exist in the literature, but none extends well for high-dimensional and complex problems. In order to solve this challenge,
we have developed the following ANN-based techniques for predicting this stabilization
parameter:
1) SPDE-Net:
As a proof of concept, we have developed an ANN called SPDE-Net for one-dimensional
SPDEs. In the proposed method, we predict the stabilization parameter for the Streamline
Upwind Petrov Galerkin (SUPG) stabilization technique. The prediction task is modelled as a regression problem using equation coefficients and domain parameters as
inputs to the neural network. Three training strategies have been proposed, i.e. supervised learning, L
2-Error minimization (global) and L2-Error minimization (local).
The proposed method outperforms existing state-of-the-art ANN-based partial differential equations (PDE) solvers, such as Physics Informed Neural Networks (PINNs).
2) AI-stab FEM
With an aim for extending SPDE-Net for two-dimensional problems, we have also developed an optimization scheme using another Neural Network called AI-stab FEM and
showed its utility in solving higher-dimensional problems. Unlike SPDE-Net, it minimizes the equation residual along with the crosswind derivative term and can be classified as an unsupervised method. We have shown that the proposed approach yields
stable solutions for several two-dimensional benchmark problems while being more accurate than other contemporary ANN-based PDE solvers such as PINNs and Variational
Neural Networks for the Solution of Partial Differential Equations (VarNet)
3) SPDE-ConvNet
In the last phase of the thesis, we attempt to predict a cell-wise stabilization parameter to
treat the interior/boundary layer regions adequately by developing an oscillations-aware
neural network. We present SPDE-ConvNet, Convolutional Neural Network (CNN),
for predicting the local (cell-wise) stabilization parameter. For the network training,
we feed the gradient of the Galerkin solution, which is an indirect metric for representing oscillations in the numerical solution, along with the equation coefficients, to
the network. It obtains a cell-wise stabilization parameter while sharing the network
parameters among all the cells for an equation. Similar to AI-stab FEM, this technique
outperforms PINNs and VarNet.
We conclude the thesis with suggestions for future work that can leverage our current
understanding of data-driven stabilization schemes for SPDEs to develop and improve
the next-generation neural network-based numerical solvers for SPDEs
A Ligase IV/XRCC4-dependent single strand break repair pathway for the maintenance of A/T-rich regions during DNA replication in mammals
No full textThe DNA in our cells is under constant threat from various exogenous and
endogenous sources. To ensure error free and faithful transmission of hereditary material,
human cells employ several DNA repair pathways. Mammalian cells utilize multiple DNA
repair mechanisms: base excision repair (BER), mismatch repair (MMR), nucleotide excision
repair (NER), and double-strand break repair, which includes homologous recombination
(HR), nonhomologous end joining (NHEJ) and recently discovered microhomology mediated
end joining (MMEJ) or Alt-NHEJ. While NHEJ entails the direct end-to-end joining of the
broken ends, HR requires genetic information from the homologous DNA sequence from the
partner chromosome for repair.
Often cancer progression has been associated with alterations in DNA repair genes
caused by mutations. Different NHEJ repair proteins have been shown to be upregulated in
several cancers, such as breast cancer, gastric cancer, oral squamous cell carcinoma,
oesophageal cancer, and lung cancer. Ligase IV and XRCC4 play a key role in the joining of
DSBs during NHEJ and displayed increased expression in a pan-cancer manner in
comparison to healthy cells.
In the first part of the current study, we investigated the sensitivity of LIG4 mutant
cells toward different FDA-approved drugs. For that, we have generated 6 different mutants
of LIG4 in cervical cancer and normal kidney epithelial cell lines using CRISPR-Cas9
mediated genome editing. LIG4 mutant cells exhibited compromised cell viability, reduced
NHEJ, and accumulation of DSBs due to slow repair kinetics. Interestingly, LIG4 mutant cells
showed increased MMEJ-mediated repair and an increase in sensitivity toward clinically
relevant drugs. Finally, we demonstrate that when combined with IR, LIG4 mutant cervical
cancer cells are highly sensitive to FDA-approved drugs.
