This study proposed a novel framework for COVID-19 severity prediction, which is a combination of data-centric and model-centric approaches. First, we propose a data-centric pre-training for extremely scare data scenarios of the investigating dataset
. Second, we propose two hybrid convolution-attention neural architectures that leverage the self-attention from the Transformer and the Dense Associative Memory (Modern Hopfield networks). Our proposed approach achieves significant improvement from the conventional baseline approach. The best model from our proposed approach achieves $R^2 = 0.85 pm 0.05$ and Pearson correlation coefficient $rho = 0.92 pm 0.02$ in geographic extend and $R^2 = 0.72 pm 0.09, rho = 0.85pm 0.06$ in opacity prediction.
Despite their consequential applications, metastable states of antibranes in warped throats are not yet fully understood. In this thesis, we provide new information on various aspects of these metastable antibranes through applications of the blackfo
ld effective theory for higher-dimensional black holes. As concrete examples, we study the conjectured metastable state of polarised anti-D3 branes at the tip of the Klebanov-Strassler (KS) throat in type IIB supergravity and the analogous state of polarised anti-M2 branes at the tip of the Cvetic-Gibbons-Lu-Pope (CGLP) throat in eleven-dimensional supergravity. For anti-D3 branes in KS throat, we provide novel evidence for the existence of the metastable state exactly where no-go theorems are lifted. In the extremal limit, we recover directly in supergravity the metastable states originally discovered by Kachru, Pearson, and Verlinde (KPV). Away from extremality, we uncover a metastable wrapped black NS5 state. We observe that such metastability is lost when the wrapped NS5 is heated sufficiently that its horizon geometry resembles that of a black anti-D3. We study the classical stability of the KPV state under generic long-wavelength deformations. We observe that, with regards to considered perturbations and regime of parameters, the state is classically stable. A study of anti-M2 branes in CGLP throat reveals many similarities to that of the anti-D3 branes. We recover directly in supergravity the Klebanov-Pufu (KP) state at extremality, and our finite temperature results fit suggestively well with known, complementary no-go theorems. However, we discover an unexpected, exotic pattern of thermal transitions of the KP state different from that of the KPV. This thesis contains also a pedagogical introduction to the blackfold formalism, focusing on aspects immediately relevant to applications to metastable antibranes.
End-to-end multilingual speech recognition involves using a single model training on a compositional speech corpus including many languages, resulting in a single neural network to handle transcribing different languages. Due to the fact that each la
nguage in the training data has different characteristics, the shared network may struggle to optimize for all various languages simultaneously. In this paper we propose a novel multilingual architecture that targets the core operation in neural networks: linear transformation functions. The key idea of the method is to assign fast weight matrices for each language by decomposing each weight matrix into a shared component and a language dependent component. The latter is then factorized into vectors using rank-1 assumptions to reduce the number of parameters per language. This efficient factorization scheme is proved to be effective in two multilingual settings with $7$ and $27$ languages, reducing the word error rates by $26%$ and $27%$ rel. for two popular architectures LSTM and Transformer, respectively.
Understanding course enrollment patterns is valuable to predict upcoming demands for future courses, and to provide student with realistic courses to pursue given their current backgrounds. This study uses undergraduate student enrollment data to for
m networks of courses where connections are based on student co-enrollments. The course networks generated in this paper are based on eight years of undergraduate course enrollment data from a large metropolitan university. The networks are analyzed to identify hub courses often taken with many other courses. Two notions of hubs are considered: one focused on raw popularity across all students, and one focused on proportional likelihoods of co-enrollment with other courses. A variety of network metrics are calculated to evaluate the course networks. Academic departments and high-level academic categories, such as Humanities vs STEM, are studied for their influence over course groupings. The identification of hub courses has practical applications, since it can help better predict the impact of changes in course offerings and in course popularity, and in the case of interdisciplinary hub courses, can be used to increase or decrease interest and enrollments in specific academic departments and areas.
This paper proposes a novel cell-based neural architecture search algorithm (NAS), which completely alleviates the expensive costs of data labeling inherited from supervised learning. Our algorithm capitalizes on the effectiveness of self-supervised
learning for image representations, which is an increasingly crucial topic of computer vision. First, using only a small amount of unlabeled train data under contrastive self-supervised learning allow us to search on a more extensive search space, discovering better neural architectures without surging the computational resources. Second, we entirely relieve the cost for labeled data (by contrastive loss) in the search stage without compromising architectures final performance in the evaluation phase. Finally, we tackle the inherent discrete search space of the NAS problem by sequential model-based optimization via the tree-parzen estimator (SMBO-TPE), enabling us to reduce the computational expense response surface significantly. An extensive number of experiments empirically show that our search algorithm can achieve state-of-the-art results with better efficiency in data labeling cost, searching time, and accuracy in final validation.
