arXiv subjects

Wei Zhu

Publications and source records attributed to Wei Zhu.

At least 199 records · Page 11Linked to original sources

arXiv2022

ET White Paper: To Find the First Earth 2.0

We propose to develop a wide-field and ultra-high-precision photometric survey mission, temporarily named "Earth 2.0 (ET)". This mission is designed to measure, for the first time, the occurrence rate and the orbital distributions of Earth-sized planets. ET consists of seven 30cm telescopes, to be launched to the Earth-Sun's L2 point. Six of these are transit telescopes with a field of view of 500 square degrees. Staring in the direction that encompasses the original Kepler field for four continuous years, this monitoring will return tens of thousands of transiting planets, including the elusive Earth twins orbiting solar-type stars. The seventh telescope is a 30cm microlensing telescope that will monitor an area of 4 square degrees toward the galactic bulge. This, combined with simultaneous ground-based KMTNet observations, will measure masses for hundreds of long-period and free-floating planets. Together, the transit and the microlensing telescopes will revolutionize our understandings of terrestrial planets across a large swath of orbital distances and free space. In addition, the survey data will also facilitate studies in the fields of asteroseismology, Galactic archeology, time-domain sciences, and black holes in binaries.

astro-ph.IM↗

arXiv2022

Identification of stellar-mass black hole binaries and the validity of linear orbital motion approximation in microlensing

Gravitational microlensing is unique in detecting binary black (BH) holes with wide (a few au) separations. Models predict that about $1\%$ of microlensing binaries should be due to binary BHs, and yet zero has been robustly identified. Using simulated events with binary BH lenses, we show that the microlensing parallax effect in a typical binary BH event cannot be reliably detected. Given the crucial role of the parallax parameter in determining the mass of dark microlenses, this may explain the non-detection of binary BHs. Additionally, we show that in only a small fraction ($\lesssim7\%$) of the simulated events the full orbital motion of the binary lens cannot be modeled with the linear orbital motion approximation. This approximation has been frequently used in modelings of binary microlensing events.

astro-ph.HE↗

arXiv2022

Systematic KMTNet Planetary Anomaly Search, Paper I: OGLE-2019-BLG-1053Lb, A Buried Terrestrial Planet

In order to exhume the buried signatures of "missing planetary caustics" in the KMTNet data, we conducted a systematic anomaly search to the residuals from point-source point-lens fits, based on a modified version of the KMTNet EventFinder algorithm. This search reveals the lowest mass-ratio planetary caustic to date in the microlensing event OGLE-2019-BLG-1053, for which the planetary signal had not been noticed before. The planetary system has a planet-host mass ratio of $q = (1.25 \pm 0.13) \times 10^{-5}$. A Bayesian analysis yields estimates of the mass of the host star, $M_{\rm host} = 0.61_{-0.24}^{+0.29}~M_\odot$, the mass of its planet, $M_{\rm planet} = 2.48_{-0.98}^{+1.19}~M_{\oplus}$, the projected planet-host separation, $a_\perp = 3.4_{-0.5}^{+0.5}$ au, and the lens distance of $D_{\rm L} = 6.8_{-0.9}^{+0.6}$ kpc. The discovery of this very low mass-ratio planet illustrates the utility of our method and opens a new window for a large and homogeneous sample to study the microlensing planet-host mass-ratio function down to $q \sim 10^{-5}$.

astro-ph.EP↗

arXiv2022

KMT-2021-BLG-0171Lb and KMT-2021-BLG-1689Lb: Two Microlensing Planets in the KMTNet High-cadence Fields with Followup Observations