Single-strand breaks (SSBs) can arise either directly e.g., from the attack of
deoxyribose by free radicals such as reactive oxygen species (ROS) or indirectly via
enzymatic cleavage of the phosphodiester backbone e.g., as normal intermediates of DNA
base excision repair (BER). The repair of direct and indirect SSBs has collectively been
termed single-strand break repair (SSBR), primarily because the same group of proteins
appears to repair both types of breaks. SSBR can be divided into four basic steps. Most
indirect SSBs are created during BER by AP endo-nuclease (APE1), and are then “handed”
to the next enzyme in the BER process in a molecular relay. SSBR starts with DNA damage
binding by Poly (ADP-ribose) polymerase (PARP) followed by DNA end processing. The
third step is DNA gap filling primarily by DNA pol β, followed by DNA ligation by DNA Ligase
III/XRCC1. To date, there has been a widespread assumption that DNA repair inside a cell is
2
being carried out irrespective of the DNA sequence. Surprisingly, in our laboratory, it was
observed that cell-free extracts prepared from various mammalian tissues could join single stranded oligomeric DNA only when thymine DNA sequences were present (Srivastava,
2013). In the present study, we have investigated the molecular mechanism of this
interesting observation. Various lines of experimentation showed that among the different
DNA ligases (Ligase I, Ligase III/XRCC1, and Ligase IV/XRCC4), only DNA Ligase
IV/XRCC4 complex could catalyse the joining of ssDNA harbouring poly Ts. This observation
is very fascinating, considering Ligase IV/XRCC4 is only known to ligate DNA double-strand
breaks during nonhomologous end joining. EMSA studies showed that Ligase IV/XRCC4
complex could bind to ssDNA Poly (T)35, however the joining of single-stranded poly T
requires Ligase IV/XRCC4 complex and cannot be catalyzed by either of the protein alone.
Biolayer interferometry (BLI) studies showed that full-length Ligase IV is required for the
stable binding of the protein to the single-stranded DNA of poly Ts. DNA bubble substrate
mimicking the DNA replication and transcription intermediates exhibited efficient joining
when poly Ts were present at the bubble. Although Ligase IV/XRCC4 could join these
substrates when poly Ts were present, the efficacy of the joining was significantly improved
when incubated with rat testicular extract. Besides, the sequence specificity remained
consistent even when investigated in double-stranded DNA context.
Importantly, single-stranded DNA containing poly thymine in dsDNA context showed
joining by rat testicular extract but not by purified Ligase IV/XRCC4 complex, suggesting
additional factors were required. The optimum joining of single-stranded DNA containing
poly thymine in dsDNA context occurred at 25ºC with 2 µg of extract in 30 min of incubation.
The optimum concentration of MgCl2 required for joining was 7.5 mM and that of ATP was
0.5 mM.
Reconstitution experiments using proteins associated with NHEJ revealed the joining
of double-stranded DNA but not those containing ssDNA with poly thymine in dsDNA
context, suggesting that additional proteins are involved in the joining of such single
stranded DNA breaks possessing long poly T sequences. To identify the additional proteins
involved in this novel repair mode, rat testicular extract was fractionated and evaluated.
Interestingly, we observed that fractions required for the joining of NHEJ DNA substrates
and ssDNA containing poly thymine in dsDNA context substrate were different, indicating the
requirement of a different set of DNA repair proteins. Pull down using biotinylated ssDNA
containing poly thymine in dsDNA context substrate after incubation with specific fractions
(with maximum joining activity) followed by mass spectrometry analysis showed different
proteins from various repair pathways including HR, NHEJ, NER, and BER. The binding of
RAD50, Ligase IV/XRCC4, RPA, CtIP, and MRE11 with ssDNA containing poly thymine in
3
dsDNA context were confirmed by EMSA, followed by western blot analysis.
Immunoprecipitation of Ligase IV followed by western analysis with supernatant fraction
showed a reduction in the level of RPA, MRE11, RAD50, Pol μ, and CtIP protein, confirming
the presence of these proteins in the complex. Further, in vitro pull-down with purified Ligase
IV/XRCC4 showed that RPA, CtIP, and RAD50 can directly interact with Ligase IV. Joining
of ssDNA containing poly thymine in dsDNA context with immunoprecipitated extract of RPA,
MRE11, RAD50, Pol μ and CtIP showed a significant reduction in the joining, suggesting the
involvement of these proteins in the thymine dependent joining of SSBs.
