Abstract

World Health Organization (WHO) estimates that approximately 5% of adults suffer from mental issues related to depression. The rate of increase in the number of patients suffering from depression is significantly higher that the available medical practitioners (specifically psychiatrists). As a result, diagnosis and treatment can be expensive. Alternative methods are being looked into for assessment of depression. In recent times, emerging technologies like Virtual Reality (VR) has been proposed to assess mental health issues in therapeutic settings. VR based applications provide personalized motivation and meaningful engagement for assessment and treatment. External sensors or devices aid VR applications in assessing physiological and behavioural measures related to depression. One behavioural measure is gait i.e., the pattern of walking of an individual. Gait has been identifed as an important parameter that changes with the mental state of a person and gait analysis techniques can be used to aid medical practitioners in identification of depression. The focus of this thesis is on designing and building a low-cost customizable mat for gait analysis. As part of the thesis, we study the use of Virtual Reality technology in healthcare in general and more specifically on the detection of depression. We performed a systematic literature review, the results from the review helped us to categorise the different physiological and behavioural measures related to depression. We put together a list of VR applications that were used in the assessment of depression and developed a bare minimum checklist to design the scenes within VR applications for diagnosing depression. We developed a mat using Velostat, a piezoresistive based low-cost sensing material for performing gait analysis. Initially, small size mats were designed and were analysed for crosstalk, full scan time, sensitivity and range. Then, a 1.8 x 1.8 feet mat was designed, and one of the gait parameters was calculated from it and compared with a commercial mat called Sensing Tex with an accuracy of 95%. After multiple iterations, we designed a 10.5 x 2 feet pressure sensing mat and used it to perform experiments. VR applications with scenes that evoke emotions like happiness, sadness and calmness were designed and used in the experiments. The results of the experiments showed positive as well as negative correlations between different gait parameters and their PHQ-9 values. Additionally differences in gait like the differences in stride length, velocity, and cadence were observed in the depressed and the non depressed were using descriptive statistics.However, inferential statistics were not conducted due to the limited sample size.

Abstract

The impact of climate change on water quality variables is an essential topic for sustainable river water quality management in a warming environment and is a great environmental concern worldwide. River Water Quality (RWQ) models aim to simulate the behavior of various water quality variables in response to pollutants, land use changes, and climate change. However, these water quality models suffer from sparse data leading to data uncertainty. In the past decades, different models have been successfully used for RWQ modeling under different spatial and temporal scales. To simulate RWQ variables, physically based water quality models can be used, but they require large amounts of site-specific detailed data, including stream geometry, meteorological variables, and hydraulic properties of the river, which are unavailable for many river systems globally. However, unlike processbased models, statistical models possess many advantages. Additionally, statistical models do not require a large number of input variables, which are unavailable for many ungauged river systems. However, accurately describing the nonlinear characteristics of a data series is a significant shortcoming of this approach. To overcome such limitations, artificial intelligence algorithms, i.e., Machine Learning (ML) techniques, are widely used to address a range of nonlinear prediction problems. Such models are suited for information extraction from sequential data in RWQ modeling, and they serve functionalities to build models using a reduced number of variables with more accurate simulation. Machine Learning (ML) has been increasingly adopted due to its ability to model complex and nonlinearities between river water quality (RWQ) variables and their predictors (e.g., Air Temperature, AT, streamflow). To simulate RWQ parameters using data-driven algorithms, more input variables are required, which are unavailable for many ungauged river systems. Climatic variables that are readily available are the maximum, minimum, and average AT to build RWQ models with more accurate simulation and higher computational efficiency. In this context, most of these ML approaches have been applied without any detailed sensitivity analysis to identify the most influencing variables to be considered in the prediction of RWQ variables. Furthermore, the development of systematic models combined with ML under minimum data input variables has not been intensively studied in predicting RWQ variables. To address these, the present study first demonstrates how new ML approaches, such as Ridge regression (RR), K-nearest neighbors (KNN) regressor, Random Forest (RF) regressor, and Support Vector Regression (SVR), can be coupled with Sobol’ global sensitivity analysis (GSA) to predict accurate RWQ variables estimates. Air Temperature (AT) changes can affect River Water Temperature (RWT) under anthropogenic climate change, the primary variable that influences water quality. Therefore, the present study selected RWT as a water quality variable prediction with a tropical river system of India, Tunga-Bhadra River, as a case study. Further, the proposed ML approaches have been combined with the Ensemble Kalman Filter (EnKF) data assimilation (DA) technique to improve the predicted values based on the measured data. Overall, the study concluded that the SVR has been noted as the most robust ML model when coupled with a global sensitivity algorithm and DA techniques to predict RWT at a monthly time scale compared to daily and seasonal. Also, the study concluded that the SVR model is a strong choice for smaller datasets and is less sensitive to outliers in the data compared to some other models. The SVR is generally less computationally expensive than the ML models. Another data uncertainty is the lack of availability of long-time series data to capture interannual variability and consistent water quality measurement datasets in RWQ modeling. Generally, RWQ data availability is on a monthly scale and is burdened with a large number of missing values with limited durations. In this context, the selection of appropriate model inputs, development of models under limited data, processing of non-stationary data, seasonality scenarios, and different potentially influenced relevant lags of variables have not been intensively investigated in the literature, especially in the case of estimation of RWQ variables. Given the missing, limited, and non-stationary data scenarios, the present thesis developed hybrid models for RWQ variables prediction using Long Short-Term Memory (LSTM), integrated with (i) k-nearest neighbor (k-NN) bootstrap resampling algorithms (kNN-LSTM) to address the data-limitations and (ii) discrete wavelet transform (WT) approach (WT-LSTM) to address the time-frequency localized features. To demonstrate the prediction of RWQ variables and to assess the impact of climate change on the river water quality parameters, t

