arXiv subjects

Kai Xiao

Publications and source records attributed to Kai Xiao.

At least 73 records · Page 4Linked to original sources

arXiv2023

When the atoms dance: exploring mechanisms of electron-beam induced modifications of materials with machine-learning assisted high temporal resolution electron microscopy

Directed atomic fabrication using an aberration-corrected scanning transmission electron microscope (STEM) opens new pathways for atomic engineering of functional materials. In this approach, the electron beam is used to actively alter the atomic structure through electron beam induced irradiation processes. One of the impediments that has limited widespread use thus far has been the ability to understand the fundamental mechanisms of atomic transformation pathways at high spatiotemporal resolution. Here, we develop a workflow for obtaining and analyzing high-speed spiral scan STEM data, up to 100 fps, to track the atomic fabrication process during nanopore milling in monolayer MoS2. An automated feedback-controlled electron beam positioning system combined with deep convolution neural network (DCNN) was used to decipher fast but low signal-to-noise datasets and classify time-resolved atom positions and nature of their evolving atomic defect configurations. Through this automated decoding, the initial atomic disordering and reordering processes leading to nanopore formation was able to be studied across various timescales. Using these experimental workflows a greater degree of speed and information can be extracted from small datasets without compromising spatial resolution. This approach can be adapted to other 2D materials systems to gain further insights into the defect formation necessary to inform future automated fabrication techniques utilizing the STEM electron beam.

cond-mat.mtrl-sci↗

arXiv2023

S-PLUS: Photometric Re-calibration with the Stellar Color Regression Method and an Improved Gaia XP Synthetic Photometry Method

We present a comprehensive re-calibration of medium- and broad-band photometry from the Southern Photometric Local Universe Survey (S-PLUS) by leveraging two approaches: an improved Gaia XP Synthetic Photometry (XPSP) method with corrected Gaia XP spectra, the Stellar Color Regression (SCR) method with corrected Gaia EDR3 photometric data and spectroscopic data from LAMOST DR7. Through the use of millions of stars as standards per band, we demonstrate the existence of position-dependent systematic errors, up to 23 mmag for the Main Survey region, in the S-PLUS DR4 photometric data. A comparison between the XPSP and SCR methods reveals minor differences in zero-point offsets, typically within the range of 1 to 6 mmag, indicating the accuracy of the re-calibration, and a two- to three-fold improvement in the zero-point precision. During this process, we also verified and corrected for the systematic errors related to CCD position. The corrected S-PLUS DR4 photometric data will provide a solid data foundation for conducting scientific research that relies on high-calibration precision. Our results underscore the power of the XPSP method in combination with the SCR method, showcasing their effectiveness in enhancing calibration precision for wide-field surveys when combined with Gaia photometry and XP spectra, to be applied for other S-PLUS sub-surveys.

astro-ph.IM↗

arXiv2023

Autonomous synthesis of thin film materials with pulsed laser deposition enabled by in situ spectroscopy and automation

Synthesis of thin films has traditionally relied upon slow, sequential processes carried out with substantial human intervention, frequently utilizing a mix of experience and serendipity to optimize material structure and properties. With recent advances in autonomous systems which combine synthesis, characterization, and decision making with artificial intelligence (AI), large parameter spaces can be explored autonomously at rates beyond what is possible by human experimentalists, greatly accelerating discovery, optimization, and understanding in materials synthesis which directly address the grand challenges in synthesis science. Here, we demonstrate autonomous synthesis of a contemporary 2D material by combining the highly versatile pulsed laser deposition (PLD) technique with automation and machine learning (ML). We incorporated in situ and real-time spectroscopy, a high-throughput methodology, and cloud connectivity to enable autonomous synthesis workflows with PLD. Ultrathin WSe2 films were grown using co-ablation of two targets and showed a 10x increase in throughput over traditional PLD workflows. Gaussian process regression and Bayesian optimization were used with in situ Raman spectroscopy to autonomously discover two distinct growth windows and the process-property relationship after sampling only 0.25% of a large 4D parameter space. Any material that can be grown with PLD could be autonomously synthesized with our platform and workflows, enabling accelerated discovery and optimization of a vast number of materials.