The previous report indicates that poly (dA:dT) tracts could be preferential sites of
polar replication fork stalling and collapse within early-replicating fragile sites (ERFSs), late replicating common fragile sites (CFSs), and at the rDNA replication fork barrier (Tubbs et
al., Cell, 2018). To understand the role of Ligase IV at the stalled fork, DNA fiber assay was
carried out in the presence of hydroxyurea (HU), which is known to induce SSBs and, thus
fork stalling. Our results showed a significant increase in fork stalling when Ligase IV was
knocked out in cells through CRISPR-Cas9 mediated gene editing, suggesting the role of
Ligase IV in replication fork restart. Interestingly, there was no effect on replication fork
restart upon KU70 knockdown suggesting the role of Ligase IV in replication fork restart,
independent of NHEJ. Further, we found that Ligase IV/XRCC4 is present at the replication
site by employing nascent DNA pull-down assay and mass spectrometry studies. Besides,
accumulation of ssDNA upon HU treatment was detected by native BrdU incorporation
assay in Ligase IV knock out (KO) cells. Similarly, a significant increase in RPA foci in
Ligase IV KO cells following HU treatment demonstrates the accumulation of single-stranded
DNA in KO cells. Importantly, ChIP sequencing analysis revealed that Ligase IV and RPA
can colocalize at poly dT/dA sequences, suggesting the binding of Ligase IV at single stranded poly dT/dA regions inside the genome. Taken together, our results suggest that
Ligase IV is involved in repairing SSBs generated at the stalled replication fork.
The ChIP assay using anti-Ligase IV showed hydroxyurea treatment dependent
binding of Ligase IV at several AT-rich fragile sites. Ligase IV KO cells showed no binding of
Ligase IV to AT-rich fragile sites, suggesting that binding was specific to Ligase IV. ChIP
sequencing with Ligase IV antibody in presence and absence of HU showed enrichment of
Ligase IV binding peaks at AT-rich regions. These results indicates that Ligase IV is involved
in the maintenance of common AT-rich fragile sites associated with several pathological
conditions.
Thus, the results of the study provide new insights into a previously uncharacterized
single-stranded DNA break repair pathway, which is dependent on DNA Ligase IV. Our
study suggests that single-strand breaks in thymine rich DNA generated during DNA
4
replication can lead to fork arrest, and under these conditions, replication restart is
dependent on the function of Ligase IV in NHEJ independent manner. We revealed that
RPA, CtIP, RAD50, MRE11, and Ligase IV are present in the complex, and the function of
Ligase IV in fork restart appears to be because of their mutual interactions. Overall, we
provide new evidence for NHEJ-independent functions of DNA Ligase IV in novel single stranded DNA break repair pathways, which might explain the crucial nature of this protein in
the survival of organisms
Privadome: A System for Citizen Privacy in the Delivery Drone Era
No full textE-commerce companies are actively considering the use of delivery drones for customer fulfillment, leading to growing concerns around citizen privacy. Drones are equipped with cameras, and the video feed from these cameras is often required as part of routine navigation, be it for semi-autonomous or fully-autonomous drones. Footage of ground-based citizens captured in these videos may lead to privacy concerns.
This paper presents \pd, a system that implements the vision of a virtual privacy dome centered around the citizen. \pd\ is designed to be integrated with city-scale regulatory authorities that oversee delivery drone operations and realizes this vision through two components, \pdmpc\ and \pdros. \pdmpc\ allows citizens equipped with a mobile device to identify drones that have captured their footage. It uses secure two-party computation to achieve this goal without compromising the privacy of the citizen's location. \pdros\ allows the citizen to communicate with such drones and obtain an audit trail showing how the drone uses their footage and determine if privacy-preserving steps are taken to sanitize the footage. An experimental evaluation of \pd\ shows that the system scales to near-term city-scale delivery drone deployments (hundreds of drones). We show that with \pdmpc\ the mobile data usage on the citizen's mobile device is comparable to that of routine activities on the device, such as streaming videos. We also show that the workflow of \pdros\ consumes a modest amount of additional CPU resources and power on our experimental platform
Supervised Learning Approaches for Language and Speaker Recognition
No full textIn the age of artificial intelligence, it is important for machines to figure out who is speaking automatically and in what language - a task humans are naturally capable of. Developing algorithms that automatically infer the speaker, language, or accent from a given speech segment are challenging problems that have been researched for over three decades. The main aim of this doctoral research was to explore and understand the shortcomings of existing approaches to the problems and propose novel supervised approaches to overcome these shortcomings to develop robust speaker and language recognition systems.
