This paper presents a way of doing large-scale audio understanding without traditional state-of-t... more This paper presents a way of doing large-scale audio understanding without traditional state-of-the-art neural architectures. Ever since the introduction of deep learning for understanding audio signals in the past decade, convolutional architectures have been able to achieve state-of-the-art results surpassing traditional hand-crafted features. In the recent past, there has been a similar shift away from traditional convolutional and recurrent neural networks towards purely end-toend Transformer architectures. We, in this work, explore an approach, based on the Bag-of-Words model. Our approach does not have any convolutions, recurrence, attention, transformers, or other approaches such as BERT. We showcase an approach going against the mainstream research at the moment. We utilize micro and macro-level clustered vanilla embeddings and use an MLP head for classification. We only use feed-forward encoderdecoder models to get the bottlenecks of spectral envelops, spectral patches, and slices as well as multi-resolution spectra. A classification head (a feed-forward layer), similar to the approach in SimCLR is trained on a learned representation. Using simple codes learned on them, we show how we surpass traditional convolutional neural network architectures, and come strikingly close to outperforming powerful Transformer architectures. Our approach could have been carried out back in 2006, using the vanilla autoencoder and dimension reduction architectures, as proposed by Hinton et. al. [1]. We show that just by using this architecture, and doing simple statistics on the latent representations, we could have outperformed state of the art architectures audio understanding, like convolutional and recurrent architecture as late as 2018 [2]. This goal of this work is, hopefully to would pave way for exciting advancements in the field of representation learning without massive, end-to-end neural architectures.
Advances in the field of auxetics have realized fabricated auxetic materials such as foams, fabri... more Advances in the field of auxetics have realized fabricated auxetic materials such as foams, fabrics, and fibers as well as a better theoretical understanding of the auxetic response. Because of their unique properties and applications, commodity auxetic materials are particularly desirable. Needle-punched nonwovens, several kinds of paper, and many knitted and woven fabrics have the potential to be auxetic, either as-produced or through a processing or design solution. In this study, we examine the out-of-plane Poisson’s ratio of as-produced and heat-compressed wool nonwovens. The wool nonwovens were found to be out-of-plane auxetic as-produced, and their auxetic character became more pronounced at higher treatment temperatures. Their behavior could be similar to that of paper, where straightening of a bent fiber was responsible for thickness increase. The prescribed processing conditions to enhance auxeticity could potentially be incorporated in their existing production, providing...
2021 24th International Conference on Digital Audio Effects (DAFx)
This paper proposes a novel way of doing audio synthesis at the waveform level using Transformer ... more This paper proposes a novel way of doing audio synthesis at the waveform level using Transformer architectures. We propose a deep neural network for generating waveforms, similar to wavenet [1]. This is fully probabilistic, auto-regressive, and causal, i.e. each sample generated depends on only the previously observed samples. Our approach outperforms a widely used wavenet architecture by up to 9% on a similar dataset for predicting the next step. Using the attention mechanism, we enable the architecture to learn which audio samples are important for the prediction of the future sample. We show how causal transformer generative models can be used for raw waveform synthesis. We also show that this performance can be improved by another 2% by conditioning samples over a wider context. The flexibility of the current model to synthesize audio from latent representations suggests a large number of potential applications. The novel approach of using generative transformer architectures for raw audio synthesis is, however, still far away from generating any meaningful music similar to wavenet, without using latent codes/meta-data to aid the generation process.
variability between clinicians' approach to the CRC screening discussion, we asked them to use th... more variability between clinicians' approach to the CRC screening discussion, we asked them to use the information sheet with relevant test performance data and a suggested guided conversation that covered salient points about the importance of screening, differences between options, and implications of each choice. Clinic staff flagged eligible
Societies rely on individual contributions to sustain public goods that benefit the entire commun... more Societies rely on individual contributions to sustain public goods that benefit the entire community. Several mechanisms, that specify how individuals change their decisions based on past experiences, have been proposed to explain how altruists are not outcompeted by selfish counterparts. A key aspect of such strategy updates involves a comparison of an individual's latest payoff with that of a random neighbour. In reality, both the economic and social milieu often shapes cooperative behaviour. We propose a new decision heuristic, where the propensity of an individual to cooperate depends on the local strategy environment in which she is embedded as well as her wealth relative to that of her neighbours. Our decision-making model allows cooperation to be sustained and also explains the results of recent experiments on social dilemmas in dynamic networks. Final cooperation levels depend only on the extent to which the strategy environment influences altruistic behaviour but are la...
