arXiv subjects

Bin Yang

Publications and source records attributed to Bin Yang.

At least 325 records · Page 18Linked to original sources

arXiv2021

BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule

Differentiable Architecture Search (DARTS) has received massive attention in recent years, mainly because it significantly reduces the computational cost through weight sharing and continuous relaxation. However, more recent works find that existing differentiable NAS techniques struggle to outperform naive baselines, yielding deteriorative architectures as the search proceeds. Rather than directly optimizing the architecture parameters, this paper formulates the neural architecture search as a distribution learning problem through relaxing the architecture weights into Gaussian distributions. By leveraging the natural-gradient variational inference (NGVI), the architecture distribution can be easily optimized based on existing codebases without incurring more memory and computational consumption. We demonstrate how the differentiable NAS benefits from Bayesian principles, enhancing exploration and improving stability. The experimental results on NAS-Bench-201 and NAS-Bench-1shot1 benchmark datasets confirm the significant improvements the proposed framework can make. In addition, instead of simply applying the argmax on the learned parameters, we further leverage the recently-proposed training-free proxies in NAS to select the optimal architecture from a group architectures drawn from the optimized distribution, where we achieve state-of-the-art results on the NAS-Bench-201 and NAS-Bench-1shot1 benchmarks. Our best architecture in the DARTS search space also obtains competitive test errors with 2.37\%, 15.72\%, and 24.2\% on CIFAR-10, CIFAR-100, and ImageNet datasets, respectively.

cs.LG↗

arXiv2021

Unsupervised Time Series Outlier Detection with Diversity-Driven Convolutional Ensembles -- Extended Version

With the sweeping digitalization of societal, medical, industrial, and scientific processes, sensing technologies are being deployed that produce increasing volumes of time series data, thus fueling a plethora of new or improved applications. In this setting, outlier detection is frequently important, and while solutions based on neural networks exist, they leave room for improvement in terms of both accuracy and efficiency. With the objective of achieving such improvements, we propose a diversity-driven, convolutional ensemble. To improve accuracy, the ensemble employs multiple basic outlier detection models built on convolutional sequence-to-sequence autoencoders that can capture temporal dependencies in time series. Further, a novel diversity-driven training method maintains diversity among the basic models, with the aim of improving the ensemble's accuracy. To improve efficiency, the approach enables a high degree of parallelism during training. In addition, it is able to transfer some model parameters from one basic model to another, which reduces training time. We report on extensive experiments using real-world multivariate time series that offer insight into the design choices underlying the new approach and offer evidence that it is capable of improved accuracy and efficiency. This is an extended version of "Unsupervised Time Series Outlier Detection with Diversity-Driven Convolutional Ensembles", to appear in PVLDB 2022.

cs.LG↗

arXiv2021

BlueMUSE: Project Overview and Science Cases

We present the concept of BlueMUSE, a blue-optimised, medium spectral resolution, panoramic integral field spectrograph based on the MUSE concept and proposed for the Very Large Telescope. With an optimised transmission down to 350 nm, a larger FoV (1.4 x 1.4 arcmin$^2$) and a higher spectral resolution compared to MUSE, BlueMUSE will open up a new range of galactic and extragalactic science cases allowed by its specific capabilities, beyond those possible with MUSE. For example a survey of massive stars in our galaxy and the Local Group will increase the known population of massive stars by a factor $>$100, to answer key questions about their evolution. Deep field observations with BlueMUSE will also significantly increase samples of Lyman-alpha emitters, spanning the era of Cosmic Noon. This will revolutionise the study of the distant Universe: allowing the intergalactic medium to be detected unambiguously in emission, enabling the study of the exchange of baryons between galaxies and their surroundings. By 2030, at a time when the focus of most of the new large facilities (ELT, JWST) will be on the infra-red, BlueMUSE will be a unique facility, outperforming any ELT instrument in the Blue/UV. It will have a strong synergy with ELT, JWST as well as ALMA, SKA, Euclid and Athena.

astro-ph.IM↗

arXiv2021

BuffetFS: Serve Yourself Permission Checks without Remote Procedure Calls

The remote procedure call (a.k.a. RPC) latency becomes increasingly significant in a distributed file system. We propose BuffetFS, a user-level file system that optimizes I/O performance by eliminating the RPCs caused by \texttt{open()} operation. By leveraging \texttt{open()} from file servers to clients, BuffetFS can restrain the procedure calls for permission checks locally, hence avoid RPCs during the initial stage to access a file. BuffetFS can further reduce response time when users are accessing a large number of small files. We implement a BuffetFS prototype and integrate it into a storage cluster. Our preliminary evaluation results show that BuffetFS can offer up to 70\% performance gain compared to the Lustre file system.

