arXiv subjects

Yao Liu

Publications and source records attributed to Yao Liu.

At least 127 records · Page 7Linked to original sources

arXiv2023

TD Convergence: An Optimization Perspective

We study the convergence behavior of the celebrated temporal-difference (TD) learning algorithm. By looking at the algorithm through the lens of optimization, we first argue that TD can be viewed as an iterative optimization algorithm where the function to be minimized changes per iteration. By carefully investigating the divergence displayed by TD on a classical counter example, we identify two forces that determine the convergent or divergent behavior of the algorithm. We next formalize our discovery in the linear TD setting with quadratic loss and prove that convergence of TD hinges on the interplay between these two forces. We extend this optimization perspective to prove convergence of TD in a much broader setting than just linear approximation and squared loss. Our results provide a theoretical explanation for the successful application of TD in reinforcement learning.

cs.LG↗

arXiv2023

Sciences with the 2.5-meter Wide Field Survey Telescope (WFST)

The Wide Field Survey Telescope (WFST) is a dedicated photometric surveying facility being built jointly by the University of Science and Technology of China and the Purple Mountain Observatory. It is equipped with a 2.5-meter diameter primary mirror, an active optics system, and a mosaic CCD camera with 0.73 gigapixels on the primary focal plane for high-quality image capture over an FOV of 6.5-square-degree. It is anticipated that WFST will be set up at the Lenghu site in the summer of 2023 and begin to observe the northern sky in four optical bands (u, g, r, and i) with a range of cadences, from hourly/daily in the Deep High-Cadence Survey (DHS) program to semiweekly in the Wide-Field Survey (WFS) program, three months later. During a photometric night, a nominal 30 s exposure in the WFS program will reach a depth of 22.27, 23.32, 22.84, and 22.31 (AB magnitudes) in these four bands, respectively, allowing for the detection of a tremendous amount of transients in the low-z universe and a systematic investigation of the variability of Galactic and extragalactic objects. In the DHS program, intranight 90 s exposures as deep as 23 (u) and 24 mag (g), in combination with target of opportunity follow-ups, will provide a unique opportunity to explore energetic transients in demand for high sensitivities, including the electromagnetic counterparts of gravitational wave events, supernovae within a few hours of their explosions, tidal disruption events and fast, luminous optical transients even beyond a redshift of unity. In addition, the final 6-year co-added images, anticipated to reach g=25.8 mag in WFS or 1.5 mags deeper in DHS, will be of fundamental importance to general Galactic and extragalactic science. The highly uniform legacy surveys of WFST will serve as an indispensable complement to those of LSST that monitor the southern sky.

astro-ph.IM↗

arXiv2023

Data-Driven Next-Generation Wireless Networking: Embracing AI for Performance and Security

New network architectures, such as the Internet of Things (IoT), 5G, and next-generation (NextG) cellular systems, put forward emerging challenges to the design of future wireless networks toward ultra-high data rate, massive data processing, smart designs, low-cost deployment, reliability and security in dynamic environments. As one of the most promising techniques today, artificial intelligence (AI) is advocated to enable a data-driven paradigm for wireless network design. In this paper, we are motivated to review existing AI techniques and their applications for the full wireless network protocol stack toward improving network performance and security. Our goal is to summarize the current motivation, challenges, and methodology of using AI to enhance wireless networking from the physical to the application layer, and shed light on creating new AI-enabled algorithms, mechanisms, protocols, and system designs for future data-driven wireless networking.

cs.NI↗

arXiv2023

Molecules with ALMA at Planet-forming Scales (MAPS). Complex Kinematics in the AS 209 Disk Induced by a Forming Planet and Disk Winds