Five transverse rod (TR)-containing main-chain liquid crystalline polymers (LCPs) are examined in... more Five transverse rod (TR)-containing main-chain liquid crystalline polymers (LCPs) are examined in both the powder and fiber forms by wide angle X-ray diffraction. Data from the diffraction experiments are consistent with the site-connectivitydriven TR-reorientation mechanism for intrinsic auxetic character in these macromolecules. Shifts in peak maxima, intensity distributions in the interchain interaction region, and calculated volume increase accompanying fiber formation are all in line with expectations from considerations of this mechanism. The narrow window for experimental conditions necessary for fiber preparation and subsequent X-ray observation of this phenomenon is detailed. A close examination of space-filling Corey-Pauling-Koltun (CPK) molecular models reveals the details of local packing of rods and connected polymer chains in the quiescent nematic polymer melt prior to fiber drawing. It is found that two rod-reorientation possibilities exist each of which leads to increased interchain separation as required for auxetic response. Suggestions for future experiments and materials design for new LCPs are described.
Human-robot interactions are less efficient and communicative than human-to-human interactions, a... more Human-robot interactions are less efficient and communicative than human-to-human interactions, and a key reason is a lack of informed sense of touch in robotic systems. Existing literature demonstrates robot success in executing handovers with humans, albeit with substantial reliance on external sensing or with primitive signal processing methods, deficient compared to the rich set of information humans can detect. In contrast, we present models capable of distinguishing between four classes of human tactile gestures at a robot's end effector, using only a non-collocated six-axis force sensor at the wrist. Due to the absence in the literature, this work describes 1) the collection of an extensive force dataset characterized by human-robot contact events, and 2) classification models informed by this dataset to determine the nature of the interaction. We demonstrate high classification accuracies among our proposed gesture definitions on a test set, emphasizing that neural netwo...
In music and speech, meaning is derived at multiple levels of context. Affect, for example, can b... more In music and speech, meaning is derived at multiple levels of context. Affect, for example, can be inferred both by a short sound token and by sonic patterns over a longer temporal window such as an entire recording. In this paper we focus on inferring meaning from this dichotomy of contexts. We show how contextual representations of short sung vocal lines can be implicitly learned from fundamental frequency ($F_0$) and thus be used as a meaningful feature space for downstream Music Information Retrieval (MIR) tasks. We propose three self-supervised deep learning paradigms which leverage pseudotask learning of these two levels of context to produce latent representation spaces. We evaluate the usefulness of these representations by embedding unseen vocal contours into each space and conducting downstream classification tasks. Our results show that contextual representation can enhance downstream classification by as much as 15 % as compared to using traditional statistical contour f...
We propose a system that learns from artistic pairings of music and corresponding album cover art... more We propose a system that learns from artistic pairings of music and corresponding album cover art. The goal is to 'translate' paintings into music and, in further stages of development, the converse. We aim to deploy this system as an artistic tool for real time 'translations' between musicians and painters. The system's outputs serve as elements to be employed in a joint live performance of music and painting, or as generative material to be used by the artists as inspiration for their improvisation.
their continuous and prompt support, guidance and encouragement all along my Doctoral research. I... more their continuous and prompt support, guidance and encouragement all along my Doctoral research. I learned from them how to think scientifically and critically. Their trust and suggestions helped me to improve and expand my problem solving skills. I hope that I could be half as smart and patient as them one day. I would like to thank my committee members-Dr. Singh, Dr. Wilkinson, Dr. Thadhani and Dr. Bucknall for their continual guidance and immensely useful insights that gave direction and helped shape this research. At the outset, it is my duty to acknowledge with gratitude the consummate facilities and generous help that I have received from the School of Material Science and Engineering at Georgia Tech. I am indebted to the fellowship award from Renewable Bioproducts Institute at Georgia Tech that supported my PhD work. I would like to thank Dr. Kamath for helping me understand nonwovens better, Dr. Realff for letting me use her microscope, Dr. Guldberg for letting me use his micro-CT facility, Dr. Parachuru for frequent help with Instron and with understanding textiles, and last but not the least Dr. Marsolan for letting me use the facilities at RBI, Georgia Tech, for valuable discussions, and for letting me be a part of the wonderful paper science community. This research could not have been accomplished without the splendid support, input and cooperation from Angela Lin, who helped me in every phase of the complex micro-CT scanning work and Matthew Priddy, for helping me with Abaqus and building and running finite element analysis on my networks. My special thanks goes out to Dr. v Rallming Yang, Dr. Dongho Kim and my friends and colleagues, Sudhir Sharma and Sandeep Mora, for providing me with materials and helping me mastering the art of papermaking. This thesis could not have been completed without the invaluable research work done by my dear undergraduate students and researchers-C. J. Layer, Tony Shu and Karla Wagner. I would like to thank them for their time, their dedication, their input, and the interest they showed in this research. I would also like to thank Yolande Berta (MSE, Georgia Tech) and Brooke Barta (GTRI, Georgia Tech) for their help with SEM and the entire MSE staff at Georgia Tech-Hope, Susan, Angie, Teresa, Rusty, Rod, Linda, Sherry, Jamar and Jasmine for their beautiful smiles and willingness to help in all situations. I acknowledge my deep gratitude to Mike Allen and Keith Hubbard of TenCate Protective Fabrics, for their valuable time, for helpful and insightful discussions and for providing us with the nonwoven samples. It was also my immense pleasure to participate