Abstract

Epoch is the instant at which significant excitation of vocal tract filter takes place during the production of speech phonation. The extraction of epochs helps in speech enhancement and multi-speaker separation. For most voiced speech, the significant excitation takes place around the instant of glottal closure. Several methods have been proposed for estimating the instants of glottal closure from speech signal without using the electro-glottograph (EGG) signal. These methods are based on short-time energy of the speech signal, predictability of an all pole linear predictor, and properties of group-delay. From lot of different methods, Zero Frequency Filtering (ZFF) and Zero Phase Zero Frequency Resonator (ZP-ZFR) is one of the the best and simplest method to find prominent locations of epochs which gives highest accuracy detection rate, and it also outperforms many other parameters. This thesis work focuses on the implementation of the ZP-ZFR algorithm and its application on Field Programmable Gate Array (FPGA). The ZP-ZFR algorithm explains its principles and advantages over the traditional ZFF algorithm. The implementation details of both algorithms on FPGA are presented, including the software simulation phase using MATLAB and the subsequent implementation using FPGA boards. Moreover, the application ZP-ZFR based Voiced Activity Detector (VAD) is implemented. The primary objective is to evaluate the effectiveness and efficiency of the ZP-ZFR algorithm and detecting voiced and unvoiced segments in speech signals which are useful for real time applications. FPGAs are well-suited for also many applications, such as object detection, tracking, and recognition, radar systems, digital signal processing (DSP) and mainly for vision based applications. FPGA implementation is carried out because, these are inherently parallel devices, allowing multiple operations to be executed simultaneously. This makes them well-suited for applications that require realtime processing, such as speech recognition and speech synthesis etc. FPGAs can also be customized to accelerate specific tasks by implementing dedicated hardware. This is highly beneficial for applications requiring real-time processing or demanding computational tasks. It is also known for low-latency operations and can be more power-efficient for specific tasks.

