Enhancing MR Image Segmentation with Realistic Adversarial Data Augmentation

The success of neural networks on medical image segmentation tasks typically relies on large labeled datasets for model training. However, acquiring and manually labeling a large medical image set is resource-intensive, expensive, and sometimes impractical due to data sharing and privacy issues. To address this challenge, we propose an adversarial data augmentation approach to improve the efficiency in utilizing training data and to enlarge the dataset via simulated but realistic transformations. Specifically, we present a generic task-driven learning framework, which jointly optimizes a data augmentation model and a segmentation network during training, generating informative examples to enhance network generalizability for the downstream task. The data augmentation model utilizes a set of photometric and geometric image transformations and chains them to simulate realistic complex imaging variations that could exist in magnetic resonance (MR) imaging. The proposed adversarial data augmentation does not rely on generative networks and can be used as a plug-in module in general segmentation networks. It is computationally efficient and applicable for both supervised and semi-supervised learning. We analyze and evaluate the method on two MR image segmentation tasks: cardiac segmentation and prostate segmentation. Results show that the proposed approach can alleviate the need for labeled data while improving model generalization ability, indicating its practical value in medical imaging applications.

Classifying Seyfert galaxies with deep learning

Traditional classification for subclass of the Seyfert galaxies is visual inspection or using a quantity defined as a flux ratio between the Balmer line and forbidden line. One algorithm of deep learning is Convolution Neural Network (CNN) and has shown successful classification results. We building a 1-dimension CNN model to distinguish Seyfert 1.9 spectra from Seyfert 2 galaxies. We find our model can recognize Seyfert 1.9 and Seyfert 2 spectra with an accuracy over 80% and pick out an additional Seyfert 1.9 sample which was missed by visual inspection. We use the new Seyfert 1.9 sample to improve performance of our model and obtain a 91% precision of Seyfert 1.9. These results indicate our model can pick out Seyfert 1.9 spectra among Seyfert 2 spectra. We decompose H{alpha} emission line of our Seyfert 1.9 galaxies by fitting 2 Gaussian components and derive line width and flux. We find velocity distribution of broad H{alpha} component of the new Seyfert 1.9 sample has an extending tail toward the higher end and luminosity of the new Seyfert 1.9 sample is slightly weaker than the original Seyfert 1.9 sample. This result indicates that our model can pick out the sources that have relatively weak broad H{alpha} component. Besides, we check distributions of the host galaxy morphology of our Seyfert 1.9 samples and find the distribution of the host galaxy morphology is dominant by large bulge galaxy. In the end, we present an online catalog of 1297 Seyfert 1.9 galaxies with measurement of H{alpha} emission line.

Cooperative Training and Latent Space Data Augmentation for Robust Medical Image Segmentation

Deep learning-based segmentation methods are vulnerable to unforeseen data distribution shifts during deployment, e.g. change of image appearances or contrasts caused by different scanners, unexpected imaging artifacts etc. In this paper, we present a cooperative framework for training image segmentation models and a latent space augmentation method for generating hard examples. Both contributions improve model generalization and robustness with limited data. The cooperative training framework consists of a fast-thinking network (FTN) and a slow-thinking network (STN). The FTN learns decoupled image features and shape features for image reconstruction and segmentation tasks. The STN learns shape priors for segmentation correction and refinement. The two networks are trained in a cooperative manner. The latent space augmentation generates challenging examples for training by masking the decoupled latent space in both channel-wise and spatial-wise manners. We performed extensive experiments on public cardiac imaging datasets. Using only 10 subjects from a single site for training, we demonstrated improved cross-site segmentation performance and increased robustness against various unforeseen imaging artifacts compared to strong baseline methods. Particularly, cooperative training with latent space data augmentation yields 15% improvement in terms of average Dice score when compared to a standard training method.