We study the kinematics of the AS 209 disk using the J=2-1 transitions of $^{12}$CO, $^{13}$CO, and C$^{18}$O. We derive the radial, azimuthal, and vertical velocity of the gas, taking into account the lowered emission surface near the annular gap at ~1.7 (200 au) within which a candidate circumplanetary disk-hosting planet has been reported previously. In $^{12}$CO and $^{13}$CO, we find a coherent upward flow arising from the gap. The upward gas flow is as fast as $150~{\rm m~s}^{-1}$ in the regions traced by $^{12}$CO emission, which corresponds to about 50% of the local sound speed or $6\%$ of the local Keplerian speed. Such an upward gas flow is difficult to reconcile with an embedded planet alone. Instead, we propose that magnetically driven winds via ambipolar diffusion are triggered by the low gas density within the planet-carved gap, dominating the kinematics of the gap region. We estimate the ambipolar Elsasser number, Am, using the HCO$^+$ column density as a proxy for ion density and find that Am is ~0.1 at the radial location of the upward flow. This value is broadly consistent with the value at which numerical simulations find ambipolar diffusion drives strong winds. We hypothesize the activation of magnetically-driven winds in a planet-carved gap can control the growth of the embedded planet. We provide a scaling relationship which describes the wind-regulated terminal mass: adopting parameters relevant to 100 au from a solar-mass star, we find the wind-regulated terminal mass is about one Jupiter mass, which may help explain the dearth of directly imaged super-Jovian-mass planets.

astro-ph.EP↗

arXiv2023

Symbolic Discovery of Optimization Algorithms

We present a method to formulate algorithm discovery as program search, and apply it to discover optimization algorithms for deep neural network training. We leverage efficient search techniques to explore an infinite and sparse program space. To bridge the large generalization gap between proxy and target tasks, we also introduce program selection and simplification strategies. Our method discovers a simple and effective optimization algorithm, $\textbf{Lion}$ ($\textit{Evo$\textbf{L}$ved S$\textbf{i}$gn M$\textbf{o}$me$\textbf{n}$tum}$). It is more memory-efficient than Adam as it only keeps track of the momentum. Different from adaptive optimizers, its update has the same magnitude for each parameter calculated through the sign operation. We compare Lion with widely used optimizers, such as Adam and Adafactor, for training a variety of models on different tasks. On image classification, Lion boosts the accuracy of ViT by up to 2% on ImageNet and saves up to 5x the pre-training compute on JFT. On vision-language contrastive learning, we achieve 88.3% $\textit{zero-shot}$ and 91.1% $\textit{fine-tuning}$ accuracy on ImageNet, surpassing the previous best results by 2% and 0.1%, respectively. On diffusion models, Lion outperforms Adam by achieving a better FID score and reducing the training compute by up to 2.3x. For autoregressive, masked language modeling, and fine-tuning, Lion exhibits a similar or better performance compared to Adam. Our analysis of Lion reveals that its performance gain grows with the training batch size. It also requires a smaller learning rate than Adam due to the larger norm of the update produced by the sign function. Additionally, we examine the limitations of Lion and identify scenarios where its improvements are small or not statistically significant. Lion is also successfully deployed in production systems such as Google search ads CTR model.

cs.LG↗

arXiv2023

FAST discovery of long tidal tails in NGC 4490/85

We report the discovery of a 100 kpc HI tail in the merging galaxy pair NGC 4490/85 detected by the Five-Hundred-meter Aperture Spherical radio Telescope (FAST). The tidal tails extended in both the south and north directions, and they are much longer than that reported previously based on the VLA interferometric maps. The NGC 4490/85 is surrounded by a large gas envelope, and a starburst low metallicity dwarf galaxy MAPS 1231+42 is found to be connected with the gas envelope, indicating that galaxy interaction trigged the intense star formation in it. Based on the fact that the metallicity in MAPS 1231+42 is one order of magnitude lower than that in the two disks of NGC 4490 and NGC 4485, we speculate that the gas near this galaxy should be primordial and could be due to gas inflow from the circum-galactic medium (CGM). We also found a collimated gas component pointing at a nearby dwarf galaxy KK 149, suggesting that this galaxy might also be interacting with the NGC 4490 pair. We discuss the possible origin of the long tidal tails and the extended gas envelope in this merging system based on the new data from FAST.

