Respiratory diseases, including asthma, bronchitis, pneumonia, and upper respiratory tract infection (RTI), are among the most common diseases in clinics. The similarities among the symptoms of these diseases precludes prompt diagnosis upon the patie
nts arrival. In pediatrics, the patients limited ability in expressing their situation makes precise diagnosis even harder. This becomes worse in primary hospitals, where the lack of medical imaging devices and the doctors limited experience further increase the difficulty of distinguishing among similar diseases. In this paper, a pediatric fine-grained diagnosis-assistant system is proposed to provide prompt and precise diagnosis using solely clinical notes upon admission, which would assist clinicians without changing the diagnostic process. The proposed system consists of two stages: a test result structuralization stage and a disease identification stage. The first stage structuralizes test results by extracting relevant numerical values from clinical notes, and the disease identification stage provides a diagnosis based on text-form clinical notes and the structured data obtained from the first stage. A novel deep learning algorithm was developed for the disease identification stage, where techniques including adaptive feature infusion and multi-modal attentive fusion were introduced to fuse structured and text data together. Clinical notes from over 12000 patients with respiratory diseases were used to train a deep learning model, and clinical notes from a non-overlapping set of about 1800 patients were used to evaluate the performance of the trained model. The average precisions (AP) for pneumonia, RTI, bronchitis and asthma are 0.878, 0.857, 0.714, and 0.825, respectively, achieving a mean AP (mAP) of 0.819.
Exciton condensates (EC) are macroscopic coherent states arising from condensation of electron-hole pairs. Bilayer heterostructures, consisting of two-dimensional electron and hole layers separated by a tunnel barrier, provide a versatile platform to
realize and study EC. The tunnel barrier suppresses recombination yielding long-lived excitons. However, this separation also reduces interlayer Coulomb interactions, limiting the exciton binding strength. Here, we report the observation of EC in naturally occurring 2H-stacked bilayer WSe$_2$. In this system, the intrinsic spin-valley structure suppresses interlayer tunneling even when the separation is reduced to the atomic limit, providing access to a previously unattainable regime of strong interlayer coupling. Using capacitance spectroscopy, we investigate magneto-EC, formed when partially filled Landau levels (LL) couple between the layers. We find that the strong-coupling EC show dramatically different behaviour compared with previous reports, including an unanticipated variation of the EC robustness with the orbital number, and find evidence for a transition between two types of low-energy charged excitations. Our results provide a demonstration of tuning EC properties by varying the constituent single-particle wavefunctions.
Ocean current, fluid mechanics, and many other spatio-temporal physical dynamical systems are essential components of the universe. One key characteristic of such systems is that certain physics laws -- represented as ordinary/partial differential eq
uations (ODEs/PDEs) -- largely dominate the whole process, irrespective of time or location. Physics-informed learning has recently emerged to learn physics for accurate prediction, but they often lack a mechanism to leverage localized spatial and temporal correlation or rely on hard-coded physics parameters. In this paper, we advocate a physics-coupled neural network model to learn parameters governing the physics of the system, and further couple the learned physics to assist the learning of recurring dynamics. A spatio-temporal physics-coupled neural network (ST-PCNN) model is proposed to achieve three goals: (1) learning the underlying physics parameters, (2) transition of local information between spatio-temporal regions, and (3) forecasting future values for the dynamical system. The physics-coupled learning ensures that the proposed model can be tremendously improved by using learned physics parameters, and can achieve good long-range forecasting (e.g., more than 30-steps). Experiments, using simulated and field-collected ocean current data, validate that ST-PCNN outperforms existing physics-informed models.
In this work, we study the behavior of blow-up solutions to the multidimensional restricted Euler--Poisson equations which are the localized version of the full Euler--Poisson system. We provide necessary conditions for the existence of finite-time b
low-up solutions in terms of the initial data, and describe the asymptotic behavior of the solutions near blow up times. We also identify a rich set of the initial data which yields global bounded solutions.
Federated learning (FL) allows edge devices to collectively learn a model without directly sharing data within each device, thus preserving privacy and eliminating the need to store data globally. While there are promising results under the assumptio
n of independent and identically distributed (iid) local data, current state-of-the-art algorithms suffer from performance degradation as the heterogeneity of local data across clients increases. To resolve this issue, we propose a simple framework, Mean Augmented Federated Learning (MAFL), where clients send and receive averaged local data, subject to the privacy requirements of target applications. Under our framework, we propose a new augmentation algorithm, named FedMix, which is inspired by a phenomenal yet simple data augmentation method, Mixup, but does not require local raw data to be directly shared among devices. Our method shows greatly improved performance in the standard benchmark datasets of FL, under highly non-iid federated settings, compared to conventional algorithms.