cond-mat.mtrl-sci↗

arXiv2023

Improvement of the Pan-STARRS Photometric Calibration with LAMOST and Gaia

In this work, we perform the re-calibration of PS1 photometry by correcting for position-dependent systematic errors using the spectroscopy-based Stellar Color Regression method (SCR), the photometry-based SCR method (SCR$'$), and the Gaia XP synthetic photometry method (XPSP). We confirm the significant large-scale and small-scale spatial variation of magnitude offsets for all the $grizy$ filters. We show that the PS1 photometric calibration precisions in the $grizy$ filters are around 5--7\,mmag when averaged over 14$'$ regions. We note a much larger calibration error up to 0.04 mag in the Galactic plane, which is probably caused by the systematic errors of the PS1 magnitudes in crowded fields. The results of the three methods are consistent with each other within 1--2\,mmag or better for all the filters. We provide two-dimensional maps and a python package ({\url{https://doi.org/10.12149/101283}}) to correct for position-dependent magnitude offsets of PS1, which can be used for high-precision investigations and as a reference to calibrate other surveys.

astro-ph.IM↗

arXiv2023

Photometric calibration of the Stellar Abundance and Galactic Evolution Survey (SAGES): Nanshan One-meter Wide-field Telescope g, r, and i band imaging data

In this paper, a total of approximately 2.6 million dwarfs were constructed as standard stars, with an accuracy of about 0.01-0.02 mag for each band, by combining spectroscopic data from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope Data Release 7, photometric data from the corrected Gaia Early Data Release 3, and photometric metallicities. Using the spectroscopy based stellar color regression method (SCR method) and the photometric-based SCR method (SCR' method), we performed the relative calibration of the Nanshan One-meter Wide-field Telescope imaging data. Based on the corrected Pan-STARRS DR1 photometry, the absolute calibration was also performed. In the photometric calibration process, we analyzed the dependence of the calibration zero points on different images (observation time), different gates of the CCD detector, and different CCD positions. We found that the stellar flat and the relative gain between different gates depend on time. The amplitude of gain variation in three channels is approximately 0.5%-0.7% relative to the other channel, with a maximum value of 4%. In addition, significant spatial variations of the stellar flat fitting residual are found and corrected. Using repeated sources in the adjacent images, we checked and discovered internal consistency of about 1-2 mmag in all the filters. Using the PS1 magnitudes synthesized by Gaia DR3 BP/RP spectra by the synthetic photometry method, we found that the photometric calibration uniformity is about 1-2 mmag for all the bands, at a spatial resolution of 1.3 degree. A detailed comparison between the spectroscopy-based SCR and photometric-based SCR method magnitude offsets was performed, and we achieved an internal consistency precision of about 2 mmag or better with resolutions of 1.3 degree for all the filters. Which is mainly from the position-dependent errors of the E(B-V) used in SCR' method.

astro-ph.SR↗

arXiv2023

Quantifying and Defending against Privacy Threats on Federated Knowledge Graph Embedding

Knowledge Graph Embedding (KGE) is a fundamental technique that extracts expressive representation from knowledge graph (KG) to facilitate diverse downstream tasks. The emerging federated KGE (FKGE) collaboratively trains from distributed KGs held among clients while avoiding exchanging clients' sensitive raw KGs, which can still suffer from privacy threats as evidenced in other federated model trainings (e.g., neural networks). However, quantifying and defending against such privacy threats remain unexplored for FKGE which possesses unique properties not shared by previously studied models. In this paper, we conduct the first holistic study of the privacy threat on FKGE from both attack and defense perspectives. For the attack, we quantify the privacy threat by proposing three new inference attacks, which reveal substantial privacy risk by successfully inferring the existence of the KG triple from victim clients. For the defense, we propose DP-Flames, a novel differentially private FKGE with private selection, which offers a better privacy-utility tradeoff by exploiting the entity-binding sparse gradient property of FKGE and comes with a tight privacy accountant by incorporating the state-of-the-art private selection technique. We further propose an adaptive privacy budget allocation policy to dynamically adjust defense magnitude across the training procedure. Comprehensive evaluations demonstrate that the proposed defense can successfully mitigate the privacy threat by effectively reducing the success rate of inference attacks from $83.1\%$ to $59.4\%$ on average with only a modest utility decrease.