astro-ph.GA↗

arXiv2023

Reinforcement Learning Tutor Better Supported Lower Performers in a Math Task

Resource limitations make it hard to provide all students with one of the most effective educational interventions: personalized instruction. Reinforcement learning could be a key tool to reduce the development cost and improve the effectiveness of intelligent tutoring software that aims to provide the right support, at the right time, to a student. Here we illustrate that deep reinforcement learning can be used to provide adaptive pedagogical support to students learning about the concept of volume in a narrative storyline software. Using explainable artificial intelligence tools, we extracted interpretable insights about the pedagogical policy learned and demonstrated that the resulting policy had similar performance in a different student population. Most importantly, in both studies, the reinforcement-learning narrative system had the largest benefit for those students with the lowest initial pretest scores, suggesting the opportunity for AI to adapt and provide support for those most in need.

cs.AI↗

arXiv2023

A Large Double-ring Disk around the Taurus M Dwarf J04124068+2438157

Planet formation imprints signatures on the physical structures of disks. In this paper, we present high-resolution ($\sim$50 mas, 8 au) Atacama Large Millimeter/submillimeter Array (ALMA) observations of 1.3 mm dust continuum and CO line emission toward the disk around the M3.5 star 2MASS J04124068+2438157. The dust disk consists only of two narrow rings at radial distances of 0.47 and 0.78 arcsec ($\sim$70 and 116 au), with Gaussian $σ$ widths of 5.6 and 8.5 au, respectively. The width of the outer ring is smaller than the estimated pressure scale height by $\sim25\%$, suggesting dust trapping in a radial pressure bump. The dust disk size, set by the location of the outermost ring, is significantly larger (by $3σ$) than other disks with similar millimeter luminosity, which can be explained by an early formation of local pressure bump to stop radial drift of millimeter dust grains. After considering the disk's physical structure and accretion properties, we prefer planet--disk interaction over dead zone or photoevaporation models to explain the observed dust disk morphology. We carry out high-contrast imaging at $L'$ band using Keck/NIRC2 to search for potential young planets, but do not identify any source above $5σ$. Within the dust gap between the two rings, we reach a contrast level of $\sim$7 mag, constraining the possible planet below $\sim$2--4 $M_{\rm Jup}$. Analyses of the gap/ring properties suggest a $\sim$Saturn mass planet at $\sim$90 au is likely responsible for the formation of the outer ring, which can be potentially revealed with JWST.

astro-ph.EP↗

arXiv2022

The Radial Profile of Dust Grain Size in the Protoplanetary Disk of DS Tau

How do dust grains in protoplanetary disks overcome rapid radial drift and grow from micron size particles to planets is not well understood. The key is to search for evidence of dust accumulation and growth as a function of radius in the disk. We investigate the radial profile of grain size in the DS Tau disk by fitting multi-band ALMA observations with self-consistent radiative transfer models. The best-fit grain sizes range from centimeters in the inner disk down to 30 micron in the outer regions. Such an inside-out decreasing tendency is consistent with theories of dust evolution. Based on the best-fit model, we find that dust of 2 Jupiter masses has been depleted within the gap. By taking the gas-to-dust mass ratio into account, the lost mass is enough to form the 3.5 Jupiter mass planet inferred by literature hydrodynamic simulations. Moreover, our modeling also indicates that at the interface region between the gap and the ring, the grain size profile shows a discontinuity, with its amplitude dependent on the dust model adopted in the radiative transfer analysis. Future multi-wavelength observations at higher angular resolutions are required to better constrain the grain size and its variation in the vicinity of disk substructures.

astro-ph.EP↗

arXiv2022

On the underestimation of dust mass in protoplanetary disks: Effects of disk structure and dust properties