In the first part of this doctoral research, we developed a supervised version of a popular embedding extraction approach called the i-vector, typically used as front-end embeddings for speaker and language recognition. In this approach, a database of speech recordings (in the form of a sequence of short-term feature vectors) is modeled with a Gaussian Mixture Model, called the Universal Background Model (GMM-UBM). The deviation in the mean components is captured in a lower dimensional latent space called the i-vector space using a factor analysis framework. In our research, we proposed a fully supervised version of the i-vector model, where each label class is associated with a Gaussian prior with a class-specific mean parameter. The joint prior (marginalized over the sample space of classes) on the latent variable becomes a GMM. The choice of the prior distribution is motivated by the Gaussian back-end, where the conventional i-vectors for each language are modeled with a single Gaussian distribution. With detailed data analysis and visualization, we show that the s-vector features yield representations that succinctly capture the language (accent) label information, and do a significantly better job distinguishing the various accents of the same language. We performed language recognition experiments on the NIST Language Recognition Evaluation (LRE) 2017 challenge dataset, which has test segments ranging from 3 to 30 seconds. With the s-vector framework, we observed relative improvements between 8% to 20% in terms of the Bayesian detection cost function, 4% to 24% in terms of EER, and 9% to 18% in terms of classification accuracy over the conventional i-vector framework. We also performed language recognition experiments on the RATS dataset and Mozilla CommonVoice dataset, and speaker classification experiments using LibriSpeech, demonstrating similar improvements.
In the second part, we explored the problem of speaker verification, where a binary decision has to be made on a test speech segment as to whether it is spoken by a target speaker or not, based on a limited duration of enrollment speech. We proposed a neural network approach for back-end modeling. The likelihood ratio score of the generative PLDA model was posed as a discriminative similarity function, and the learnable parameters of the score function are optimized using a verification cost, proposed to be an approximation of the minimum detection cost function (DCF). The speaker recognition experiments using the NPLDA model are performed on the speaker verification task in the VOiCES datasets and the SITW challenge dataset. Further, we explore a fully neural approach where the neural model outputs the verification score directly, given the acoustic feature inputs. This Siamese neural network (E2E-NPLDA) model combines the embedding extraction and back-end modeling stages into a single processing pipeline. The development of the single neural Siamese model allows the joint optimization of all the modules using a verification cost. We provide a detailed analysis of the influence of hyper-parameters, choice of loss functions, and data sampling strategies for training these models. Several speaker recognition experiments were performed using Speakers in the Wild (SITW), VOiCES, and NIST SRE datasets where the proposed NPLDA and E2E-NPLDA models are shown to improve over the state-of-art significantly x-vector PLDA baseline system
Atto-second Dynamics in Molecular Systems
No full textThe thesis describes the attosecond dynamics of small molecular systems, focusing on ionization delay times and hole migration delay times. These time delays are some of the fastest events in electronic dynamics. When an electromagnetic pulse with sufficient energy interacts with a molecule, an electron is released to the free state and
subsequently a positive charge, a hole, is created in the molecule. The time scale of the hole dynamics in the molecule is comparable with the time scale of the ionization process. We have used simple mathematical models and molecular quantum chemistry calculations to analyze the coupled ionization process and hole dynamics focusing on their time scales
Understanding cholesterol homeostasis and its impact on GPCR signalling in aging
No full textFull text embargo One year up to July 2024
Aging is defined as the time-related deterioration of physiological functions necessary for survival of organisms. Despite significant progress in the extension of human lifespan due to modern medicine, improved sanitation and nutrition, the undesirable effects associated with aging have not been alleviated. There is increasing evidence that aging occurs at the cellular level and cellular senescence is a major contributing factor to the process of aging. Accumulation of cellular damage over the years results in tissue dysfunction leading to organismal aging. Thus, it is important to understand the molecular basis of aging and identify possible therapeutic intervention approaches and targets. Senescence is associated with high inflammation as well as high oxidative and nitrosative stress which causes persistent damage to cellular components such as DNA, lipids and proteins. In this study, I have examined the effect of altered cholesterol homeostasis on CXCR4 signalling during cellular senescence.