cs.CR↗

arXiv2022

Magneto-thermomechanically triggered active mechanical metamaterials -- untethered, reversible, reprogrammable transformations with shape locking

Future active metamaterials for reconfigurable structural applications require fast, untethered, reversible, and reprogrammable (multimodal) transformability with shape locking. Herein, we aim to construct and demonstrate a magneto-thermomechanical tool that enables a single material system to transform with untethered, reversible, low-powered reprogrammable deformations and shape locking via the application of magneto-thermomechanically triggered prestress on a shape memory polymer and structural instability with asymmetric magnetic torque. We demonstrate the mutual assistance of two physics concepts - magnetic control combined with the thermomechanical behavior of shape memory polymers, without requiring new materials synthesis and high-power energy for reprogramming. Our approach can open a new path of active metamaterials, flexible yet stiff soft robots, and multimodal morphing structures, where we can design them in reversible and reprogrammable ways.

cond-mat.mtrl-sci↗

arXiv2022

Photometric calibration methods for wide-field photometric surveys

Uniform and accurate photometric calibration plays an important role in the current and next-generation wide-field imaging surveys. Herein, we review the modern photometric calibration methods, including the classic standard star method, "hardware/observation-driven" methods (such as the Ubercalibration, Hypercalibration, and Forward Global Calibration Methods), and "software/physics-driven" methods (e.g., the Stellar Locus Regression, Stellar Locus, and Stellar Color Regression Methods). Further, we discuss their advantages, limitations, and future developments toward millimagnitude precision calibration.

astro-ph.IM↗

arXiv2022

Volumetric-mapping-based inverse design of 3D architected materials and mobility control by topology reconstruction

The recent development of modular origami structures has ushered in a new era for active metamaterials with multiple degrees of freedom (multi-DOF). Notably, no systematic inverse design approach for volumetric modular origami structures has been reported. Moreover, very few topologies of modular origami have been studied for the design of active metamaterials with multi-DOF. Herein, we develop an inverse design method and reconfigurable algorithm for constructing 3D active architected structures - we synthesize modular origami structures that can be volumetrically mapped to a target 3D shape. We can control the reconfigurability by reconstructing the topology of the architected structures. Our inverse design based on volumetric mapping with mobility control by topology reconstruction can be used to construct architected metamaterials with any 3D complex shape that are also transformable with multi-DOF. Our work opens a new path toward 3D reconfigurable structures based on volumetric inverse design. This work is significant for the design of 3D active metamaterials and 3D morphing devices for automotive, aerospace, and biomedical engineering applications.

cond-mat.mtrl-sci↗

arXiv2022

Validation and Improvement of the Pan-STARRS Photometric Calibration with the Stellar Color Regression Method

As one of the best ground-based photometric dataset, Pan-STARRS1 (PS1) has been widely used as the reference to calibrate other surveys. In this work, we present an independent validation and re-calibration of the PS1 photometry using spectroscopic data from the LAMOST DR7 and photometric data from the corrected Gaia EDR3 with the Stellar Color Regression (SCR) method. Using per band typically a total of 1.5 million LAMOST-PS1-Gaia stars as standards, we show that the PS1 photometric calibration precisions in the $grizy$ filters are around $4\sim 5$ mmag when averaged over $20'$ regions. However, significant large- and small-scale spatial variation of magnitude offset, up to over 1 per cent, probably caused by the calibration errors in the PS1, are found for all the $grizy$ filters. The calibration errors in different filters are un-correlated, and are slightly larger for the $g$ and $y$ filters. We also detect moderate magnitude-dependent errors (0.005, 0.005, 0.005, 0.004, 0.003 mag per magnitude in the 14 -- 17 magnitude range for the $grizy$ filters, respectively) in the PS1 photometry by comparing with the Gaia EDR3 and other catalogs. The errors are likely caused by the systematic uncertainties in the PSF magnitudes. We provide two-dimensional maps to correct for such magnitude offsets in the LAMOST footprint at different spatial resolutions from $20'$ to $160'$. The results demonstrate the power of the SCR method in improving the calibration precision of wide-field surveys when combined with the LAMOST spectroscopy and Gaia photometry.