Deep Learning-Aided OFDM-Based Generalized Optical Quadrature Spatial Modulation

In this paper, we propose an orthogonal frequency division multiplexing (OFDM)-based generalized optical quadrature spatial modulation (GOQSM) technique for multiple-input multiple-output optical wireless communication (MIMO-OWC) systems. Considering the error propagation and noise amplification effects when applying maximum likelihood and maximum ratio combining (ML-MRC)-based detection, we further propose a deep neural network (DNN)-aided detection for OFDM-based GOQSM systems. The proposed DNN-aided detection scheme performs the GOQSM detection in a joint manner, which can efficiently eliminate the adverse effects of both error propagation and noise amplification. The obtained simulation results successfully verify the superiority of the deep learning-aided OFDM-based GOQSM technique for high-speed MIMO-OWC systems.

Pedestrian Attribute Recognition in Video Surveillance Scenarios Based on View-attribute Attention Localization

Pedestrian attribute recognition in surveillance scenarios is still a challenging task due to inaccurate localization of specific attributes. In this paper, we propose a novel view-attribute localization method based on attention (VALA), which relies on the strong relevance between attributes and views to capture specific view-attributes and to localize attribute-corresponding areas by attention mechanism. A specific view-attribute is composed by the extracted attribute feature and four view scores which are predicted by view predictor as the confidences for attribute from different views. View-attribute is then delivered back to shallow network layers for supervising deep feature extraction. To explore the location of a view-attribute, regional attention is introduced to aggregate spatial information of the input attribute feature in height and width direction for constraining the image into a narrow range. Moreover, the inter-channel dependency of view-feature is embedded in the above two spatial directions. An attention attribute-specific region is gained after fining the narrow range by balancing the ratio of channel dependencies between height and width branches. The final view-attribute recognition outcome is obtained by combining the output of regional attention with the view scores from view predictor. Experiments on three wide datasets (RAP, RAPv2, PETA, and PA-100K) demonstrate the effectiveness of our approach compared with state-of-the-art methods.

Spectral Independence via Stability and Applications to Holant-Type Problems

This paper formalizes connections between stability of polynomials and convergence rates of Markov Chain Monte Carlo (MCMC) algorithms. We prove that if a (multivariate) partition function is nonzero in a region around a real point $lambda$ then spectral independence holds at $lambda$. As a consequence, for Holant-type problems (e.g., spin systems) on bounded-degree graphs, we obtain optimal $O(nlog n)$ mixing time bounds for the single-site update Markov chain known as the Glauber dynamics. Our result significantly improves the running time guarantees obtained via the polynomial interpolation method of Barvinok (2017), refined by Patel and Regts (2017). There are a variety of applications of our results. In this paper, we focus on Holant-type (i.e., edge-coloring) problems, including weighted edge covers and weighted even subgraphs. For the weighted edge cover problem (and several natural generalizations) we obtain an $O(nlog{n})$ sampling algorithm on bounded-degree graphs. The even subgraphs problem corresponds to the high-temperature expansion of the ferromagnetic Ising model. We obtain an $O(nlog{n})$ sampling algorithm for the ferromagnetic Ising model with a nonzero external field on bounded-degree graphs, which improves upon the classical result of Jerrum and Sinclair (1993) for this class of graphs. We obtain further applications to antiferromagnetic two-spin models on line graphs, weighted graph homomorphisms, tensor networks, and more.

Deep learning based on PINN for solving 2 D0F vortex induced vibration of cylinder with high Reynolds number