cs.DC↗

arXiv2021

USIS: Unsupervised Semantic Image Synthesis

Semantic Image Synthesis (SIS) is a subclass of image-to-image translation where a photorealistic image is synthesized from a segmentation mask. SIS has mostly been addressed as a supervised problem. However, state-of-the-art methods depend on a huge amount of labeled data and cannot be applied in an unpaired setting. On the other hand, generic unpaired image-to-image translation frameworks underperform in comparison, because they color-code semantic layouts and feed them to traditional convolutional networks, which then learn correspondences in appearance instead of semantic content. In this initial work, we propose a new Unsupervised paradigm for Semantic Image Synthesis (USIS) as a first step towards closing the performance gap between paired and unpaired settings. Notably, the framework deploys a SPADE generator that learns to output images with visually separable semantic classes using a self-supervised segmentation loss. Furthermore, in order to match the color and texture distribution of real images without losing high-frequency information, we propose to use whole image wavelet-based discrimination. We test our methodology on 3 challenging datasets and demonstrate its ability to generate multimodal photorealistic images with an improved quality in the unpaired setting.

cs.CV↗

arXiv2021

Improving Uncertainty of Deep Learning-based Object Classification on Radar Spectra using Label Smoothing

Object type classification for automotive radar has greatly improved with recent deep learning (DL) solutions, however these developments have mostly focused on the classification accuracy. Before employing DL solutions in safety-critical applications, such as automated driving, an indispensable prerequisite is the accurate quantification of the classifiers' reliability. Unfortunately, DL classifiers are characterized as black-box systems which output severely over-confident predictions, leading downstream decision-making systems to false conclusions with possibly catastrophic consequences. We find that deep radar classifiers maintain high-confidences for ambiguous, difficult samples, e.g. small objects measured at large distances, under domain shift and signal corruptions, regardless of the correctness of the predictions. The focus of this article is to learn deep radar spectra classifiers which offer robust real-time uncertainty estimates using label smoothing during training. Label smoothing is a technique of refining, or softening, the hard labels typically available in classification datasets. In this article, we exploit radar-specific know-how to define soft labels which encourage the classifiers to learn to output high-quality calibrated uncertainty estimates, thereby partially resolving the problem of over-confidence. Our investigations show how simple radar knowledge can easily be combined with complex data-driven learning algorithms to yield safe automotive radar perception.

cs.LG↗

arXiv2021

Estimation of Bivariate Structural Causal Models by Variational Gaussian Process Regression Under Likelihoods Parametrised by Normalising Flows

One major drawback of state-of-the-art artificial intelligence is its lack of explainability. One approach to solve the problem is taking causality into account. Causal mechanisms can be described by structural causal models. In this work, we propose a method for estimating bivariate structural causal models using a combination of normalising flows applied to density estimation and variational Gaussian process regression for post-nonlinear models. It facilitates causal discovery, i.e. distinguishing cause and effect, by either the independence of cause and residual or a likelihood ratio test. Our method which estimates post-nonlinear models can better explain a variety of real-world cause-effect pairs than a simple additive noise model. Though it remains difficult to exploit this benefit regarding all pairs from the Tübingen benchmark database, we demonstrate that combining the additive noise model approach with our method significantly enhances causal discovery.

stat.ML↗

arXiv2021

PLUMENet: Efficient 3D Object Detection from Stereo Images

3D object detection is a key component of many robotic applications such as self-driving vehicles. While many approaches rely on expensive 3D sensors such as LiDAR to produce accurate 3D estimates, methods that exploit stereo cameras have recently shown promising results at a lower cost. Existing approaches tackle this problem in two steps: first depth estimation from stereo images is performed to produce a pseudo LiDAR point cloud, which is then used as input to a 3D object detector. However, this approach is suboptimal due to the representation mismatch, as the two tasks are optimized in two different metric spaces. In this paper we propose a model that unifies these two tasks and performs them in the same metric space. Specifically, we directly construct a pseudo LiDAR feature volume (PLUME) in 3D space, which is then used to solve both depth estimation and object detection tasks. Our approach achieves state-of-the-art performance with much faster inference times when compared to existing methods on the challenging KITTI benchmark.