Abstract

Advertising is a part of marketing activity and brings awareness to potential viewers about the prod- ucts/services. It helps businesses to increase sales and create a brand image of the products. Online advertising is a form of advertising which uses the Internet to promote products/services to viewers. With the proliferation of eCommerce, online advertising has become a popular mode of advertising. It enables a higher reach of users at reduced costs. Search engine advertising, display advertising, social media advertising, native advertising, video advertising, and email advertising are some of the modes of online advertising. In this thesis, we have made an effort to improve display advertising. Web banners or display advertising are important among the types of online advertising. Display advertising mainly consists of visitors, advertisers, and the publisher. The publisher owns the website (a collection of web pages). A web page contains ad slots. Advertisers are interested in placing the banners in the ad slots to expose banners to potential users. Normally, given a website, several advertisers bid for slots. Given the website, click stream data of user visits, and the demands of multiple advertisers for user visits, the issue is to develop an approach to allocating ad slots so that the maximum number of advertisers’ demands can be satisfied and the publisher’s revenue can be improved. The naive approach suffers from the issues of ad repeatability and underutilization of slots. Existing works addressed this problem by modeling it as an optimization problem by employing reinforcement learning and data mining techniques. Also, in the literature, there is an effort to solve the problem by exploiting the knowledge of coverage patterns and assuming a single slot per web page. However, there are multiple slots per page in practice, and the existing approach can not be extended to a practical scenario. In this thesis, we have tried to improve the existing coverage pattern-based approach, which has been proposed by assuming a single slot per page by considering a practical scenario, i.e., by considering multiple ad slots for each web page. As part of this effort, it was noted that the existing coverage pattern algorithms are not scalable to very large-size click-stream transactions. So, in this thesis, to improve the scalability of the existing approach, we have developed a distributed approach to mine coverage patterns; next, we have proposed an ad slot allocation approach by considering multiple ad slots per page. In the proposed distributed coverage pattern mining (DCPM) approach, we employ a notion of the summarized form of Inverse Transactional Database (ITD) and replicate it at every node. We also em- ploy an efficient clustering-based method to distribute the computational load among the Worker nodes. We have performed extensive experiments using two real-world datasets and one synthetic dataset. The results show that the proposed approach significantly improves the performance over the state-of-the-art approaches in terms of execution time. We have also proposed a framework for display advertising by considering multiple slots per page. In this framework, we employ the DCPM approach and compute the ad-slot patterns ASPs. We then calculate the impressions of each ASP and propose an efficient allocation approach to meet the advertiser demand in the form of impressions. Our extensive performance evaluation using two real-time click- stream datasets demonstrates the efficiency of the proposed framework in terms of improved publisher revenue and reduced ad repeatability. Ad revenue is vital for several e-commerce companies and revenue from display advertising is one of the opportunities. We have proposed a scalable pattern mining approach to improve the publisher’s rev- enue through display advertising. We hope that the proposed framework will facilitate the improvement of the approaches that are being followed by display advertisement companies

Abstract

Dementia is a chronic and progressive syndrome that affects the cognitive functioning of an individual. Alzheimer’s, a neurodegenerative disorder, is the leading cause of dementia. The only way to control the progression of the disease is its early detection, followed by drug and non-drug interventions. The speech production chain is the presentation of cognitive abilities and is majorly affected in the early stages of Alzheimer’s disease. Speech signals are ubiquitous and facilitate easy recording, storage, and transfer. For these reasons, researchers have long strived to develop complementary tools for Alzheimer’s Dementia detection using acoustic and linguistic clues derived from speech utterances. This thesis is one more attempt at finding the differentiating auditory and linguistic clues in the utterance(audio and transcript) of an Alzheimer’s Dementia patient. The ADReSS INTERSPEECH-2020 and ADReSSo INTERSPEECH-2021 challenges provide a balanced dataset for the Alzheimer’s Dementia classification task. This thesis explores the efficacy of different acoustic features to capture distinct patterns in the speech utterance of AD patients. The features are obtained by evaluating Cepstral Coefficients over different acoustic algorithms and techniques. Mel-frequency and Linear Prediction methods are used to capture the Vocal tract characteristics; Residual Coefficient, Glottal Volume Velocity, and Zero Frequency Filtering approach for excitation source characteristics; Envelope Modulation Spectrum and Long Term Averaging Spectrum capture the prosody characteristics of speech utterance. The next part of this thesis explores the Single-frequency-filtering-based(SFF) high spectrotemporal resolution feature using the filter bank approach for Alzheimer’s Dementia detection. Experiments are performed using different machine learning classifiers over the acoustic features extracted from the challenge datasets. The current study also demonstrates the performance of the BERT model for the dementia classification task. Finally, the performance of individual and combined acoustic features is reported. Also, the classification score is evaluated by score level fusion of the acoustic models and BERT model to observe the complementary characteristics of acoustic features to the BERT model. Acoustic models perform best when combined with linguistic features, suggesting the complementary nature of acoustic and linguistic features. Also, the high spectro-temporal resolution Single Frequency Filtering feature captures the characteristics of speech patterns better for Alzheimer’s Dementia classification than traditional source-filter model-based features.