The total amount of dust grains in protoplanetary disks is one of the key properties that characterize the potential for planet formation. With (sub-)millimeter flux measurements, literature studies usually derive the dust mass using an analytic form under the assumption of optically thin emission, which may lead to substantial underestimation. In this work, we conduct a parameter study with the goal of investigating the effects of disk structure and dust properties on the underestimation through self-consistent radiative transfer models. Different dust models, scattering modes and approaches for dust settling are considered and compared. The influences of disk substructures, such as rings and crescents, on the mass derivation are investigated as well. The results indicate that the traditional analytic method can underestimate the mass by a factor of a few to hundreds, depending on the optical depth along the line of sight set mainly by the true dust mass, disk size and inclination. As an application, we perform a detailed radiative transfer modeling of the spectral energy distribution of DoAr 33, one of the observed DSHARP disks. When the DSHARP dust opacities are adopted, the most probable dust mass returned from the Bayesian analysis is roughly 7 times higher than the value given by the analytic calculation. Our study demonstrates that estimating disk dust masses from radiative transfer modeling is one solution for alleviating the problem of insufficient mass for planet formation raised in the ALMA era.

astro-ph.EP↗

arXiv2022

Evolution of Galaxy Types and HI Gas in Hickson Compact Groups

Compact groups have high galaxy densities and low velocity dispersions, and their group members have experienced numerous and frequent interactions during their lifetimes. They provide a unique environment to study the evolution of galaxies. We examined the galaxies types and HI contents in groups to make a study on the galaxy evolution in compact groups. We used the group crossing time as an age indicator for galaxy groups. Our sample is derived from the Hickson Compact Group catalog. We obtained group morphology data from the Hyper-Leda database and the IR classification based on Wide-Field Infrared Survey Explorer (WISE) fluxes from Zucker et al. (2016). By cross-matching the latest released ALFALFA 100% HI source catalog and supplemented by data found in literature, we obtained 40 galaxy groups with HI data available. We confirmed that the weak correlation between HI mass fraction and group crossing time found by Ai & Zhu (2018) in SDSS groups also exists in compact groups. We also found that the group spiral galaxy fraction is correlated with the group crossing time, but the actively star-forming galaxy fraction is not correlated with the group crossing time. These results seem to fit with the hypothesis that the sequential acquisition of neighbors from surrounding larger-scale structures has affected the morphology transition and star formation efficiency in compact groups.

astro-ph.GA↗

arXiv2022

Millimeter Gap Contrast as a Probe for Turbulence Level in Protoplanetary Disks

Turbulent motions are believed to regulate angular momentum transport and influence dust evolution in protoplanetary disks. Measuring the strength of turbulence is challenging through gas line observations because of the requirement for high spatial and spectral resolution data, and an exquisite determination of the temperature. In this work, taking the well-known HD 163296 disk as an example, we investigated the contrast of gaps identified in high angular resolution continuum images as a probe for the level of turbulence. With self-consistent radiative transfer models, we simultaneously analyzed the radial brightness profiles along the disk major and minor axes, and the azimuthal brightness profiles of the B67 and B100 rings. By fitting all the gap contrasts measured from these profiles, we constrained the gas-to-dust scale height ratio $Λ$ to be $3.0_{-0.8}^{+0.3}$, $1.2_{-0.1}^{+0.1}$ and ${\ge}\,6.5$ for the D48, B67 and B100 regions, respectively. The varying gas-to-dust scale height ratios indicate that the degree of dust settling changes with radius. The inferred values for $Λ$ translate into a turbulence level of $α_{\rm turb}\,{<}\,3\times10^{-3}$ in the D48 and B100 regions, which is consistent with previous upper limits set by gas line observations. However, turbulent motions in the B67 ring are strong with $α_{\rm turb}\,{\sim}1.2\,{\times}\,10^{-2}$. Due to the degeneracy between $Λ$ and the depth of dust surface density drops, the turbulence strength in the D86 gap region is not constrained.