with identity fraud in our society reaching unprecedented proportions and with an increasing emph... more with identity fraud in our society reaching unprecedented proportions and with an increasing emphasis on the emerging automatic personal identification applications, biometrics-based verification, especially fingerprint-based identification, is receiving a lot of attention Biometrics deals with identifying individuals with help of their biological data. Fingerprint scanning is the most common method of the biometric methods available today. The security of fingerprint scanners has however been questioned and previous studies have shown that fingerprint scanners can be fooled with artificial fingerprints, i.e. copies of real fingerprints. The fingerprint recognition systems are evolving and this paper will discuss the situation of today. We match the finger prints, one that is already in the database of the sensor and second the fingerprint that we enrolled in the sensor currently by using the Boolean function XORING. We get the matching score and decide the result on the matching sc...
ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2019
We propose spoken sentence embeddings which capture both acoustic and linguistic content. While e... more We propose spoken sentence embeddings which capture both acoustic and linguistic content. While existing works operate at the character, phoneme, or word level, our method learns long-term dependencies by modeling speech at the sentence level. Formulated as an audio-linguistic multitask learning problem, our encoder-decoder model simultaneously reconstructs acoustic and natural language features from audio. Our results show that spoken sentence embeddings outperform phoneme and word-level baselines on speech recognition and emotion recognition tasks. Ablation studies show that our embeddings can better model high-level acoustic concepts while retaining linguistic content. Overall, our work illustrates the viability of generic, multi-modal sentence embeddings for spoken language understanding.
Gene drive technology is being presented as a means to deliver on some of the global challenges h... more Gene drive technology is being presented as a means to deliver on some of the global challenges humanity faces today in healthcare, agriculture and conservation. However, there is a limited understanding of the consequences of releasing self-perpetuating transgenic organisms into the wild populations under complex ecological conditions. In this study, we analyze the impact of three factors, mate-choice, mating systems and spatial mating network, on the population dynamics for two distinct classes of modification gene drive systems; distortion and viability-based ones. All three factors had a high impact on the modelling outcome. First, we demonstrate that distortion based gene drives appear to be more robust against the mate-choice than viability-based gene drives. Second, we find that gene drive spread is much faster for higher degrees of polygamy. With fitness cost, speed is the highest for intermediate levels of polygamy. Finally, the spread of gene drive is faster and more effec...
Societies rely on individual contributions to sustain public goods that benefit the entire commun... more Societies rely on individual contributions to sustain public goods that benefit the entire community. Several mechanisms, that specify how individuals change their decisions based on past experiences, have been proposed to explain how altruists are not outcompeted by selfish counterparts. A key aspect of such strategy updates involves a comparison of an individual’s latest payoff with that of a random neighbour. In reality, both the economic and social milieu often shapes cooperative behaviour. We propose a new decision heuristic, where the propensity of an individual to cooperate depends on the local strategy environment in which she is embedded as well as her wealth relative to that of her neighbours. Our decision-making model allows cooperation to be sustained and also explains the results of recent experiments on social dilemmas in dynamic networks. Final cooperation levels depend only on the extent to which the strategy environment influences altruistic behaviour but are largel...
Background While schizophrenia is observed in different parts of the world across countries, ethn... more Background While schizophrenia is observed in different parts of the world across countries, ethnicities, and races, research indicates cultural factors play significant roles in the phenomenology of this illness. Cultural norms and values affect manifestations of this pathology; more specifically, they affect how symptoms are expressed, experienced, and interpreted. Given that culture affects manifestations of schizophrenia, cultural factors should be considered in the assessment of its symptoms in clinical trials. This study explores the differences and patterns in the Positive and Negative Syndrome Scale (PANSS) item ratings across different geocultural regions. Identifying such patterns can give insights into culturally sensitive assessment practices and aid in developing more effective rater training and data surveillance that consider unique cultural factors. Methods Data were obtained from an international group of raters from 37 different countries, representing 6 geocultura...