Abstract

Discovering new reaction pathways lies at the heart of drug discovery and chemical experimentation. Chemical patent texts contain new reactions and reaction pathways. Thus, a huge amount of drug reaction data lies in unannotated patent texts which are not machine-readable. Reaction roles play an important part in analysing chemical pathways, and tracing chemicals through them, and while there is a vast body of chemical data available, the unavailability of reaction role annotated data is a blocker to effectively deploy deep learning methods for reaction discovery. To overcome this hurdle, this work introduces a new dataset, WEAVE 2.0, an expansion of the existing WEAVE dataset, a chemical NER dataset obtained from chemical patents. WEAVE 2.0 augments WEAVE named entities along with full, manual, annotations of novel chemical reactions with reaction role information. We provide baseline models for chemical entity recognition from our raw dataset, using simple architectures commonly used for chemical NER and related tasks, such as biomedical NER. As the size of the dataset is small, we introduce dataset augmentation techniques to improve learning. These techniques can be used to generate further data from other patent-based datasets, such as WEAVE [17]. We also introduce and test improved models, which structure the problem as two smaller parts instead of one, both against the raw dataset, as well as data augmented using the above methods. Further, we compare our best models against other similar datasets for chemical NER, showing its performance across multiple similar tasks. Our dataset and associated models form the foundation of neural understanding of chemical reaction pathways via reaction roles and will allow models trained for downstream tasks to utilise this information to generally lead to better predictions.

Abstract

In the current digital era, there has been an exponential surge in the demand for data storage and retrieval. This escalating need is propelled by a diverse range of applications, spanning from cloud computing and data centers to mobile devices and embedded systems. NAND flash memory, a non-volatile storage technology, plays a pivotal role in meeting these demands and is extensively employed in solid-state drives (SSDs), memory cards, and various other applications. Its notable feature is its high storage density, enabling the storage of substantial volumes of data in a compact and efficient form factor. Industries have recently favored 3D NAND Flash over Traditional NAND Flash, where the ”3D” denotes the vertical stacking of memory cells in multiple layers, akin to a skyscraper, to amplify storage density. While this stacking brings benefits, it presents challenges for analog circuit design engineers. As the number of layers increases, so does the load capacitance, introducing complexities in design as each new generation witnesses a rise in Load Capacitance. This challenge persists despite the growing number of vertical layers, particularly affecting the development of a robust and stable Low Dropout Regulator (LDO) amidst the increasing load capacitance in each generation. Moreover, the escalating layer count necessitates a proportionate increase in the number of LDOs to adeptly supply power to this expansive distributed network. To tackle the aforementioned challenges, the thesis introduces a novel LDO featuring a Flipped Voltage Follower based driver (FVF), deviating from conventional drivers. This innovative design offers increased robustness, especially when applied to a 3D NAND Flash memory array, surpassing the capabilities of traditional architectures. The thesis advocates for a dualloop scheme to enhance the transient performance metrics of the LDO. This proposed scheme not only diminishes settling time but also reduce the overshoot/undershoot voltage in comparison to existing state-of-the-art designs. Moreover, the thesis demonstrates how this structural approach can be extended to efficiently handle extensive distributed loads.