Transformer models are powerful sequence-to-sequence architectures that are capable of directly mapping speech inputs to transcriptions or translations. However, the mechanism for modeling positions in this model was tailored for text modeling, and t
hus is less ideal for acoustic inputs. In this work, we adapt the relative position encoding scheme to the Speech Transformer, where the key addition is relative distance between input states in the self-attention network. As a result, the network can better adapt to the variable distributions present in speech data. Our experiments show that our resulting model achieves the best recognition result on the Switchboard benchmark in the non-augmentation condition, and the best published result in the MuST-C speech translation benchmark. We also show that this model is able to better utilize synthetic data than the Transformer, and adapts better to variable sentence segmentation quality for speech translation.
As the advancement of deep learning (DL), the Internet of Things and cloud computing techniques for biomedical and healthcare problems, mobile healthcare systems have received unprecedented attention. Since DL techniques usually require enormous amou
nt of computation, most of them cannot be directly deployed on the resource-constrained mobile and IoT devices. Hence, most of the mobile healthcare systems leverage the cloud computing infrastructure, where the data collected by the mobile and IoT devices would be transmitted to the cloud computing platforms for analysis. However, in the contested environments, relying on the cloud might not be practical at all times. For instance, the satellite communication might be denied or disrupted. We propose SAIA, a Split Artificial Intelligence Architecture for mobile healthcare systems. Unlike traditional approaches for artificial intelligence (AI) which solely exploits the computational power of the cloud server, SAIA could not only relies on the cloud computing infrastructure while the wireless communication is available, but also utilizes the lightweight AI solutions that work locally on the client side, hence, it can work even when the communication is impeded. In SAIA, we propose a meta-information based decision unit, that could tune whether a sample captured by the client should be operated by the embedded AI (i.e., keeping on the client) or the networked AI (i.e., sending to the server), under different conditions. In our experimental evaluation, extensive experiments have been conducted on two popular healthcare datasets. Our results show that SAIA consistently outperforms its baselines in terms of both effectiveness and efficiency.
Opening up a new window of millimeter-wave observations that span frequency bands in the range of 30 to 500 GHz, survey half the sky, and are both an order of magnitude deeper (about 0.5 uK-arcmin) and of higher-resolution (about 10 arcseconds) than
currently funded surveys would yield an enormous gain in understanding of both fundamental physics and astrophysics. In particular, such a survey would allow for major advances in measuring the distribution of dark matter and gas on small-scales, and yield needed insight on 1.) dark matter particle properties, 2.) the evolution of gas and galaxies, 3.) new light particle species, 4.) the epoch of inflation, and 5.) the census of bodies orbiting in the outer Solar System.
We find new and compelling evidence for the meta-stability of SUSY-breaking states in holographic backgrounds whose consistency has been the source of ongoing disagreements in the literature. As a concrete example, we analyse anti-D3 branes at the ti
p of the Klebanov-Strassler (KS) throat. Using the blackfold formalism we examine how temperature affects the conjectured meta-stable state and determine whether and how the existing extremal results generalize when going beyond extremality. In the extremal limit we exactly recover the results of Kachru, Pearson and Verlinde (KPV), in a regime of parameter space that was previously inaccesible. Away from extremality we uncover a meta-stable black NS5 state that disappears near a geometric transition where black anti-D3 branes and black NS5 branes become indistinguishable. This is remarkably consistent with complementary earlier results based on the analysis of regularity conditions of backreacted solutions. We therefore provide highly non-trivial evidence for the meta-stability of anti-branes in non-compact throat geometries since we find a consistent picture over different regimes in parameter space.
We present a method to measure the small-scale matter power spectrum using high-resolution measurements of the gravitational lensing of the Cosmic Microwave Background (CMB). To determine whether small-scale structure today is suppressed on scales be
low 10 kiloparsecs (corresponding to M < 10^9 M_sun), one needs to probe CMB-lensing modes out to L ~ 35,000, requiring a CMB experiment with about 20 arcsecond resolution or better. We show that a CMB survey covering 4,000 square degrees of sky, with an instrumental sensitivity of 0.5 uK-arcmin at 18 arcsecond resolution, could distinguish between cold dark matter and an alternative, such as 1 keV warm dark matter or 10^(-22) eV fuzzy dark matter with about 4-sigma significance. A survey of the same resolution with 0.1 uK-arcmin noise could distinguish between cold dark matter and these alternatives at better than 20-sigma significance; such high-significance measurements may also allow one to distinguish between a suppression of power due to either baryonic effects or the particle nature of dark matter, since each impacts the shape of the lensing power spectrum differently. CMB temperature maps yield higher signal-to-noise than polarization maps in this small-scale regime; thus, systematic effects, such as from extragalactic astrophysical foregrounds, need to be carefully considered. However, these systematic concerns can likely be mitigated with known techniques. Next-generation CMB lensing may thus provide a robust and powerful method of measuring the small-scale matter power spectrum.