Background: The exponential growth of COVID-19 cases and testing has created supply shortages at ... more Background: The exponential growth of COVID-19 cases and testing has created supply shortages at various points in the testing workflow. As of April 15, 2020 FDA recommendations only allowed for the use of nasopharyngeal, flocked mid turbinate, or foam nasal swabs, all of which are in very low supply. Polyester swabs are more readily available and mass producible. We compare the performance of polyester and foam swabs stored in different transport media. Methods: Both polyester and foam nasal swabs were collected from convalescent COVID-19 patients at a single visit. Using the foam nasal swabs as the comparator, sensitivity of the polyester swabs in each media were calculated, three by three tables were constructed to measure concordance, and cycle threshold (Ct) values were compared. Findings: 126 visits had polyester and foam swabs stored in viral transport media (VTM), 51 had polyester and foam swabs stored in saline, and 63 had a foam swab in VTM and a polyester swab stored in a...
ABSTRACTBackgroundCurrent testing for SARS-CoV-2 requires health care workers to collect a nasoph... more ABSTRACTBackgroundCurrent testing for SARS-CoV-2 requires health care workers to collect a nasopharyngeal (NP) sample from a patient. NP sampling requires the use of personal protective equipment that are in limited supply, is uncomfortable for the patient, and reduces clinical efficiency. This study explored the equivalency of patient-collected tongue, anterior nares (nasal), and mid-turbinate (MT) samples to health care worker-collected NP samples for detecting SARS-CoV-2.MethodsPatients presenting to five urgent care facilities with symptoms indicative of an upper respiratory infection provided self-collected samples from three anatomic sites along with a health care worker-collected NP sample. Using NP as the comparator, sensitivities and one-sided 95% confidence intervals for the tongue, nasal, and MT samples for detection of SARS-CoV-2 were calculated.ResultsThe sensitivity for detecting SARS-CoV-2 in patient-collected tongue, nasal, and mid-turbinate samples was 89.8% (95% CI...
Synthetic gene drive technologies aim to spread transgenic constructs into wild populations even ... more Synthetic gene drive technologies aim to spread transgenic constructs into wild populations even when they impose organismal fitness disadvantages. The extraordinary diversity of plausible drive mechanisms and the range of selective parameters they may encounter makes it very difficult to convey their relative predicted properties, particularly where multiple approaches are combined. The sheer number of published manuscripts in this field, experimental and theoretical, the numerous techniques resulting in an explosion in the gene drive vocabulary hinder the regulators’ point of view. We address this concern by defining a simplified parameter based language of synthetic drives. Employing the classical population dynamics approach, we show that different drive construct (replacement) mechanisms can be condensed and evaluated on an equal footing even where they incorporate multiple replacement drives approaches. Using a common language, it is then possible to compare various model prop...
We consider competition between antibiotic producing bacteria, non-producers (or cheaters), and s... more We consider competition between antibiotic producing bacteria, non-producers (or cheaters), and sensitive cells in a two-dimensional lattice model. Previous work has shown that these three cell types can survive in spatial models due to the presence of spatial patterns, whereas coexistence is not possible in a well-mixed system. We extend this to consider the evolution of the antibiotic production rate, assuming that the cost of antibiotic production leads to a reduction in growth rate of the producers. We find that coexistence occurs for an intermediate range of antibiotic production rate. If production rate is too high or too low, only sensitive cells survive. When evolution of production rate is allowed, a mixture of cell types arises in which there is a dominant producer strain that produces sufficient to limit the growth of sensitive cells and which is able to withstand the presence of cheaters in its own species. The mixture includes a range of low-rate producers and non-producers, none of which could survive without the presence of the dominant producer strain. We also consider the case of evolution of antibiotic resistance within the sensitive species. In order for the resistant cells to survive, they must grow faster than both the non-producers and the producers. However, if the resistant cells grow too rapidly, the producing species is eliminated, after which the resistance mutation is no longer useful, and sensitive cells take over the system. We show that there is a range of growth rates of the resistant cells where the two species coexist, and where the production mechanism is maintained as a polymorphism in the producing species and the resistance mechanism is maintained as a polymorphism in the sensitive species.