The chemokine receptor CXCR4 is upregulated in senescent cells and however the signalling upon stimulation is impaired. However, during senescence the CXCR4 mediated calcium oscillations are disrupted and signalling is altered. To investigate the mechanism of altered signalling, I studied various steps in the signalling cascade, starting at the membrane level. Plasma membrane composition plays an important role in receptor organisation, dynamics and signalling and senescent cells show changes in several membrane properties. Cholesterol is the most abundant lipid species found in eukaryotic membranes which when oxidised disrupts the membrane dynamics. The oxidation of cholesterol to form oxysterols during aging led to altered CXCR4 signalling, which was recorded as delayed or disrupted calcium oscillations after ligand stimulation. Senescent cells behave like a cholesterol depleted and oxysterol rich environment. CXCR4 receptor signalling through the Gi G-protein, however the presence of oxysterols switches the affinity towards Gs leading to an anti-inflammatory arm of signalling.
The cholesterol homeostasis pathway is very tightly regulated leading to balanced free cholesterol levels inside the cell. The total cholesterol levels were found to be very high in senescent cells, at the same time the cell size is also larger than non-senescent cells. On analysis of per unit area cholesterol, the senescent cells had lower cholesterol per unit area and higher oxysterols. To understand this dysregulation in the homeostasis pathway, I analysed the proteins which are involved in the cholesterol biosynthesis and its efflux from cells. The cholesterol biosynthesis genes like HMGCS were found to be much lower in senescent cells, while the efflux genes like ABCA1 were high. This indicated that senescent cells were constantly eliminating cholesterol and oxysterols from cells.
Oxysterols are toxic to the cells in high amounts and need to be eliminated upon generation. The cells maintain a fine balance between the cholesterol and oxysterol levels, which is disrupted during senescence. I investigated the effect of cholesterol biosynthesis and efflux inhibition on the viability of senescent cells. Senescent cells died when treated with biosynthesis inhibitor Simvastatin, or efflux inhibitor GSK2033. This led to the discovery of a a novel mechanism for the maintenance of senescent cell viability and identified targets for senolytic development. Targeting the cholesterol homeostasis pathway for development of novel senolytics and effects on health-span of acceler-aged mice were also explored. Simvastatin, which is a novel potential senolytic, was used to study aging signatures in radiation induced aging mouse model. Simvastatin reduced liver injury and inflammation in irradiated mice. The present study has implications in cancer co-therapies, age-related disease prevention as well as rejuvenation therapies. Cholesterol homeostasis pathway can be targeted to develop novel compounds and therapies that reduce age related injury and systemic inflammation
Empirical Study on a Class of Problems for Omnichannel Retailing in India with Special Reference to Apparel Industry
No full textOmnichannel retailing is an integrated experience that melds the advantages of physical stores with information-rich online shopping. To offer a seamless omnichannel shopping experience, retailers need to understand what drives customers towards omnichannel and what dissuades them away from omnichannel. Also, retailers need to build cross channel synergies with effective decision-making regarding order fulfillment. Thus, with this premise, this study addresses a class of problems related to omnichannel with specific research objectives to (a) understand customer-specific drivers and barriers for the adoption of omnichannel (b) rank those drivers and barriers according to their importance (c) classify omnichannel customers according to sociodemographic characteristics and buying behavior, and (d) propose a solution methodology for fulfillment decision problem in omnichannel. To address these objectives, this study is confined to the apparel industry in India and the required primary data is collected from Mumbai and Bangalore as they are among top five Indian cities in terms of omnichannel orders and their price-value.