Vortex-induced vibration (VIV) exists widely in natural and industrial fields. The main approaches for solving VIV problems are numerical simulations and experimental methods. However, experiment methods are difficult to obtain the whole flow field information and also high-cost while numerical simulation is extraordinary time-consuming and limited in low Reynolds number and simple geometric configuration. In addition, numerical simulations are difficult to handle the moving mesh technique. In this paper, physics informed neural network (PINN) is proposed to solve the VIV and wake-induced vibration (WIV) of cylinder with high Reynolds number. Compared to tradition data-driven neural network, the Reynolds Average Navier-Stokes (RANS) equation, by implanting an additional turbulent eddy viscosity, coupled with structures dynamic motion equation are also embedded into the loss function. Training and validation data is obtained by computational fluid dynamic (CFD) technique. Three scenarios are proposed to validate the performance of PINN in solving VIV and WIV of cylinders. In the first place, the stiffness parameter and damping parameter are calculated via limited force data and displacement data; secondly, the flow field and lifting force/drag force are inferred by scattered velocity information; eventually, the displacement can be directly predicted only through lifting forces and drag forces based on LSTM. Results demonstrate that,compared with traditional neural network, PINN method is more effective in inferring and re-constructing the unknown parameters and flow field with high Reynolds number under VIV and WIV circumstances.

SpeechNet: A Universal Modularized Model for Speech Processing Tasks

There is a wide variety of speech processing tasks ranging from extracting content information from speech signals to generating speech signals. For different tasks, model networks are usually designed and tuned separately. If a universal model can perform multiple speech processing tasks, some tasks might be improved with the related abilities learned from other tasks. The multi-task learning of a wide variety of speech processing tasks with a universal model has not been studied. This paper proposes a universal modularized model, SpeechNet, which treats all speech processing tasks into a speech/text input and speech/text output format. We select five essential speech processing tasks for multi-task learning experiments with SpeechNet. We show that SpeechNet learns all of the above tasks, and we further analyze which tasks can be improved by other tasks. SpeechNet is modularized and flexible for incorporating more modules, tasks, or training approaches in the future. We release the code and experimental settings to facilitate the research of modularized universal models and multi-task learning of speech processing tasks.

Experimental demonstration of conjugate-Franson interferometry

Franson interferometry is a well-known quantum measurement technique for probing photon-pair frequency correlations that is often used to certify time-energy entanglement. We demonstrate the complementary technique in the time basis, called conjugate-Franson interferometry, that measures photon-pair arrival-time correlations, thus providing a valuable addition to the quantum toolbox. We obtain a conjugate-Franson interference visibility of $96pm 1$% without background subtraction for entangled photon pairs generated by spontaneous parametric down-conversion. Our measured result surpasses the quantum-classical threshold by 25 standard deviations and validates the conjugate-Franson interferometer (CFI) as an alternative method for certifying time-energy entanglement. Moreover, the CFI visibility is a function of the biphotons joint temporal intensity and is therefore sensitive to that states spectral phase variation, something which is not the case for Franson interferometry or Hong-Ou-Mandel interferometry. We highlight the CFIs utility by measuring its visibilities for two different biphoton states, one without and the other with spectral phase variation, and observing a 21% reduction in the CFI visibility for the latter. The CFI is potentially useful for applications in areas of photonic entanglement, quantum communications, and quantum networking.

First search for direct $CP$-violating asymmetry in $B^0 to K^0 pi^0$ decays at Belle II

We report on the first measurement of the direct $CP$-violating asymmetry ($mathcal{A}$) in the charmless decay $B^0 to K^0pi^0$ at Belle II and an updated measurement of its branching fraction ($mathcal{B}$). We use a sample of electron-positron collisions collected in 2019 and 2020 at the $Upsilon(4S)$ resonance and corresponding to $62.8$ $text{fb}^{-1}$ of integrated luminosity. We reconstruct and select about $50$ $B^0 to K_S^0 pi^0$ candidates, and we measure $mathcal{A}_{{K^0pi^0}} = -0.40_{-0.44}^{+0.46} (text{stat}) pm 0.04 (text{syst})$ and $mathcal{B}(B^0 to K^0 pi^0) = [8.5_{-1.6}^{+1.7} (text{stat}) pm 1.2 (text{syst})] times 10^{-6}$. This is the first measurement of $CP$ violation in $B^0 to K^0pi^0$ decays reported by Belle II. The results agree with previous determinations and show a detector performance comparable with the best Belle results.