Follow-up observations of high-magnification gravitational microlensing events can fully exploit their intrinsic sensitivity to detect extrasolar planets, especially those with small mass ratios. To make followup more uniform and efficient, we develop a system, HighMagFinder, based on the real-time data from the Korean Microlensing Telescope Network (KMTNet) to automatically alert possible ongoing high-magnification events. We started a new phase of follow-up observations with the help of HighMagFinder in 2021. Here we report the discovery of two planets in high-magnification microlensing events, KMT-2021-BLG-0171 and KMT-2021-BLG-1689, which were identified by the HighMagFinder. We find that both events suffer the ``central-resonant'' caustic degeneracy. The planet-host mass-ratio is $q\sim4.7\times10^{-5}$ or $q\sim 2.2\times10^{-5}$ for KMT-2021-BLG-0171, and $q\sim2.5\times10^{-4}$ or $q\sim 1.8\times10^{-4}$ for KMT-2021-BLG-1689. Together with two events reported by Ryu et al. (2022), four cases that suffer such degeneracy have been discovered in the 2021 season alone, indicating that the degenerate solutions may have been missed in some previous studies. We also propose a new factor for weighting the probability of each solution from the phase-space. The resonant interpretations for the two events are disfavored under this consideration. This factor can be included in future statistical studies to weight degenerate solutions.

astro-ph.EP↗

arXiv2022

The intrinsic multiplicity distribution of exoplanets revealed from the radial velocity method

Planet multiplicities are useful in constraining the formation and evolution of planetary systems but usually difficult to constrain observationally. Here, we develop a general method that can properly take into account the survey incompleteness and recover the intrinsic planet multiplicity distribution. We then apply it to the radial velocity (RV) planet sample from the California Legacy Survey (CLS). Within the $1\,$au ($10\,$au) region, we find $21 \pm 4\%$ ($19.2 \pm 2.8\%$) of Sun-like stars host planets with masses above $10\,M_\oplus$ ($0.3\,M_{\rm J}$), about 30\% (40\%) of which are multi-planet systems; in terms of the RV semi-amplitude $K$, $33 \pm 7\%$ ($25 \pm 3\%$) of Sun-like stars contain planets of $K>1\,$m/s ($3\,$m/s), and each system hosts on average $1.8 \pm 0.4$ ($1.63 \pm 0.16$) planets. We note that the hot Jupiter rate in the CLS Sun-like sample is higher than the consensus value of $\sim$1\% by a factor of about three. We also confirm previous studies on the correlation between inner ($<1\,$au) and outer ($>1\,$au) planets.

astro-ph.EP↗

arXiv2022

Federated Learning of Molecular Properties with Graph Neural Networks in a Heterogeneous Setting

Chemistry research has both high material and computational costs to conduct experiments. Institutions thus consider chemical data to be valuable and there have been few efforts to construct large public datasets for machine learning. Another challenge is that different intuitions are interested in different classes of molecules, creating heterogeneous data that cannot be easily joined by conventional distributed training. In this work, we introduce federated heterogeneous molecular learning to address these challenges. Federated learning allows end-users to build a global model collaboratively while keeping the training data distributed over isolated clients. Due to the lack of related research, we first simulate a heterogeneous federated learning benchmark (FedChem) by jointly performing scaffold splitting and latent Dirichlet allocation on existing datasets for heterogeneously distributed client data. Our results on FedChem show that significant learning challenges arise when working with heterogeneous molecules across clients. We then propose a method to alleviate the problem, namely Federated Learning by Instance reweighTing (FLIT(+)). FLIT(+) can align the local training across heterogeneous clients by improving the performance for uncertain samples. Comprehensive experiments conducted on our new benchmark FedChem validate the advantages of this method over other federated learning schemes. FedChem should enable a new type of collaboration for improving AI in chemistry that mitigates concerns about valuable chemical data.

cs.LG↗

arXiv2022

Deep Federated Anomaly Detection for Multivariate Time Series Data

Despite the fact that many anomaly detection approaches have been developed for multivariate time series data, limited effort has been made on federated settings in which multivariate time series data are heterogeneously distributed among different edge devices while data sharing is prohibited. In this paper, we investigate the problem of federated unsupervised anomaly detection and present a Federated Exemplar-based Deep Neural Network (Fed-ExDNN) to conduct anomaly detection for multivariate time series data on different edge devices. Specifically, we first design an Exemplar-based Deep Neural network (ExDNN) to learn local time series representations based on their compatibility with an exemplar module which consists of hidden parameters learned to capture varieties of normal patterns on each edge device. Next, a constrained clustering mechanism (FedCC) is employed on the centralized server to align and aggregate the parameters of different local exemplar modules to obtain a unified global exemplar module. Finally, the global exemplar module is deployed together with a shared feature encoder to each edge device and anomaly detection is conducted by examining the compatibility of testing data to the exemplar module. Fed-ExDNN captures local normal time series patterns with ExDNN and aggregates these patterns by FedCC, and thus can handle the heterogeneous data distributed over different edge devices simultaneously. Thoroughly empirical studies on six public datasets show that ExDNN and Fed-ExDNN can outperform state-of-the-art anomaly detection algorithms and federated learning techniques.