This paper presents a way of doing large-scale audio understanding without traditional state-of-t... more This paper presents a way of doing large-scale audio understanding without traditional state-of-the-art neural architectures. Ever since the introduction of deep learning for understanding audio signals in the past decade, convolutional architectures have been able to achieve state-of-the-art results surpassing traditional hand-crafted features. In the recent past, there has been a similar shift away from traditional convolutional and recurrent neural networks towards purely end-toend Transformer architectures. We, in this work, explore an approach, based on the Bag-of-Words model. Our approach does not have any convolutions, recurrence, attention, transformers, or other approaches such as BERT. We showcase an approach going against the mainstream research at the moment. We utilize micro and macro-level clustered vanilla embeddings and use an MLP head for classification. We only use feed-forward encoderdecoder models to get the bottlenecks of spectral envelops, spectral patches, and slices as well as multi-resolution spectra. A classification head (a feed-forward layer), similar to the approach in SimCLR is trained on a learned representation. Using simple codes learned on them, we show how we surpass traditional convolutional neural network architectures, and come strikingly close to outperforming powerful Transformer architectures. Our approach could have been carried out back in 2006, using the vanilla autoencoder and dimension reduction architectures, as proposed by Hinton et. al. [1]. We show that just by using this architecture, and doing simple statistics on the latent representations, we could have outperformed state of the art architectures audio understanding, like convolutional and recurrent architecture as late as 2018 [2]. This goal of this work is, hopefully to would pave way for exciting advancements in the field of representation learning without massive, end-to-end neural architectures.
Advances in the field of auxetics have realized fabricated auxetic materials such as foams, fabri... more Advances in the field of auxetics have realized fabricated auxetic materials such as foams, fabrics, and fibers as well as a better theoretical understanding of the auxetic response. Because of their unique properties and applications, commodity auxetic materials are particularly desirable. Needle-punched nonwovens, several kinds of paper, and many knitted and woven fabrics have the potential to be auxetic, either as-produced or through a processing or design solution. In this study, we examine the out-of-plane Poisson’s ratio of as-produced and heat-compressed wool nonwovens. The wool nonwovens were found to be out-of-plane auxetic as-produced, and their auxetic character became more pronounced at higher treatment temperatures. Their behavior could be similar to that of paper, where straightening of a bent fiber was responsible for thickness increase. The prescribed processing conditions to enhance auxeticity could potentially be incorporated in their existing production, providing...
2021 24th International Conference on Digital Audio Effects (DAFx)
This paper proposes a novel way of doing audio synthesis at the waveform level using Transformer ... more This paper proposes a novel way of doing audio synthesis at the waveform level using Transformer architectures. We propose a deep neural network for generating waveforms, similar to wavenet [1]. This is fully probabilistic, auto-regressive, and causal, i.e. each sample generated depends on only the previously observed samples. Our approach outperforms a widely used wavenet architecture by up to 9% on a similar dataset for predicting the next step. Using the attention mechanism, we enable the architecture to learn which audio samples are important for the prediction of the future sample. We show how causal transformer generative models can be used for raw waveform synthesis. We also show that this performance can be improved by another 2% by conditioning samples over a wider context. The flexibility of the current model to synthesize audio from latent representations suggests a large number of potential applications. The novel approach of using generative transformer architectures for raw audio synthesis is, however, still far away from generating any meaningful music similar to wavenet, without using latent codes/meta-data to aid the generation process.
variability between clinicians' approach to the CRC screening discussion, we asked them to use th... more variability between clinicians' approach to the CRC screening discussion, we asked them to use the information sheet with relevant test performance data and a suggested guided conversation that covered salient points about the importance of screening, differences between options, and implications of each choice. Clinic staff flagged eligible
Societies rely on individual contributions to sustain public goods that benefit the entire commun... more Societies rely on individual contributions to sustain public goods that benefit the entire community. Several mechanisms, that specify how individuals change their decisions based on past experiences, have been proposed to explain how altruists are not outcompeted by selfish counterparts. A key aspect of such strategy updates involves a comparison of an individual's latest payoff with that of a random neighbour. In reality, both the economic and social milieu often shapes cooperative behaviour. We propose a new decision heuristic, where the propensity of an individual to cooperate depends on the local strategy environment in which she is embedded as well as her wealth relative to that of her neighbours. Our decision-making model allows cooperation to be sustained and also explains the results of recent experiments on social dilemmas in dynamic networks. Final cooperation levels depend only on the extent to which the strategy environment influences altruistic behaviour but are la...