Abstract

General Circulation Models (GCMs) aid in developing climate-resilient policies, preparing for extreme events and implementing governance and disaster management mitigation strategies. Thus, assessing GCMs’ climate forecasting capabilities is crucial to establishing the credibility and reliability of climate projections, facilitating well-informed decision-making and climate change readiness. The UK Met Office Hadley Centre Coupled Model, version 3 (HadCM3), is extensively evaluated in this study. HadCM3 is used to simulate global and regional climate patterns and has been crucial to climate change assessments. This thesis explores the spatio-temporal dynamics of temperature in contiguous Peninsular India by identifying regions of high & low variability, with an emphasis on Elevation-Dependent Warming (EDW). The central objective of this study is to improve our understanding of the attributes, patterns, trends, and fundamental factors that affect temperature variations across diverse spatial and temporal dimensions and the HadCM3 model’s ability to accurately represent this phenomenon. Through a 30-year analysis of observed and modelled minimum (Tmin) and maximum temperature (Tmax) data (1991-2020 and 1990-2019, respectively) over the study region, this study explores the relationship between elevation and warming trends. The observed and modelled Tmin and Tmax have been increasing at a rate of +0.13°C/decade and +0.17°C/decade for observed data & +0.43°C/decade and +0.45°C/decade for modelled data, whereas the observed and modelled diurnal temperature range (DTR) increased at +0.03°C /decade & +0.02°C/decade. The lapse rates of observed and modelled minimum and maximum temperatures are found to be positive, but the DTR lapse rate is found to be negative, indicating increasing DTR with elevation. Modelled data shows more pronounced trends (+6.344°C/km for Tmin, +5.954°C/km for Tmax, and -0.39°C/km for DTR) than observed temperature (+2.376°C/km for Tmin, +2.341°C/km for Tmax, and -0.193°C/km for DTR). This discrepancy suggests that the global average lapse rate (+6.5°C/km) underlines the modelled data, resulting in much higher lapse rates in the study region, which implies that the model uses a uniform lapse rate to model temperature and fails to account for the study region’s unique topographic temperature relationships. The study proposes the P-MiSTIC method, a multivariate extension of MiSTIC, to identify zones with spatio-temporal consistency and investigate these inconsistencies. This study uses Tmin and Tmax as P-MiSTIC inputs with weights of -1 and 1. Using MiSTIC, the study finds 12, 9, 8, and 4 spatio-temporally invariant zones for observed and modelled Tmin and Tmax, respectively. Importantly, the elevation-based change rates of variables for these zones (-2.57°C/km and -2.19°C/km for observed Tmin and Tmax, and -7.48°C/km and -6.67°C/km for modelled Tmin and Tmax) closely match the data-derived lapse rates. Using the P-MiSTIC method, 11 and 1 zones are identified for observed and modelled data, respectively, indicating that modelled data is consistent over the Peninsular region. DTR change rates for zones within observed temperature data increase with elevation, indicating elevation-dependent warming in specific study regions. In particular, the Western Himalayan region and the Karakoram region, with the highest elevations, drive the phenomenon in the study region with the Western Himalayan and Karakoram DTRs increasing at 0.19°C/decade and +0.28°C/decade over the study period. The GCM model’s DTR data shows a much slower increase of +0.02°C/decade in the study region. Using the P-MiSTIC method on three decade-wise subsets, only one zone is identified in modelled data across all decades, while observed data identifies 12, 11, and 9 zones for decades 1, 2, and 3. The elevation-based DTR change rates for the observed temperature zones are estimated at -0.03°C/km, -0.028°C/km, and +0.533°C/km for decades 1, 2, and 3. DTR trends change from decreasing to increasing as the study progresses from decades 1 and 2 to decade 3, suggesting elevation-dependent warming in the study region is accelerating. The Western Himalayas and the Karakoram experience considerable variations of DTR. In the Western Himalayas, the DTR decreases at -0.14°C/decade, -0.29°C/decade, and - 0.41°C/decade for decades 1, 2, and 3, while in the Karakoram, it increases at +0.11°C/decade, +0.13°C/decade, and +2.46°C/decade. However, the model’s DTR data shows a uniform increase of +0.02°C/decade across all decades, indicating the model’s uniform rate. By dividing the datasets into DJF, MAM, JJA, and SON seasons, the study examined seasonal spatio-temporal variations. Seasonal observations identified 11, 10, 7, and 8 zones where the diurnal temperature range (DTR) changed with elevation at -0.518°C/km, -0.089°C/km, 0.668°C/km, and 0.453°C/km for the four seasons. Western Himalayan and Karakor