cs.LG↗

arXiv2022

Localized Adversarial Domain Generalization

Deep learning methods can struggle to handle domain shifts not seen in training data, which can cause them to not generalize well to unseen domains. This has led to research attention on domain generalization (DG), which aims to the model's generalization ability to out-of-distribution. Adversarial domain generalization is a popular approach to DG, but conventional approaches (1) struggle to sufficiently align features so that local neighborhoods are mixed across domains; and (2) can suffer from feature space over collapse which can threaten generalization performance. To address these limitations, we propose localized adversarial domain generalization with space compactness maintenance~(LADG) which constitutes two major contributions. First, we propose an adversarial localized classifier as the domain discriminator, along with a principled primary branch. This constructs a min-max game whereby the aim of the featurizer is to produce locally mixed domains. Second, we propose to use a coding-rate loss to alleviate feature space over collapse. We conduct comprehensive experiments on the Wilds DG benchmark to validate our approach, where LADG outperforms leading competitors on most datasets.

cs.LG↗

arXiv2022

A Simple Hash-Based Early Exiting Approach For Language Understanding and Generation

Early exiting allows instances to exit at different layers according to the estimation of difficulty. Previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty, which suffers from generalization and threshold-tuning. In contrast, learning to exit, or learning to predict instance difficulty is a more appealing way. Though some effort has been devoted to employing such "learn-to-exit" modules, it is still unknown whether and how well the instance difficulty can be learned. As a response, we first conduct experiments on the learnability of instance difficulty, which demonstrates that modern neural models perform poorly on predicting instance difficulty. Based on this observation, we propose a simple-yet-effective Hash-based Early Exiting approach (HashEE) that replaces the learn-to-exit modules with hash functions to assign each token to a fixed exiting layer. Different from previous methods, HashEE requires no internal classifiers nor extra parameters, and therefore is more efficient. Experimental results on classification, regression, and generation tasks demonstrate that HashEE can achieve higher performance with fewer FLOPs and inference time compared with previous state-of-the-art early exiting methods.

cs.CL↗

arXiv2022

Fluid Simulation System Based on Graph Neural Network

Traditional computational fluid dynamics calculates the physical information of the flow field by solving partial differential equations, which takes a long time to calculate and consumes a lot of computational resources. We build a fluid simulation simulator based on the graph neural network architecture. The simulator has fast computing speed and low consumption of computing resources. We regard the computational domain as a structural graph, and the computational nodes in the structural graph determine neighbor nodes through adaptive sampling. Building deep learning architectures with attention graph neural networks. The fluid simulation simulator is trained according to the simulation results of the flow field around the cylinder with different Reynolds numbers. The trained fluid simulation simulator not only has a very high accuracy for the prediction of the flow field in the training set, but also can extrapolate the flow field outside the training set. Compared to traditional CFD solvers, the fluid simulation simulator achieves a speedup of 2-3 orders of magnitude. The fluid simulation simulator provides new ideas for the rapid optimization and design of fluid mechanics models and the real-time control of intelligent fluid mechanisms.

physics.flu-dyn↗

arXiv2022

AlphaFold Accelerates Artificial Intelligence Powered Drug Discovery: Efficient Discovery of a Novel Cyclin-dependent Kinase 20 (CDK20) Small Molecule Inhibitor