Five transverse rod (TR)-containing main-chain liquid crystalline polymers (LCPs) are examined in... more Five transverse rod (TR)-containing main-chain liquid crystalline polymers (LCPs) are examined in both the powder and fiber forms by wide angle X-ray diffraction. Data from the diffraction experiments are consistent with the site-connectivitydriven TR-reorientation mechanism for intrinsic auxetic character in these macromolecules. Shifts in peak maxima, intensity distributions in the interchain interaction region, and calculated volume increase accompanying fiber formation are all in line with expectations from considerations of this mechanism. The narrow window for experimental conditions necessary for fiber preparation and subsequent X-ray observation of this phenomenon is detailed. A close examination of space-filling Corey-Pauling-Koltun (CPK) molecular models reveals the details of local packing of rods and connected polymer chains in the quiescent nematic polymer melt prior to fiber drawing. It is found that two rod-reorientation possibilities exist each of which leads to increased interchain separation as required for auxetic response. Suggestions for future experiments and materials design for new LCPs are described.
Human-robot interactions are less efficient and communicative than human-to-human interactions, a... more Human-robot interactions are less efficient and communicative than human-to-human interactions, and a key reason is a lack of informed sense of touch in robotic systems. Existing literature demonstrates robot success in executing handovers with humans, albeit with substantial reliance on external sensing or with primitive signal processing methods, deficient compared to the rich set of information humans can detect. In contrast, we present models capable of distinguishing between four classes of human tactile gestures at a robot's end effector, using only a non-collocated six-axis force sensor at the wrist. Due to the absence in the literature, this work describes 1) the collection of an extensive force dataset characterized by human-robot contact events, and 2) classification models informed by this dataset to determine the nature of the interaction. We demonstrate high classification accuracies among our proposed gesture definitions on a test set, emphasizing that neural netwo...
In music and speech, meaning is derived at multiple levels of context. Affect, for example, can b... more In music and speech, meaning is derived at multiple levels of context. Affect, for example, can be inferred both by a short sound token and by sonic patterns over a longer temporal window such as an entire recording. In this paper we focus on inferring meaning from this dichotomy of contexts. We show how contextual representations of short sung vocal lines can be implicitly learned from fundamental frequency ($F_0$) and thus be used as a meaningful feature space for downstream Music Information Retrieval (MIR) tasks. We propose three self-supervised deep learning paradigms which leverage pseudotask learning of these two levels of context to produce latent representation spaces. We evaluate the usefulness of these representations by embedding unseen vocal contours into each space and conducting downstream classification tasks. Our results show that contextual representation can enhance downstream classification by as much as 15 % as compared to using traditional statistical contour f...
We propose a system that learns from artistic pairings of music and corresponding album cover art... more We propose a system that learns from artistic pairings of music and corresponding album cover art. The goal is to 'translate' paintings into music and, in further stages of development, the converse. We aim to deploy this system as an artistic tool for real time 'translations' between musicians and painters. The system's outputs serve as elements to be employed in a joint live performance of music and painting, or as generative material to be used by the artists as inspiration for their improvisation.
their continuous and prompt support, guidance and encouragement all along my Doctoral research. I... more their continuous and prompt support, guidance and encouragement all along my Doctoral research. I learned from them how to think scientifically and critically. Their trust and suggestions helped me to improve and expand my problem solving skills. I hope that I could be half as smart and patient as them one day. I would like to thank my committee members-Dr. Singh, Dr. Wilkinson, Dr. Thadhani and Dr. Bucknall for their continual guidance and immensely useful insights that gave direction and helped shape this research. At the outset, it is my duty to acknowledge with gratitude the consummate facilities and generous help that I have received from the School of Material Science and Engineering at Georgia Tech. I am indebted to the fellowship award from Renewable Bioproducts Institute at Georgia Tech that supported my PhD work. I would like to thank Dr. Kamath for helping me understand nonwovens better, Dr. Realff for letting me use her microscope, Dr. Guldberg for letting me use his micro-CT facility, Dr. Parachuru for frequent help with Instron and with understanding textiles, and last but not the least Dr. Marsolan for letting me use the facilities at RBI, Georgia Tech, for valuable discussions, and for letting me be a part of the wonderful paper science community. This research could not have been accomplished without the splendid support, input and cooperation from Angela Lin, who helped me in every phase of the complex micro-CT scanning work and Matthew Priddy, for helping me with Abaqus and building and running finite element analysis on my networks. My special thanks goes out to Dr. v Rallming Yang, Dr. Dongho Kim and my friends and colleagues, Sudhir Sharma and Sandeep Mora, for providing me with materials and helping me mastering the art of papermaking. This thesis could not have been completed without the invaluable research work done by my dear undergraduate students and researchers-C. J. Layer, Tony Shu and Karla Wagner. I would like to thank them for their time, their dedication, their input, and the interest they showed in this research. I would also like to thank Yolande Berta (MSE, Georgia Tech) and Brooke Barta (GTRI, Georgia Tech) for their help with SEM and the entire MSE staff at Georgia Tech-Hope, Susan, Angie, Teresa, Rusty, Rod, Linda, Sherry, Jamar and Jasmine for their beautiful smiles and willingness to help in all situations. I acknowledge my deep gratitude to Mike Allen and Keith Hubbard of TenCate Protective Fabrics, for their valuable time, for helpful and insightful discussions and for providing us with the nonwoven samples. It was also my immense pleasure to participate