For addressing the first research objective, 10 unique drivers for the adoption of omnichannel retailing in the Indian apparel industry (D-AOCRIAI): Improved Shopping Experience, Reduced Effort, Social Influence, Habit, Hedonic Motivation, Technology Development, Enhanced Promotion, Sporadic Event, Personalization, and Integrated Supply chain and 9 unique barriers (B-AOCRIAI): Inconsistency in Offering, Channel Discord, Lack of Trust, Data Privacy Concern, Lack of Infrastructure and Resources, Psychological Hinderance, Poor Customer Support, Inefficient Order Fulfillment, and Difficulty due to Sporadic Events and their measurement variables are identified. Total Interpretive Structural Modelling (TISM) and Decision Making Trial and Evaluation Laboratory (DEMATEL) approaches are used to propose an initial conceptual framework for each of D-AOCRIAI and B-AOCRIAI using data from 12 domain experts and 23 customers.
The proposed conceptual framework for each of D-AOCRIAI and B-AOCRIAI is statistically validated and finalized by following descriptive research methods including confirmatory factor analysis and structural equation modeling. The data required for statistically finalizing the proposed framework is collected from 850 customers comprising of 448 adopters and 402 non-adopters of omnichannel by developing an appropriate questionnaire. From the total variance (R2) explained with respect to D-AOCRIAI (R2 = 0.86) and B-AOCRIAI (R2 = 0.71), final version for each of the frameworks is constructed. Further, it is evident that Drivers: Improved Shopping Experience, Reduced Efforts, and Social Influence directly and positively impact omnichannel adoption. Also, Barriers: Lack of Trust, Psychological Hindrance, and Lack of Infrastructure and Resources directly and negatively impact omnichannel adoption.
To address the second research objective, the identified unique drivers and barriers are ranked using MCDM methods: Analytical Hierarchy Process (AHP), and Best Worst Method (BWM) respectively using data collected from the group of 36 customers by developing suitable questionnaire for each of these MCDM methods. From the results obtained, it is observed that the most important drivers are: Reduced Effort, Improved Shopping Experience and Enhanced Promotion. Similarly, the most important barriers are Data Privacy Concerns, Psychological Hinderance, and Inconsistency in Offering.
Classification models are developed with approaches: decision tree, random forest, and adaptive boosting to address third objective – that is classifying the customers into adopters and non-adopters of omnichannel based on their sociodemographic characteristics using data from all 850 respondents. It appears from the results that Age, Profession, and Income were found to be the most significant sociodemographic characteristics impacting the adoption of omnichannel. Further, adopters of omnichannel are segmented based on the buying behavior (Recency, Frequency, and Monetary Value) using K-Means clustering into 4 different clusters: Omni-connected, Omni-consistent, Omni-spenders, and Omni-hesitant.
Finally, to address the fourth research objective of the research, a solution methodology for the decision problem of online order fulfillment in omnichannel is developed by formulating the problem as a mixed integer linear programming (MILP) model with the objective of cost-minimization. The problem configuration considers a deterministic demand over multiple periods and computes product flow across various locations in a retailer’s network. From the optimal solution obtained, it appears that fulfillment of online orders through ‘warehouse’ and ‘direct-to-customer center’ is more cost effective.
The research problems considered in this study can serve as empirical evidence to assist retailers in developing cluster-wise omnichannel strategies. Retailers should focus on offering an improved, personalized, and convenient shopping experience while ensuring data privacy and mitigating inconsistencies. Though this research has achieved all the planned objectives, there are certain limitations. The inferences of the study cannot be generalized at the pan-India level across different product categories. The proposed classification models for clustering use self-reported ordinal values from customers regarding their buying behavior. The proposed mathematical model for order fulfillment uses deterministic demand. Future research could validate the proposed models for across different geographies and product categories. Also, the independent decision problem of ‘omnichannel fulfillment’ can be treated as an integrated decision problem with pricing, inventory, and last-mile delivery
Advances in Architecturing of Large-Scale Photovoltaics: Packaging and Incidence Angle Agnostic Hierarchical Nanostructures
No full textEncapsulant of Graphene embedded polymer that exhibits low permeability to moisture and oxygen were used as a packaging material for OPVDs. Graphene-Surlyn encapsulated and unencapsulated devices were stored both in ambient and inert conditions to evaluate the performance of the encapsulant material. The devices with encapsulation and stored inside the glove box exhibited 88% of the initial performance after more than 50000 hours of real-time ageing. These results demonstrate an enormous increase in the lifetime of the devices, paving the
way for robust and long-life OPV devices.