cs.CV↗

arXiv2021

A Deep Learning-based Quality Assessment and Segmentation System with a Large-scale Benchmark Dataset for Optical Coherence Tomographic Angiography Image

Optical Coherence Tomography Angiography (OCTA) is a non-invasive and non-contacting imaging technique providing visualization of microvasculature of retina and optic nerve head in human eyes in vivo. The adequate image quality of OCTA is the prerequisite for the subsequent quantification of retinal microvasculature. Traditionally, the image quality score based on signal strength is used for discriminating low quality. However, it is insufficient for identifying artefacts such as motion and off-centration, which rely specialized knowledge and need tedious and time-consuming manual identification. One of the most primary issues in OCTA analysis is to sort out the foveal avascular zone (FAZ) region in the retina, which highly correlates with any visual acuity disease. However, the variations in OCTA visual quality affect the performance of deep learning in any downstream marginally. Moreover, filtering the low-quality OCTA images out is both labor-intensive and time-consuming. To address these issues, we develop an automated computer-aided OCTA image processing system using deep neural networks as the classifier and segmentor to help ophthalmologists in clinical diagnosis and research. This system can be an assistive tool as it can process OCTA images of different formats to assess the quality and segment the FAZ area. The source code is freely available at https://github.com/shanzha09/COIPS.git. Another major contribution is the large-scale OCTA dataset, namely OCTA-25K-IQA-SEG we publicize for performance evaluation. It is comprised of four subsets, namely sOCTA-3$\times$3-10k, sOCTA-6$\times$6-14k, sOCTA-3$\times$3-1.1k-seg, and dOCTA-6$\times$6-1.1k-seg, which contains a total number of 25,665 images. The large-scale OCTA dataset is available at https://doi.org/10.5281/zenodo.5111975, https://doi.org/10.5281/zenodo.5111972.

eess.IV↗

arXiv2021

LAPNet: Non-rigid Registration derived in k-space for Magnetic Resonance Imaging

Physiological motion, such as cardiac and respiratory motion, during Magnetic Resonance (MR) image acquisition can cause image artifacts. Motion correction techniques have been proposed to compensate for these types of motion during thoracic scans, relying on accurate motion estimation from undersampled motion-resolved reconstruction. A particular interest and challenge lie in the derivation of reliable non-rigid motion fields from the undersampled motion-resolved data. Motion estimation is usually formulated in image space via diffusion, parametric-spline, or optical flow methods. However, image-based registration can be impaired by remaining aliasing artifacts due to the undersampled motion-resolved reconstruction. In this work, we describe a formalism to perform non-rigid registration directly in the sampled Fourier space, i.e. k-space. We propose a deep-learning based approach to perform fast and accurate non-rigid registration from the undersampled k-space data. The basic working principle originates from the Local All-Pass (LAP) technique, a recently introduced optical flow-based registration. The proposed LAPNet is compared against traditional and deep learning image-based registrations and tested on fully-sampled and highly-accelerated (with two undersampling strategies) 3D respiratory motion-resolved MR images in a cohort of 40 patients with suspected liver or lung metastases and 25 healthy subjects. The proposed LAPNet provided consistent and superior performance to image-based approaches throughout different sampling trajectories and acceleration factors.

eess.IV↗

arXiv2021

Non-detection of water-ice grains in the coma of comet 46P/Wirtanen and implications for hyperactivity

Hyperactive comets have high water production rates, with inferred sublimation areas of order the surface area of the nucleus. Comets 46P/Wirtanen and 103P/Hartley 2 are two examples of this cometary class. Based on observations of comet Hartley 2 by the Deep Impact spacecraft, hyperactivity appears to be caused by the ejection of water-ice grains and/or water-ice rich chunks of nucleus into the coma. These materials increase the sublimating surface area, and yield high water production rates. The historic close approach of comet Wirtanen to Earth in 2018 afforded an opportunity to test Hartley 2 style hyperactivity in a second Jupiter-family comet. We present high spatial resolution, near-infrared spectroscopy of the inner coma of Wirtanen. No evidence for the 1.5- or 2.0-$μ$m water-ice absorption bands is found in six 0.8-2.5 $μ$m spectra taken around perihelion and closest approach to Earth. In addition, the strong 3.0-$μ$m water-ice absorption band is absent in a 2.0-5.3 $μ$m spectrum taken near perihelion. Using spectroscopic and sublimation lifetime models we set constraints on the physical properties of the ice grains in the coma, assuming they are responsible for the comet's hyperactivity. We rule out pure water-ice grains of any size, given their long lifetime. Instead, the hyperactivity of the nucleus and lack of water-ice absorption features in our spectra can be explained either by icy grains on the order of 1 $μ$m in size with a small amount of low albedo dust (greater than 0.5% by volume), or large chunks containing significant amounts of water ice.

