We propose a deep switching state space model (DS$^3$M) for efficient inference and forecasting of nonlinear time series with irregularly switching among various regimes. The switching among regimes is captured by both discrete and continuous latent
variables with recurrent neural networks. The model is estimated with variational inference using a reparameterization trick. We test the approach on a variety of simulated and real datasets. In all cases, DS$^3$M achieves competitive performance compared to several state-of-the-art methods (e.g. GRU, SRNN, DSARF, SNLDS), with superior forecasting accuracy, convincing interpretability of the discrete latent variables, and powerful representation of the continuous latent variables for different kinds of time series. Specifically, the MAPE values increase by 0.09% to 15.71% against the second-best performing alternative models.
Solving physical problems by deep learning is accurate and efficient mainly accounting for the use of an elaborate neural network. We propose a novel hybrid network which integrates two different kinds of neural networks: LSTM and ResNet, in order to
overcome the difficulty met in solving strongly-oscillating dynamics of the systems time evolution. By taking the double-well model as an example we show that our new method can benefit from a pre-learning and verification of the periodicity of frequency by using the LSTM network, simultaneously making a high-fidelity prediction about the whole dynamics of system with ResNet, which is impossibly achieved in the case of single network. Such a hybrid network can be applied for solving cooperative dynamics in a system with fast spatial or temporal modulations, promising for realistic oscillation calculations under experimental conditions.
Volatility for financial assets returns can be used to gauge the risk for financial market. We propose a deep stochastic volatility model (DSVM) based on the framework of deep latent variable models. It uses flexible deep learning models to automatic
ally detect the dependence of the future volatility on past returns, past volatilities and the stochastic noise, and thus provides a flexible volatility model without the need to manually select features. We develop a scalable inference and learning algorithm based on variational inference. In real data analysis, the DSVM outperforms several popular alternative volatility models. In addition, the predicted volatility of the DSVM provides a more reliable risk measure that can better reflex the risk in the financial market, reaching more quickly to a higher level when the market becomes more risky and to a lower level when the market is more stable, compared with the commonly used GARCH type model with a huge data set on the U.S. stock market.
Atrial Fibrillation (AF) is a common cardiac arrhythmia affecting a large number of people around the world. If left undetected, it will develop into chronic disability or even early mortality. However, patients who have this problem can barely feel
its presence, especially in its early stage. A non-invasive, automatic, and effective detection method is therefore needed to help early detection so that medical intervention can be implemented in time to prevent its progression. Electrocardiogram (ECG), which records the electrical activities of the heart, has been widely used for detecting the presence of AF. However, due to the subtle patterns of AF, the performance of detection models have largely depended on complicated data pre-processing and expertly engineered features. In our work, we developed DenseECG, an end-to-end model based on 5 layers 1D densely connected convolutional neural network. We trained our model using the publicly available dataset from 2017 PhysioNet Computing in Cardiology(CinC) Challenge containing 8528 single-lead ECG recordings of short-term heart rhythms (9-61s). Our trained model was able to outperform the other state-of-the-art AF detection models on this dataset without complicated data pre-processing and expert-supervised feature engineering.
Deep learning-based image denoising approaches have been extensively studied in recent years, prevailing in many public benchmark datasets. However, the stat-of-the-art networks are computationally too expensive to be directly applied on mobile devic
es. In this work, we propose a light-weight, efficient neural network-based raw image denoiser that runs smoothly on mainstream mobile devices, and produces high quality denoising results. Our key insights are twofold: (1) by measuring and estimating sensor noise level, a smaller network trained on synthetic sensor-specific data can out-perform larger ones trained on general data; (2) the large noise level variation under different ISO settings can be removed by a novel k-Sigma Transform, allowing a small network to efficiently handle a wide range of noise levels. We conduct extensive experiments to demonstrate the efficiency and accuracy of our approach. Our proposed mobile-friendly denoising model runs at ~70 milliseconds per megapixel on Qualcomm Snapdragon 855 chipset, and it is the basis of the night shot feature of several flagship smartphones released in 2019.