Hierarchically nanostructured coatings (HNC), crucial for adequate light trapping and enhancing the capacity of silicon solar cells to convert light into energy, provide the potential for better photon management. To absorb light from virtually any angle, an effort is made to emulate the biomimetic design of a particular type of insect. We coated the HNC to polycrystalline silicon solar cells (epoxy) using a thermosetting polymer and the PDMS stamping technique. The hierarchically built solar cell outperformed solar cells made of bare silicon by around 25% in terms of power production at a maximum angle of incidence of 41.53°.
The hierarchically formed solar cell might replace the demand for tracking-based
solar panels. Superior optoelectronic properties are embraced by this self-assembled HNC method
for a range of solar cell applications. The HNC on silicon solar cells was performed on larger (1m
x 2m) panels to assess the scalability of these structures. The performance improvement of the
panels was ~80% of that of the small-scale version, with ~20% enhancement in the cumulative
power generation throughout the day.
To tune the optical losses caused by refractive index mismatch and spectrum conversion from the IR to the visible area, which can be a potential rival in silicon photovoltaics, this work has included both HNC and up-conversion material (NaYF4:Yb/Er). Infrared light absorbers and green, blue, and red light emitters are two functions of the rare-earth-doped up-conversion nanoparticles employed in this thesis. When subjected to constant IR irradiation for 15 days, the integration of NaYF4:Yb/Er up-conversion nanoparticles decreased the temperature of the cells by a margin of around 20%. A 32% increase in power output over reference solar cells was also discovered
An MLIR-Based High-Level Synthesis Compiler for Hardware Accelerator Design
No full textThe emergence of machine learning, image and audio processing on edge devices has motivated research towards power-efficient custom hardware accelerators. Though FPGAs are an ideal target for custom accelerators, the difficulty of hardware design and the lack of vendor-agnostic, standardized hardware compilation infrastructure has hindered their adoption. High-level synthesis (HLS) offers a more compiler-centric alternative to the traditional Verilog-based hardware design improving developer productivity.
Though HLS offers many advantages over traditional HDL-based hardware design flow, it is still not a mature ecosystem. There is a need for research in both programming abstraction for hardware design and compiler optimizations to meet the efficiency of hand-optimized HDL designs. In the software world, LLVM has enabled rapid prototyping of programming languages. A new programming language can target the LLVM compiler and benefit from the existing optimizations and backend code generation. LLVM also enables the development of new compiler optimizations. A new optimization pass can be plugged into the existing compiler pipeline to evaluate its benefits on existing programming languages and benchmarks. This decoupling of different stages of the compiler pipeline can be largely attributed to the LLVM intermediate representation.
The high-level synthesis ecosystem still lacks such extensible modular compiler infrastructure which could be used for the development of new HLS programming languages and optimizations. In this work, we propose an MLIR based end-to-end HLS compiler and an intermediate representation that is suitable for the design and implementation of domain-specific accelerators for affine workloads. Our compiler brings similar levels of modularity and extensibility to the HLS compilation domain, which LLVM brought in the area of software compilation. A modular compiler infrastructure offers the advantage of incrementally introducing new language frontends and optimization passes without the need to reinvent the whole HLS compiler stack.
Our compiler converts a high-level description of the accelerator specified in the C programming language into a register-transfer-level(RTL) design (SystemVerilog). We use memory dependence analysis and integer-linear-program(ILP) based automatic scheduling to improve loop pipelining, and introduce parallelization between producer-consumer kernels. Our ILP-based optimizer beats the state-of-the-art Vitis HLS compiler by 1.3x in performance over a representative set of benchmarks, while requiring fewer FPGA resources