astro-ph.EP↗

arXiv2021

Unsupervised Path Representation Learning with Curriculum Negative Sampling

Path representations are critical in a variety of transportation applications, such as estimating path ranking in path recommendation systems and estimating path travel time in navigation systems. Existing studies often learn task-specific path representations in a supervised manner, which require a large amount of labeled training data and generalize poorly to other tasks. We propose an unsupervised learning framework Path InfoMax (PIM) to learn generic path representations that work for different downstream tasks. We first propose a curriculum negative sampling method, for each input path, to generate a small amount of negative paths, by following the principles of curriculum learning. Next, \emph{PIM} employs mutual information maximization to learn path representations from both a global and a local view. In the global view, PIM distinguishes the representations of the input paths from those of the negative paths. In the local view, \emph{PIM} distinguishes the input path representations from the representations of the nodes that appear only in the negative paths. This enables the learned path representations to encode both global and local information at different scales. Extensive experiments on two downstream tasks, ranking score estimation and travel time estimation, using two road network datasets suggest that PIM significantly outperforms other unsupervised methods and is also able to be used as a pre-training method to enhance supervised path representation learning.

cs.LG↗

arXiv2021

Investigation of Uncertainty of Deep Learning-based Object Classification on Radar Spectra

Deep learning (DL) has recently attracted increasing interest to improve object type classification for automotive radar.In addition to high accuracy, it is crucial for decision making in autonomous vehicles to evaluate the reliability of the predictions; however, decisions of DL networks are non-transparent. Current DL research has investigated how uncertainties of predictions can be quantified, and in this article, we evaluate the potential of these methods for safe, automotive radar perception. In particular we evaluate how uncertainty quantification can support radar perception under (1) domain shift, (2) corruptions of input signals, and (3) in the presence of unknown objects. We find that in agreement with phenomena observed in the literature,deep radar classifiers are overly confident, even in their wrong predictions. This raises concerns about the use of the confidence values for decision making under uncertainty, as the model fails to notify when it cannot handle an unknown situation. Accurate confidence values would allow optimal integration of multiple information sources, e.g. via sensor fusion. We show that by applying state-of-the-art post-hoc uncertainty calibration, the quality of confidence measures can be significantly improved,thereby partially resolving the over-confidence problem. Our investigation shows that further research into training and calibrating DL networks is necessary and offers great potential for safe automotive object classification with radar sensors.

cs.LG↗

arXiv2021

Investigating Cross-Domain Losses for Speech Enhancement

Recent years have seen a surge in the number of available frameworks for speech enhancement (SE) and recognition. Whether model-based or constructed via deep learning, these frameworks often rely in isolation on either time-domain signals or time-frequency (TF) representations of speech data. In this study, we investigate the advantages of each set of approaches by separately examining their impact on speech intelligibility and quality. Furthermore, we combine the fragmented benefits of time-domain and TF speech representations by introducing two new cross-domain SE frameworks. A quantitative comparative analysis against recent model-based and deep learning SE approaches is performed to illustrate the merit of the proposed frameworks.

cs.SD↗

arXiv2021

Discovery of Carbon Monoxide in Distant Comet C/2017 K2 (PANSTARRS)

Optical observations of the Oort cloud comet C/2017 K2 (PANSTARRS) show that its activity began at large heliocentric distances (up to 35 au), which cannot be explained by either the sublimation or the crystallization of water ice. Supervolatile sublimation, most likely of carbon monoxide (CO), has been proposed as a plausible driver of the observed mass loss. Here, we present the detection of the J = 2$-$1 rotational transition in outgassed CO from C/2017 K2 when at heliocentric distance $r_H$ = 6.72 au, using the James Clerk Maxwell Telescope. The CO line is blue-shifted by 0.20$\pm$0.03 km s$^{-1}$ with an area and width of 8.3$\pm$2.3 mK km s$^{-1}$ and $0.28\pm$0.08 km s$^{-1}$, respectively. The CO production rate is $Q_{CO} = (1.6\pm0.5) \times10^{27}$ s$^{-1}$. These are the first observations of a gaseous species in C/2017 K2 and provide observational confirmation of the role of supervolatile sublimation in this comet.