with identity fraud in our society reaching unprecedented proportions and with an increasing emph... more with identity fraud in our society reaching unprecedented proportions and with an increasing emphasis on the emerging automatic personal identification applications, biometrics-based verification, especially fingerprint-based identification, is receiving a lot of attention Biometrics deals with identifying individuals with help of their biological data. Fingerprint scanning is the most common method of the biometric methods available today. The security of fingerprint scanners has however been questioned and previous studies have shown that fingerprint scanners can be fooled with artificial fingerprints, i.e. copies of real fingerprints. The fingerprint recognition systems are evolving and this paper will discuss the situation of today. We match the finger prints, one that is already in the database of the sensor and second the fingerprint that we enrolled in the sensor currently by using the Boolean function XORING. We get the matching score and decide the result on the matching sc...
ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2019
We propose spoken sentence embeddings which capture both acoustic and linguistic content. While e... more We propose spoken sentence embeddings which capture both acoustic and linguistic content. While existing works operate at the character, phoneme, or word level, our method learns long-term dependencies by modeling speech at the sentence level. Formulated as an audio-linguistic multitask learning problem, our encoder-decoder model simultaneously reconstructs acoustic and natural language features from audio. Our results show that spoken sentence embeddings outperform phoneme and word-level baselines on speech recognition and emotion recognition tasks. Ablation studies show that our embeddings can better model high-level acoustic concepts while retaining linguistic content. Overall, our work illustrates the viability of generic, multi-modal sentence embeddings for spoken language understanding.
Gene drive technology is being presented as a means to deliver on some of the global challenges h... more Gene drive technology is being presented as a means to deliver on some of the global challenges humanity faces today in healthcare, agriculture and conservation. However, there is a limited understanding of the consequences of releasing self-perpetuating transgenic organisms into the wild populations under complex ecological conditions. In this study, we analyze the impact of three factors, mate-choice, mating systems and spatial mating network, on the population dynamics for two distinct classes of modification gene drive systems; distortion and viability-based ones. All three factors had a high impact on the modelling outcome. First, we demonstrate that distortion based gene drives appear to be more robust against the mate-choice than viability-based gene drives. Second, we find that gene drive spread is much faster for higher degrees of polygamy. With fitness cost, speed is the highest for intermediate levels of polygamy. Finally, the spread of gene drive is faster and more effec...
Societies rely on individual contributions to sustain public goods that benefit the entire commun... more Societies rely on individual contributions to sustain public goods that benefit the entire community. Several mechanisms, that specify how individuals change their decisions based on past experiences, have been proposed to explain how altruists are not outcompeted by selfish counterparts. A key aspect of such strategy updates involves a comparison of an individual’s latest payoff with that of a random neighbour. In reality, both the economic and social milieu often shapes cooperative behaviour. We propose a new decision heuristic, where the propensity of an individual to cooperate depends on the local strategy environment in which she is embedded as well as her wealth relative to that of her neighbours. Our decision-making model allows cooperation to be sustained and also explains the results of recent experiments on social dilemmas in dynamic networks. Final cooperation levels depend only on the extent to which the strategy environment influences altruistic behaviour but are largel...
Background While schizophrenia is observed in different parts of the world across countries, ethn... more Background While schizophrenia is observed in different parts of the world across countries, ethnicities, and races, research indicates cultural factors play significant roles in the phenomenology of this illness. Cultural norms and values affect manifestations of this pathology; more specifically, they affect how symptoms are expressed, experienced, and interpreted. Given that culture affects manifestations of schizophrenia, cultural factors should be considered in the assessment of its symptoms in clinical trials. This study explores the differences and patterns in the Positive and Negative Syndrome Scale (PANSS) item ratings across different geocultural regions. Identifying such patterns can give insights into culturally sensitive assessment practices and aid in developing more effective rater training and data surveillance that consider unique cultural factors. Methods Data were obtained from an international group of raters from 37 different countries, representing 6 geocultura...