Understanding surface mechanics of soft solids, such as soft polymeric gels, is crucial in many engineering processes, such as dynamic wetting and adhesive failure. In these situations, a combination of capillary and elastic forces drives the motion,
which is balanced by dissipative mechanisms to determine the rate. While shear rheology (i.e. viscoelasticity) has long been assumed to dominate the dissipation, recent works have suggested that compressibility effects (i.e. poroelasticity) could play roles in swollen networks. We use fast interferometric imaging to quantify the relaxation of surface deformations due to a displaced contact line. By systematically measuring the profiles at different time and length scales, we experimentally observe a crossover from viscoelastic to poroelastic surface relaxations.
Inspired by a recent experiment [Phys. Rev. Letts. textbf{122}, 253201(2019)] that an unprecedented quantum interference was observed in the way of Stimulated Raman adiabatic passage (STIRAP) due to the coexisting resonant- and detuned-STIRAPs, we co
mprehensively study this effect for uncovering its robustness towards the external-field fluctuations of laser noise, imperfect resonance condition as well as the excited-state decaying. We verify that, an auxiliary dynamical phase accumulated in hold time caused by the quasi-dark state can sensitively manipulate the visibility and frequency of the interference fringe, representing a new hallmark to measure the hyperfine energy accurately. The robust stability of scheme comes from the intrinsic superiority embedded in STIRAP itself, which promises a remarkable preservation of the quantum interference quality in a practical implementation.
We study the equivalence principle, regarded as the cornerstone of general relativity, by analyzing the deformation observable of black hole shadows. Such deformation can arise from new physics and may be expressed as a phenomenological violation of
the equivalence principle. Specifically, we assume that there is an additional background vector field that couples to the photons around the supermassive black hole. This type of coupling yields impact on the way the system depends on initial conditions, and affects the black hole shadow at different wavelengths by a different amount, and therefore observations of the shadow in different wavelengths could constrain such couplings. This can be tested by future multi-band observations. Adopting a specific form of the vector field, we obtain constraints on model parameters from Event Horizon Telescope observations and measurements of gas/stellar orbits.
We experimentally investigate the rheology and stress fluctuations of granules densely suspended in silicone oil. We find that both thickening strength and stress fluctuations significantly weaken with oil viscosity $eta_0$. Comparison of our rheolog
ical results to the Wyart-Cates model for describing different dynamic jamming states suggests a transition from frictional contacts to lubrication interactions as $eta_0$ increases. To clarify the contribution from viscous interactions to the rheology, we systematically measure stress fluctuations in various flow states. Reduction of stress fluctuations with $eta_0$ indicates that a strong lubrication layer greatly inhibits force correlations among particles. Measuring stress fluctuations in the strong shear thickening regime, we observe a crossover from asymmetric Gamma to symmetric Gaussian distributions and associated with it a decrease of lateral (radial) correlation length $xi$ with increasing shear rate.
This paper considers a realistic problem in person re-identification (re-ID) task, i.e., partial re-ID. Under partial re-ID scenario, the images may contain a partial observation of a pedestrian. If we directly compare a partial pedestrian image with
a holistic one, the extreme spatial misalignment significantly compromises the discriminative ability of the learned representation. We propose a Visibility-aware Part Model (VPM), which learns to perceive the visibility of regions through self-supervision. The visibility awareness allows VPM to extract region-level features and compare two images with focus on their shared regions (which are visible on both images). VPM gains two-fold benefit toward higher accuracy for partial re-ID. On the one hand, compared with learning a global feature, VPM learns region-level features and benefits from fine-grained information. On the other hand, with visibility awareness, VPM is capable to estimate the shared regions between two images and thus suppresses the spatial misalignment. Experimental results confirm that our method significantly improves the learned representation and the achieved accuracy is on par with the state of the art.