The AlphaFold computer program predicted protein structures for the whole human genome, which has been considered as a remarkable breakthrough both in artificial intelligence (AI) application and structural biology. Despite the varying confidence level, these predicted structures still could significantly contribute to structure-based drug design of novel targets, especially the ones with no or limited structural information. In this work, we successfully applied AlphaFold in our end-to-end AI-powered drug discovery engines constituted of a biocomputational platform PandaOmics and a generative chemistry platform Chemistry42, to identify a first-in-class hit molecule of a novel target without an experimental structure starting from target selection towards hit identification in a cost- and time-efficient manner. PandaOmics provided the targets of interest and Chemistry42 generated the molecules based on the AlphaFold predicted structure, and the selected molecules were synthesized and tested in biological assays. Through this approach, we identified a small molecule hit compound for CDK20 with a Kd value of 8.9 +/- 1.6 uM (n = 4) within 30 days from target selection and after only synthesizing 7 compounds. Based on the available data, the second round of AI-powered compound generation was conducted and through which, a more potent hit molecule, ISM042-2 048, was discovered with a Kd value of 210.0 +/- 42.4 nM (n = 2), within 30 days and after synthesizing 6 compounds from the discovery of the first hit ISM042-2-001. To the best of our knowledge, this is the first reported small molecule targeting CDK20 and more importantly, this work is the first demonstration of AlphaFold application in the hit identification process in early drug discovery.

q-bio.BM↗

arXiv2022

Scaling-Translation-Equivariant Networks with Decomposed Convolutional Filters

Encoding the scale information explicitly into the representation learned by a convolutional neural network (CNN) is beneficial for many computer vision tasks especially when dealing with multiscale inputs. We study, in this paper, a scaling-translation-equivariant (ST-equivariant) CNN with joint convolutions across the space and the scaling group, which is shown to be both sufficient and necessary to achieve equivariance for the regular representation of the scaling-translation group ST . To reduce the model complexity and computational burden, we decompose the convolutional filters under two pre-fixed separable bases and truncate the expansion to low-frequency components. A further benefit of the truncated filter expansion is the improved deformation robustness of the equivariant representation, a property which is theoretically analyzed and empirically verified. Numerical experiments demonstrate that the proposed scaling-translation-equivariant network with decomposed convolutional filters (ScDCFNet) achieves significantly improved performance in multiscale image classification and better interpretability than regular CNNs at a reduced model size.

cs.LG↗

arXiv2022

No Significant Correlation between Line-emission and Continuum Substructures in the Molecules with ALMA at Planet-forming Scales Program

Recently, the Molecules with ALMA at Planet-forming Scales (MAPS) ALMA Large Program reported a high number of line emission substructures coincident with dust rings and gaps in the continuum emission, suggesting a causal link between these axisymmetric line emission and dust continuum substructures. To test the robustness of the claimed correlation, we compare the observed spatial overlap fraction in substructures with that from the null hypothesis, in which the overlap is assumed to arise from the random placement of line emission substructures. Our results reveal that there is no statistically significant evidence for a universal correlation between line emission and continuum substructures, questioning the frequently-made link between continuum rings and pressure bumps. The analysis also clearly identifies outliers. The chemical rings and the dust gaps in MWC 480 appear to be strongly correlated (${>}4σ$), and the gaps in the CO isotopologues tend to moderately (${\sim}3σ$) correlate with dust rings.

astro-ph.EP↗

arXiv2022

Learning to Aggregate and Refine Noisy Labels for Visual Sentiment Analysis

Visual sentiment analysis has received increasing attention in recent years. However, the dataset's quality is a concern because the sentiment labels are crowd-sourcing, subjective, and prone to mistakes, and poses a severe threat to the data-driven models, especially the deep neural networks. The deep models would generalize poorly on the testing cases when trained to over-fit the training samples with noisy sentiment labels. Inspired by the recent progress on learning with noisy labels, we propose a robust learning method to perform robust visual sentiment analysis. Our method relies on external memory to aggregate and filters noisy labels during training. The memory is composed of the prototypes with corresponding labels, which can be updated online. The learned prototypes and their labels can be regarded as denoising features and labels for the local regions and can guide the training process to prevent the model from overfitting the noisy cases. We establish a benchmark for visual sentiment analysis with label noise using publicly available datasets. The experiment results of the proposed benchmark settings comprehensively show the effectiveness of our method.

cs.CV↗

arXiv2022

OGLE-2016-BLG-1093Lb: A Sub-Jupiter-mass Spitzer Planet Located in Galactic Bulge