Background: The exponential growth of COVID-19 cases and testing has created supply shortages at ... more Background: The exponential growth of COVID-19 cases and testing has created supply shortages at various points in the testing workflow. As of April 15, 2020 FDA recommendations only allowed for the use of nasopharyngeal, flocked mid turbinate, or foam nasal swabs, all of which are in very low supply. Polyester swabs are more readily available and mass producible. We compare the performance of polyester and foam swabs stored in different transport media. Methods: Both polyester and foam nasal swabs were collected from convalescent COVID-19 patients at a single visit. Using the foam nasal swabs as the comparator, sensitivity of the polyester swabs in each media were calculated, three by three tables were constructed to measure concordance, and cycle threshold (Ct) values were compared. Findings: 126 visits had polyester and foam swabs stored in viral transport media (VTM), 51 had polyester and foam swabs stored in saline, and 63 had a foam swab in VTM and a polyester swab stored in a...
ABSTRACTBackgroundCurrent testing for SARS-CoV-2 requires health care workers to collect a nasoph... more ABSTRACTBackgroundCurrent testing for SARS-CoV-2 requires health care workers to collect a nasopharyngeal (NP) sample from a patient. NP sampling requires the use of personal protective equipment that are in limited supply, is uncomfortable for the patient, and reduces clinical efficiency. This study explored the equivalency of patient-collected tongue, anterior nares (nasal), and mid-turbinate (MT) samples to health care worker-collected NP samples for detecting SARS-CoV-2.MethodsPatients presenting to five urgent care facilities with symptoms indicative of an upper respiratory infection provided self-collected samples from three anatomic sites along with a health care worker-collected NP sample. Using NP as the comparator, sensitivities and one-sided 95% confidence intervals for the tongue, nasal, and MT samples for detection of SARS-CoV-2 were calculated.ResultsThe sensitivity for detecting SARS-CoV-2 in patient-collected tongue, nasal, and mid-turbinate samples was 89.8% (95% CI...
Synthetic gene drive technologies aim to spread transgenic constructs into wild populations even ... more Synthetic gene drive technologies aim to spread transgenic constructs into wild populations even when they impose organismal fitness disadvantages. The extraordinary diversity of plausible drive mechanisms and the range of selective parameters they may encounter makes it very difficult to convey their relative predicted properties, particularly where multiple approaches are combined. The sheer number of published manuscripts in this field, experimental and theoretical, the numerous techniques resulting in an explosion in the gene drive vocabulary hinder the regulators’ point of view. We address this concern by defining a simplified parameter based language of synthetic drives. Employing the classical population dynamics approach, we show that different drive construct (replacement) mechanisms can be condensed and evaluated on an equal footing even where they incorporate multiple replacement drives approaches. Using a common language, it is then possible to compare various model prop...
We consider competition between antibiotic producing bacteria, non-producers (or cheaters), and s... more We consider competition between antibiotic producing bacteria, non-producers (or cheaters), and sensitive cells in a two-dimensional lattice model. Previous work has shown that these three cell types can survive in spatial models due to the presence of spatial patterns, whereas coexistence is not possible in a well-mixed system. We extend this to consider the evolution of the antibiotic production rate, assuming that the cost of antibiotic production leads to a reduction in growth rate of the producers. We find that coexistence occurs for an intermediate range of antibiotic production rate. If production rate is too high or too low, only sensitive cells survive. When evolution of production rate is allowed, a mixture of cell types arises in which there is a dominant producer strain that produces sufficient to limit the growth of sensitive cells and which is able to withstand the presence of cheaters in its own species. The mixture includes a range of low-rate producers and non-producers, none of which could survive without the presence of the dominant producer strain. We also consider the case of evolution of antibiotic resistance within the sensitive species. In order for the resistant cells to survive, they must grow faster than both the non-producers and the producers. However, if the resistant cells grow too rapidly, the producing species is eliminated, after which the resistance mutation is no longer useful, and sensitive cells take over the system. We show that there is a range of growth rates of the resistant cells where the two species coexist, and where the production mechanism is maintained as a polymorphism in the producing species and the resistance mechanism is maintained as a polymorphism in the sensitive species.
Uploads
Papers by Prateek Verma