Abstract

The goal of rendering is to produce a photorealistic image of the given 3D scene description. Physically based rendering simulates the physics of the light as it travels and interacts with objects in the scene before finally reaching the camera sensor. Monte Carlo methods have been the go-to approach for physically based rendering. They are general and robust but introduce noise and are computationally expensive. Recent advancements in hardware, algorithms, and denoising techniques have enabled realtime applications of Monte Carlo methods. However, complex scenes still demand high sample counts. In this thesis, we explore and present the utilization of analytic and neural approaches for physically based rendering. Analytic methods offer noise-free renderings but are less general and may introduce bias. In recent years, neural-based approaches have gained traction, offering a balance between generality and computational efficiency. We compare and contrast the traditional Monte Carlo-based methods and emerging analytic and neural network-based methods. We then propose analytic and neural solutions to two challenging cases: direct lighting with many area lights and efficient rendering of glinty appearances on specular normal-mapped surfaces. Direct lighting from many area light sources is challenging due to variance from both choosing an important light and then a point on it. Existing methods weigh the contribution of all lights by estimating their effect on the shading point. We propose to extend one such method by using analytic methods to improve the estimation of the light’s contribution. This enhancement accelerates the convergence of the algorithm, making it more efficient for scenes with many dynamic lights. The second case deals with the challenge of rendering glinty appearances on normal mapped specular surfaces efficiently. Traditional Monte Carlo methods struggle with this task due to the rapidly changing spatial characteristics of microstructures. Our solution introduces a novel method supporting spatially varying roughness based on a neural histogram, offering both memory and compute efficiency. Additionally, full direct illumination integration is computed analytically for all light directions with minimal computational effort, resulting in improved quality compared to previous approaches. Through comprehensive analysis and experimentation, this thesis contributes to the advancement of rendering techniques, shedding light on the trade-offs between different methods and providing insights into their practical applications for achieving photorealistic rendering

Abstract

Artificial intelligence (AI) has infiltrated all fields of science, from high-energy particle physics to biology to computational chemistry. In the last couple of decades, there has been tremendous advancement in machine learning (ML) applications in computational chemistry. Deep learning (DL) has achieved some success in the automation of feature design, physicochemical property prediction, accelerated chemical space search, and the design of new drug-like molecules. Much work is still needed in terms of property prediction of inorganic molecules, along with the search and design of new molecules and material design with desired properties. This research aims to develop machine learning methods for 3D structure generation and molecular geometry optimization. Use of neural network potential (NNPs) can accelerate the process of 3D structure generation and molecular geometry optimization. Various neural network potentials (NNPs) have been reported in the literature to be as fast as force fields and as accurate as DFT. There has been a lack of standard comparative evaluation of these NNPs, which motivated us to do a benchmark study on NNPs. In this benchmark study, we evaluate and compare four NNPs, i.e., ANI, PhysNet, SchNet, and BAND-NN, for their accuracy in energy prediction, transferability to larger molecules, ability to produce accurate PES, and applicability in geometry optimization. In the context of 3D structure generation (Molecules and material design), there are two major components: search algorithm and property predictor. We need a fast and accurate method to predict the energy of the given system to accelerate the search in conformational space. For this, we developed a model known as DART, which predicts the energy of Gallium clusters using a Topological Atomistic Descriptor (TAD). TAD is a very simple and elegant descriptor that tries to encode structural information by dividing the connectivity information using distance cutoffs. We show the DART models ability to predict the energies of Gallium clusters accurately. For the second component, i.e., the search algorithm, we developed an RL-based model, MeGen, to generate 3D low-energy isomers of Gallium clusters, which uses DART as a reward function. Here we showed that MeGen is significantly more efficient than the conventional workflow for generating ground-state geometries as well as low-lying isomers in terms of time and computational resources. Following a similar train of thought, we developed the MolOpt model. This multi-agent RL-based search algorithm can perform molecular geometry optimization (MGO) by searching for the low-energy structure on the potential energy surface. We show that MolOpt trained on ethane and butane can be used to optimize larger alkanes up to octane. We compare our model with other optimizers and show that MolOpt outperforms the MDMin optimizer and performs similarly to the FIRE optimizer. We further developed an improved version of MolOpt known as MolOpt2. We have made algorithmic changes in MolOpt2, and MolOpt2 is trained on a diverse set of molecules. Hence, due to algorithmic changes, our new model (MolOpt2) can perform MGO on molecules containing elements CHNO and having a size of up to nine heavy atoms. Similar to MolOpt, we compare MolOpt2 with other optimizers and show that MolOpt2 outperforms the MolOpt, MDMin optimizer and performs similarly to the FIRE optimizer.