astro-ph.EP↗

arXiv2021

Contrastive Learning and Self-Training for Unsupervised Domain Adaptation in Semantic Segmentation

Deep convolutional neural networks have considerably improved state-of-the-art results for semantic segmentation. Nevertheless, even modern architectures lack the ability to generalize well to a test dataset that originates from a different domain. To avoid the costly annotation of training data for unseen domains, unsupervised domain adaptation (UDA) attempts to provide efficient knowledge transfer from a labeled source domain to an unlabeled target domain. Previous work has mainly focused on minimizing the discrepancy between the two domains by using adversarial training or self-training. While adversarial training may fail to align the correct semantic categories as it minimizes the discrepancy between the global distributions, self-training raises the question of how to provide reliable pseudo-labels. To align the correct semantic categories across domains, we propose a contrastive learning approach that adapts category-wise centroids across domains. Furthermore, we extend our method with self-training, where we use a memory-efficient temporal ensemble to generate consistent and reliable pseudo-labels. Although both contrastive learning and self-training (CLST) through temporal ensembling enable knowledge transfer between two domains, it is their combination that leads to a symbiotic structure. We validate our approach on two domain adaptation benchmarks: GTA5 $\rightarrow$ Cityscapes and SYNTHIA $\rightarrow$ Cityscapes. Our method achieves better or comparable results than the state-of-the-art. We will make the code publicly available.

cs.CV↗

arXiv2021

Intelligent Decision Method for Main Control Parameters of Tunnel Boring Machine based on Multi-Objective Optimization of Excavation Efficiency and Cost

Timely and reasonable matching of the control parameters and geological conditions of the rock mass in tunnel excavation is crucial for hard rock tunnel boring machines (TBMs). Therefore, this paper proposes an intelligent decision method for the main control parameters of the TBM based on the multi-objective optimization of excavation efficiency and cost. The main objectives of this method are to obtain the most important parameters of the rock mass and machine, determine the optimization objective, and establish the objective function. In this study, muck information was included as an important parameter in the traditional rock mass and machine parameter database. The rock-machine interaction model was established through an improved neural network algorithm. Using 250 sets of data collected in the field, the validity of the rock-machine interaction relationship model was verified. Then, taking the cost as the optimization objective, the cost calculation model related to tunneling and the cutter was obtained. Subsequently, combined with rock-machine interaction model, the objective function of control parameter optimization based on cost was established. Finally, a tunneling test was carried out at the engineering site, and the main TBM control parameters (thrust and torque) after the optimization decision were used to excavate the test section. Compared with the values in the section where the TBM operators relied on experience, the average penetration rate of the TBM increased by 11.10%, and the average cutter life increased by 15.62%. The results indicate that this method can play an effective role in TBM tunneling in the test section.

eess.SY↗

arXiv2021

Compact pebbles and the evolution of volatiles in the interstellar comet 2I/Borisov

The interstellar traveler, 2I/Borisov, is the first clearly active extrasolar comet, ever detected in our Solar system. We obtained high-resolution interferometric observations of 2I/Borisov with the Atacama Large Millimeter/submillimeter Array (ALMA), and multi-color optical observations with the Very Large Telescope (VLT) to gain a comprehensive understanding of the dust properties of this comet. We found that the dust coma of 2I/Borisov consists of compact "pebbles" of radii exceeding ~1 mm, suggesting that the dust particles have experienced compaction through mutual impacts during the bouncing collision phase in the protoplanetary disk. We derived a dust mass loss rate of >= 200 kg/s and a dust-to-gas ratio >=3. Our long term monitoring of 2I/Borisov with VLT indicates a steady dust mass loss with no significant dust fragmentation and/or sublimation occurring in the coma. We also detected emissions from carbon monoxide gas (CO) with ALMA and derived the gas production rate of Q(CO) (3.3+/-0.8)x10^{26} mole/s. We found that the CO/H$_2$O mixing ratio of 2I/Borisov changed drastically before and after perihelion, indicating the heterogeneity of the cometary nucleus, with components formed at different locations beyond the volatile snow-line with different chemical abundances. Our observations suggest that 2I/Borisov's home system, much like our own system, experienced efficient radial mixing from the innermost parts of its protoplanetary disk to beyond the frost line of CO.

astro-ph.EP↗