The needs for precisely estimating a students academic performance have been emphasized with an increasing amount of attention paid to Intelligent Tutoring System (ITS). However, since labels for academic performance, such as test scores, are collect
ed from outside of ITS, obtaining the labels is costly, leading to label-scarcity problem which brings challenge in taking machine learning approaches for academic performance prediction. To this end, inspired by the recent advancement of pre-training method in natural language processing community, we propose DPA, a transfer learning framework with Discriminative Pre-training tasks for Academic performance prediction. DPA pre-trains two models, a generator and a discriminator, and fine-tunes the discriminator on academic performance prediction. In DPAs pre-training phase, a sequence of interactions where some tokens are masked is provided to the generator which is trained to reconstruct the original sequence. Then, the discriminator takes an interaction sequence where the masked tokens are replaced by the generators outputs, and is trained to predict the originalities of all tokens in the sequence. Compared to the previous state-of-the-art generative pre-training method, DPA is more sample efficient, leading to fast convergence to lower academic performance prediction error. We conduct extensive experimental studies on a real-world dataset obtained from a multi-platform ITS application and show that DPA outperforms the previous state-of-the-art generative pre-training method with a reduction of 4.05% in mean absolute error and more robust to increased label-scarcity.
This paper is concerned with a stochastic linear-quadratic (LQ) optimal control problem on infinite time horizon, with regime switching, random coefficients, and cone control constraint. Two new extended stochastic Riccati equations (ESREs) on infini
te time horizon are introduced. The existence of the nonnegative solutions, in both standard and singular cases, is proved through a sequence of ESREs on finite time horizon. Based on this result and some approximation techniques, we obtain the optimal state feedback control and optimal value for the stochastic LQ problem explicitly, which also implies the uniqueness of solutions for the ESREs. Finally, we apply these results to solve a lifetime portfolio selection problem of tracking a given wealth level with regime switching and portfolio constraint.
In semi-supervised domain adaptation, a few labeled samples per class in the target domain guide features of the remaining target samples to aggregate around them. However, the trained model cannot produce a highly discriminative feature representati
on for the target domain because the training data is dominated by labeled samples from the source domain. This could lead to disconnection between the labeled and unlabeled target samples as well as misalignment between unlabeled target samples and the source domain. In this paper, we propose a novel approach called Cross-domain Adaptive Clustering to address this problem. To achieve both inter-domain and intra-domain adaptation, we first introduce an adversarial adaptive clustering loss to group features of unlabeled target data into clusters and perform cluster-wise feature alignment across the source and target domains. We further apply pseudo labeling to unlabeled samples in the target domain and retain pseudo-labels with high confidence. Pseudo labeling expands the number of ``labeled samples in each class in the target domain, and thus produces a more robust and powerful cluster core for each class to facilitate adversarial learning. Extensive experiments on benchmark datasets, including DomainNet, Office-Home and Office, demonstrate that our proposed approach achieves the state-of-the-art performance in semi-supervised domain adaptation.
We study the problem of embedding edgeless nodes such as users who newly enter the underlying network, while using graph neural networks (GNNs) widely studied for effective representation learning of graphs thanks to its highly expressive capability
via message passing. Our study is motivated by the fact that existing GNNs cannot be adopted for our problem since message passing to such edgeless nodes having no connections is impossible. To tackle this challenge, we propose Edgeless-GNN, a new framework that enables GNNs to generate node embeddings even for edgeless nodes through unsupervised inductive learning. Specifically, we start by constructing a $k$-nearest neighbor graph ($k$NNG) based on the similarity of node attributes to replace the GNNs computation graph defined by the neighborhood-based aggregation of each node. As our main contributions, the known network structure is used to train model parameters, while a new loss function is established using energy-based learning in such a way that our model learns the network structure. For the edgeless nodes, we inductively infer embeddings for the edgeless nodes by using edges via $k$NNG construction as a computation graph. By evaluating the performance of various downstream machine learning (ML) tasks, we empirically demonstrate that Edgeless-GNN consistently outperforms state-of-the-art methods of inductive network embedding. Moreover, our findings corroborate the effectiveness of Edgeless-GNN in judiciously combining the replaced computation graph with our newly designed loss. Our framework is GNN-model-agnostic; thus, GNN models can be appropriately chosen according to ones needs and ML tasks.
In moire heterostructures, gate-tunable insulating phases driven by electronic correlations have been recently discovered. Here, we use transport measurements to characterize the gate-driven metal-insulator transitions and the metallic phase in twist
ed WSe$_2$ near half filling of the first moire subband. We find that the metal-insulator transition as a function of both density and displacement field is continuous. At the metal-insulator boundary, the resistivity displays strange metal behaviour at low temperature with dissipation comparable to the Planckian limit. Further into the metallic phase, Fermi-liquid behaviour is recovered at low temperature which evolves into a quantum critical fan at intermediate temperatures before eventually reaching an anomalous saturated regime near room temperature. An analysis of the residual resistivity indicates the presence of strong quantum fluctuations in the insulating phase. These results establish twisted WSe$_2$ as a new platform to study doping and bandwidth controlled metal-insulator quantum phase transitions on the triangular lattice.