astro-ph.SR↗

arXiv2021

On Distinctive Properties of Universal Perturbations

We identify properties of universal adversarial perturbations (UAPs) that distinguish them from standard adversarial perturbations. Specifically, we show that targeted UAPs generated by projected gradient descent exhibit two human-aligned properties: semantic locality and spatial invariance, which standard targeted adversarial perturbations lack. We also demonstrate that UAPs contain significantly less signal for generalization than standard adversarial perturbations -- that is, UAPs leverage non-robust features to a smaller extent than standard adversarial perturbations.

cs.LG↗

arXiv2021

The miniJPAS Survey: A Study on Wavelength Dependence of the Photon Response Non-uniformity of the JPAS-{\it Pathfinder} Camera

Understanding the origins of small-scale flats of CCDs and their wavelength-dependent variations plays an important role in high-precision photometric, astrometric, and shape measurements of astronomical objects. Based on the unique flat data of 47 narrow-band filters provided by JPAS-{\it Pathfinder}, we analyze the variations of small-scale flats as a function of wavelength. We find moderate variations (from about $1.0\%$ at 390 nm to $0.3\%$ at 890 nm) of small-scale flats among different filters, increasing towards shorter wavelengths. Small-scale flats of two filters close in central wavelengths are strongly correlated. We then use a simple physical model to reproduce the observed variations to a precision of about $\pm 0.14\%$, by considering the variations of charge collection efficiencies, effective areas and thicknesses between CCD pixels. We find that the wavelength-dependent variations of small-scale flats of the JPAS-{\it Pathfinder} camera originate from inhomogeneities of the quantum efficiency (particularly charge collection efficiency) as well as the effective area and thickness of CCD pixels. The former dominates the variations in short wavelengths while the latter two dominate at longer wavelengths. The effects on proper flat-fielding as well as on photometric/flux calibrations for photometric/slit-less spectroscopic surveys are discussed, particularly in blue filters/wavelengths. We also find that different model parameters are sensitive to flats of different wavelengths, depending on the relations between the electron absorption depth, the photon absorption length and the CCD thickness. In order to model the wavelength-dependent variations of small-scale flats, a small number (around ten) of small-scale flats with well-selected wavelengths are sufficient to reconstruct small-scale flats in other wavelengths.

astro-ph.IM↗

arXiv2021

SHORING: Design Provable Conditional High-Order Interaction Network via Symbolic Testing

Deep learning provides a promising way to extract effective representations from raw data in an end-to-end fashion and has proven its effectiveness in various domains such as computer vision, natural language processing, etc. However, in domains such as content/product recommendation and risk management, where sequence of event data is the most used raw data form and experts derived features are more commonly used, deep learning models struggle to dominate the game. In this paper, we propose a symbolic testing framework that helps to answer the question of what kinds of expert-derived features could be learned by a neural network. Inspired by this testing framework, we introduce an efficient architecture named SHORING, which contains two components: \textit{event network} and \textit{sequence network}. The \textit{event} network learns arbitrarily yet efficiently high-order \textit{event-level} embeddings via a provable reparameterization trick, the \textit{sequence} network aggregates from sequence of \textit{event-level} embeddings. We argue that SHORING is capable of learning certain standard symbolic expressions which the standard multi-head self-attention network fails to learn, and conduct comprehensive experiments and ablation studies on four synthetic datasets and three real-world datasets. The results show that SHORING empirically outperforms the state-of-the-art methods.