astro-ph.EP↗

arXiv2022

On the Convergence of Multi-Server Federated Learning with Overlapping Area

Multi-server Federated learning (FL) has been considered as a promising solution to address the limited communication resource problem of single-server FL. We consider a typical multi-server FL architecture, where the coverage areas of regional servers may overlap. The key point of this architecture is that the clients located in the overlapping areas update their local models based on the average model of all accessible regional models, which enables indirect model sharing among different regional servers. Due to the complicated network topology, the convergence analysis is much more challenging than single-server FL. In this paper, we firstly propose a novel MS-FedAvg algorithm for this multi-server FL architecture and analyze its convergence on non-iid datasets for general non-convex settings. Since the number of clients located in each regional server is much less than in single-server FL, the bandwidth of each client should be large enough to successfully communicate training models with the server, which indicates that full client participation can work in multi-server FL. Also, we provide the convergence analysis of the partial client participation scheme and develop a new biased partial participation strategy to further accelerate convergence. Our results indicate that the convergence results highly depend on the ratio of the number of clients in each area type to the total number of clients in all three strategies. The extensive experiments show remarkable performance and support our theoretical results.

cs.DC↗

arXiv2022

Perception-Aware Attack: Creating Adversarial Music via Reverse-Engineering Human Perception

Recently, adversarial machine learning attacks have posed serious security threats against practical audio signal classification systems, including speech recognition, speaker recognition, and music copyright detection. Previous studies have mainly focused on ensuring the effectiveness of attacking an audio signal classifier via creating a small noise-like perturbation on the original signal. It is still unclear if an attacker is able to create audio signal perturbations that can be well perceived by human beings in addition to its attack effectiveness. This is particularly important for music signals as they are carefully crafted with human-enjoyable audio characteristics. In this work, we formulate the adversarial attack against music signals as a new perception-aware attack framework, which integrates human study into adversarial attack design. Specifically, we conduct a human study to quantify the human perception with respect to a change of a music signal. We invite human participants to rate their perceived deviation based on pairs of original and perturbed music signals, and reverse-engineer the human perception process by regression analysis to predict the human-perceived deviation given a perturbed signal. The perception-aware attack is then formulated as an optimization problem that finds an optimal perturbation signal to minimize the prediction of perceived deviation from the regressed human perception model. We use the perception-aware framework to design a realistic adversarial music attack against YouTube's copyright detector. Experiments show that the perception-aware attack produces adversarial music with significantly better perceptual quality than prior work.

cs.SD↗

arXiv2022

Molecules with ALMA at Planet-forming Scales (MAPS). A Circumplanetary Disk Candidate in Molecular Line Emission in the AS 209 Disk

We report the discovery of a circumplanetary disk (CPD) candidate embedded in the circumstellar disk of the T Tauri star AS 209 at a radial distance of about 200 au (on-sky separation of 1."4 from the star at a position angle of $161^\circ$), isolated via $^{13}$CO $J=2-1$ emission. This is the first instance of CPD detection via gaseous emission capable of tracing the overall CPD mass. The CPD is spatially unresolved with a $117\times82$ mas beam and manifests as a point source in $^{13}$CO, indicating that its diameter is $\lesssim14$ au. The CPD is embedded within an annular gap in the circumstellar disk previously identified using $^{12}$CO and near-infrared scattered light observations, and is associated with localized velocity perturbations in $^{12}$CO. The coincidence of these features suggests that they have a common origin: an embedded giant planet. We use the $^{13}$CO intensity to constrain the CPD gas temperature and mass. We find that the CPD temperature is $\gtrsim35$ K, higher than the circumstellar disk temperature at the radial location of the CPD, 22 K, suggesting that heating sources localized to the CPD must be present. The CPD gas mass is $\gtrsim 0.095 M_{\rm Jup} \simeq 30 M_{\rm Earth}$ adopting a standard $^{13}$CO abundance. From the non-detection of millimeter continuum emission at the location of the CPD ($3σ$ flux density $\lesssim26.4~μ$Jy), we infer that the CPD dust mass is $\lesssim 0.027 M_{\rm Earth} \simeq 2.2$ lunar masses, indicating a low dust-to-gas mass ratio of $\lesssim9\times10^{-4}$. We discuss the formation mechanism of the CPD-hosting giant planet on a wide orbit in the framework of gravitational instability and pebble accretion.