OGLE-2016-BLG-1093 is a planetary microlensing event that is part of the statistical $Spitzer$ microlens parallax sample. The precise measurement of the microlens parallax effect for this event, combined with the measurement of finite source effects, leads to a direct measurement of the lens masses and system distance: $M_{\rm host} = 0.38$--$0.57\, M_{\odot}$, $m_p = 0.59$--$0.87\, M_{\rm Jup}$, and the system is located at the Galactic bulge ($D_L \sim 8.1$ kpc). Because this was a high-magnification event, we are also able to empirically show that the "cheap-space parallax" concept Gould & Yee (2012) produces well-constrained (and consistent) results for $|π_{\rm E}|$. This demonstrates that this concept can be extended to many two-body lenses. Finally, we briefly explore systematics in the $Spitzer$ light curve in this event and show that their potential impact is strongly mitigated by the color-constraint.

astro-ph.EP↗

arXiv2021

MagAO observations of the binary microlens OGLE-2014-BLG-1050 prefer the higher-mass solution

We report adaptive-optics (AO) follow-up imaging of OGLE-2014-BLG-1050, which is the second binary microlensing event with space-based parallax measurements. The degeneracy in microlens parallax pi_E led to two sets of solutions, either a ~(0.9, 0.35) M_Sun binary at ~3.5 kpc, or a ~(0.2, 0.07) M_Sun binary at ~1.1 kpc. We measure the flux blended with the microlensed source by conducting Magellan AO observations, and find that the blending is consistent with the predicted lens flux from the higher-mass solution. From the combination of the AO flux measurement together with previous lensing constraints, it is estimated that} the lens system consists of a $1.05^{+0.08}_{-0.07}$ M_Sun primary and a $0.38^{+0.07}_{-0.06}$ M_Sun secondary at $3.43^{+0.19}_{-0.21}$ kpc.

astro-ph.SR↗

arXiv2021

Systematic KMTNet Planetary Anomaly Search, Paper III: One Wide-Orbit Planet And Two Stellar Binaries

Only a few wide-orbit planets around old stars have been detected, which limits our statistical understanding of this planet population. Following the systematic search for planetary anomalies in microlensing events found by the Korea Microlensing Telescope Network (KMTNet), we present the discovery and analysis of three events that were initially thought to contain wide-orbit planets. The anomalous feature in the light curve of OGLE-2018-BLG-0383 is caused by a planet with mass ratio $q=2.1\times 10^{-4}$ and a projected separation $s=2.45$. This makes it the lowest mass-ratio microlensing planet at such wide orbits. The other two events, KMT-2018-BLG-0998 and OGLE-2018-BLG-0271, are shown to be stellar binaries ($q>0.1$) with rather close ($s<1$) separations. We briefly discuss the properties of known wide-orbit microlensing planets and show that the survey observations are crucial in discovering and further statistically constraining such a planet population.

astro-ph.EP↗

arXiv2021

Deformation Robust Roto-Scale-Translation Equivariant CNNs

Incorporating group symmetry directly into the learning process has proved to be an effective guideline for model design. By producing features that are guaranteed to transform covariantly to the group actions on the inputs, group-equivariant convolutional neural networks (G-CNNs) achieve significantly improved generalization performance in learning tasks with intrinsic symmetry. General theory and practical implementation of G-CNNs have been studied for planar images under either rotation or scaling transformation, but only individually. We present, in this paper, a roto-scale-translation equivariant CNN (RST-CNN), that is guaranteed to achieve equivariance jointly over these three groups via coupled group convolutions. Moreover, as symmetry transformations in reality are rarely perfect and typically subject to input deformation, we provide a stability analysis of the equivariance of representation to input distortion, which motivates the truncated expansion of the convolutional filters under (pre-fixed) low-frequency spatial modes. The resulting model provably achieves deformation-robust RST equivariance, i.e., the RST symmetry is still "approximately" preserved when the transformation is "contaminated" by a nuisance data deformation, a property that is especially important for out-of-distribution generalization. Numerical experiments on MNIST, Fashion-MNIST, and STL-10 demonstrate that the proposed model yields remarkable gains over prior arts, especially in the small data regime where both rotation and scaling variations are present within the data.

cs.CV↗