cs.LG↗

arXiv2021

3DB: A Framework for Debugging Computer Vision Models

We introduce 3DB: an extendable, unified framework for testing and debugging vision models using photorealistic simulation. We demonstrate, through a wide range of use cases, that 3DB allows users to discover vulnerabilities in computer vision systems and gain insights into how models make decisions. 3DB captures and generalizes many robustness analyses from prior work, and enables one to study their interplay. Finally, we find that the insights generated by the system transfer to the physical world. We are releasing 3DB as a library (https://github.com/3db/3db) alongside a set of example analyses, guides, and documentation: https://3db.github.io/3db/ .

cs.CV↗

arXiv2021

Magnetostriction of α-RuCl3 flakes in the zigzag phase

Motivated by the possibility of an intermediate U(1) quantum spin liquid phase in out-of-plane magnetic fields and enhanced magnetic fluctuations in exfoliated α-RuCl3 flakes, we study magneto-Raman spectra of exfoliated multilayer α-RuCl3 in out-of-plane magnetic fields of -6 T to 6 T at temperatures of 670 mK - 4 K. While the literature currently suggests that bulk α-RuCl3 is in an antiferromagnetic zigzag phase with R3bar symmetry at low temperature, we do not observe R3bar symmetry in exfoliated α-RuCl3 at low temperatures. While we saw no magnetic field driven transitions, the Raman modes exhibit unexpected stochastic shifts in response to applied magnetic field that are above the uncertainties inferred from Bayesian analysis. These stochastic shifts are consistent with the emergence of magnetostrictive interactions in exfoliated α-RuCl3.

cond-mat.str-el↗

arXiv2020

Noise or Signal: The Role of Image Backgrounds in Object Recognition

We assess the tendency of state-of-the-art object recognition models to depend on signals from image backgrounds. We create a toolkit for disentangling foreground and background signal on ImageNet images, and find that (a) models can achieve non-trivial accuracy by relying on the background alone, (b) models often misclassify images even in the presence of correctly classified foregrounds--up to 87.5% of the time with adversarially chosen backgrounds, and (c) more accurate models tend to depend on backgrounds less. Our analysis of backgrounds brings us closer to understanding which correlations machine learning models use, and how they determine models' out of distribution performance.

cs.CV↗

arXiv2019

Cookie Clicker

Cookie Clicker is a popular online incremental game where the goal of the game is to generate as many cookies as possible. In the game you start with an initial cookie generation rate, and you can use cookies as currency to purchase various items that increase your cookie generation rate. In this paper, we analyze strategies for playing Cookie Clicker optimally. While simple to state, the game gives rise to interesting analysis involving ideas from NP-hardness, approximation algorithms, and dynamic programming.

cs.GT↗

arXiv2019

Evaluating Robustness of Neural Networks with Mixed Integer Programming

Neural networks have demonstrated considerable success on a wide variety of real-world problems. However, networks trained only to optimize for training accuracy can often be fooled by adversarial examples - slightly perturbed inputs that are misclassified with high confidence. Verification of networks enables us to gauge their vulnerability to such adversarial examples. We formulate verification of piecewise-linear neural networks as a mixed integer program. On a representative task of finding minimum adversarial distortions, our verifier is two to three orders of magnitude quicker than the state-of-the-art. We achieve this computational speedup via tight formulations for non-linearities, as well as a novel presolve algorithm that makes full use of all information available. The computational speedup allows us to verify properties on convolutional networks with an order of magnitude more ReLUs than networks previously verified by any complete verifier. In particular, we determine for the first time the exact adversarial accuracy of an MNIST classifier to perturbations with bounded $l_\infty$ norm $ε=0.1$: for this classifier, we find an adversarial example for 4.38% of samples, and a certificate of robustness (to perturbations with bounded norm) for the remainder. Across all robust training procedures and network architectures considered, we are able to certify more samples than the state-of-the-art and find more adversarial examples than a strong first-order attack.

cs.LG↗