astro-ph.EP↗

arXiv2022

Towards Adaptive Unknown Authentication for Universal Domain Adaptation by Classifier Paradox

Universal domain adaptation (UniDA) is a general unsupervised domain adaptation setting, which addresses both domain and label shifts in adaptation. Its main challenge lies in how to identify target samples in unshared or unknown classes. Previous methods commonly strive to depict sample "confidence" along with a threshold for rejecting unknowns, and align feature distributions of shared classes across domains. However, it is still hard to pre-specify a "confidence" criterion and threshold which are adaptive to various real tasks, and a mis-prediction of unknowns further incurs misalignment of features in shared classes. In this paper, we propose a new UniDA method with adaptive Unknown Authentication by Classifier Paradox (UACP), considering that samples with paradoxical predictions are probably unknowns belonging to none of the source classes. In UACP, a composite classifier is jointly designed with two types of predictors. That is, a multi-class (MC) predictor classifies samples to one of the multiple source classes, while a binary one-vs-all (OVA) predictor further verifies the prediction by MC predictor. Samples with verification failure or paradox are identified as unknowns. Further, instead of feature alignment for shared classes, implicit domain alignment is conducted in output space such that samples across domains share the same decision boundary, though with feature discrepancy. Empirical results validate UACP under both open-set and universal UDA settings.

cs.CV↗

arXiv2022

Offline Policy Optimization with Eligible Actions

Offline policy optimization could have a large impact on many real-world decision-making problems, as online learning may be infeasible in many applications. Importance sampling and its variants are a commonly used type of estimator in offline policy evaluation, and such estimators typically do not require assumptions on the properties and representational capabilities of value function or decision process model function classes. In this paper, we identify an important overfitting phenomenon in optimizing the importance weighted return, in which it may be possible for the learned policy to essentially avoid making aligned decisions for part of the initial state space. We propose an algorithm to avoid this overfitting through a new per-state-neighborhood normalization constraint, and provide a theoretical justification of the proposed algorithm. We also show the limitations of previous attempts to this approach. We test our algorithm in a healthcare-inspired simulator, a logged dataset collected from real hospitals and continuous control tasks. These experiments show the proposed method yields less overfitting and better test performance compared to state-of-the-art batch reinforcement learning algorithms.

cs.LG↗

arXiv2022

Gas Column Density Distribution of Molecular Clouds in the Third Quadrant of the Milky Way

We have obtained column density maps for an unbiased sample of 120 molecular clouds in the third quadrant of the Milky Way mid-plane (b$\le |5|^{\circ}$) within the galactic longitude range from 195$^{\circ}$ to 225$^{\circ}$, using the high sensitivity $^{12}$CO and $^{13}$CO ($J=1-0$) data from the Milky Way Imaging Scroll Painting (MWISP) project. The probability density functions of the molecular hydrogen column density of the clouds, N-PDFs, are fitted with both log-normal (LN) function and log-normal plus power-law (LN+PL) function. The molecular clouds are classified into three categories according to their shapes of N-PDFs, i.e., LN, LN+PL, and UN (unclear), respectively. About 72\% of the molecular clouds fall into the LN category, while 18\% and 10\% into the LN+PL and UN categories, respectively. A power-law scaling relation, $σ_s\propto N_{H_2}^{0.44}$, exists between the width of the N-PDF, $σ_s$, and the average column density, $N_{H_2}$, of the molecular clouds. However, $σ_s$ shows no correlation with the mass of the clouds. A correlation is found between the dispersion of normalized column density, $σ_{N/\rm }$, and the sonic Mach number, $\mathcal{M}$, of molecular clouds. Overall, as predicted by numerical simulations, the N-PDFs of the molecular clouds with active star formation activity tend to have N-PDFs with power-law high-density tails.

astro-ph.GA↗