arXiv subjects

Bo Zhao

Publications and source records attributed to Bo Zhao.

At least 235 records · Page 13Linked to original sources

arXiv2022

DeepCore: A Comprehensive Library for Coreset Selection in Deep Learning

Coreset selection, which aims to select a subset of the most informative training samples, is a long-standing learning problem that can benefit many downstream tasks such as data-efficient learning, continual learning, neural architecture search, active learning, etc. However, many existing coreset selection methods are not designed for deep learning, which may have high complexity and poor generalization performance. In addition, the recently proposed methods are evaluated on models, datasets, and settings of different complexities. To advance the research of coreset selection in deep learning, we contribute a comprehensive code library, namely DeepCore, and provide an empirical study on popular coreset selection methods on CIFAR10 and ImageNet datasets. Extensive experiments on CIFAR10 and ImageNet datasets verify that, although various methods have advantages in certain experiment settings, random selection is still a strong baseline.

cs.LG↗

arXiv2022

One Password: An Encryption Scheme for Hiding Users' Register Information

In recent years, the attack which leverages register information (e.g. accounts and passwords) leaked from 3rd party applications to try other applications is popular and serious. We call this attack "database collision". Traditionally, people have to keep dozens of accounts and passwords for different applications to prevent this attack. In this paper, we propose a novel encryption scheme for hiding users' register information and preventing this attack. Specifically, we first hash the register information using existing safe hash function. Then the hash string is hidden, instead a coefficient vector is stored for verification. Coefficient vectors of the same register information are generated randomly for different applications. Hence, the original information is hardly cracked by dictionary based attack or database collision in practice. Using our encryption scheme, each user only needs to keep one password for dozens of applications.

cs.CR↗

arXiv2022

LIMO: Latent Inceptionism for Targeted Molecule Generation

Generation of drug-like molecules with high binding affinity to target proteins remains a difficult and resource-intensive task in drug discovery. Existing approaches primarily employ reinforcement learning, Markov sampling, or deep generative models guided by Gaussian processes, which can be prohibitively slow when generating molecules with high binding affinity calculated by computationally-expensive physics-based methods. We present Latent Inceptionism on Molecules (LIMO), which significantly accelerates molecule generation with an inceptionism-like technique. LIMO employs a variational autoencoder-generated latent space and property prediction by two neural networks in sequence to enable faster gradient-based reverse-optimization of molecular properties. Comprehensive experiments show that LIMO performs competitively on benchmark tasks and markedly outperforms state-of-the-art techniques on the novel task of generating drug-like compounds with high binding affinity, reaching nanomolar range against two protein targets. We corroborate these docking-based results with more accurate molecular dynamics-based calculations of absolute binding free energy and show that one of our generated drug-like compounds has a predicted $K_D$ (a measure of binding affinity) of $6 \cdot 10^{-14}$ M against the human estrogen receptor, well beyond the affinities of typical early-stage drug candidates and most FDA-approved drugs to their respective targets. Code is available at https://github.com/Rose-STL-Lab/LIMO.

cs.LG↗

arXiv2022

Tunable Multi-Peak Perfect Absorbers Based on Borophene for High-Performance Near-Infrared Refractive Index Sensing

Borophene has recently attracted significant attention as an emerging two-dimensional monoelemental material for refractive index sensing because of its ultra-high surface-to-volume ratio and outstanding surface sensitivity. However, current research mainly focuses on designing borophene-based sensors with single-peak absorption, which lacks reliability compared to the performance of multi-peak sensors. In this paper, high-performance borophene refractive index sensors with multiple nearly perfect absorption peaks are proposed. The geometric parameters of the metadevices are optimized using finite-difference time-domain (FDTD) simulations to obtain three strong absorption peaks in the near-infrared (near-IR) regime with intensities of 99.72%, 99.18%, and 98.13%. Further calculations show that the sensitivities of the three strong peaks reach 339.1 nm shift per refractive index units (nm/RIU), 410.1 nm/RIU, and 738.1 nm/RIU, and their corresponding figure of merits (FOMs) reach up to 15.41/RIU, 14.65/RIU and 10.85/RIU, respectively, exhibiting great potential for near-IR refractive index sensing. Moreover, the three absorption peaks can be easily tuned by adjusting the carrier density of borophene, which can be realized by applying different external electric bias voltages. These results can provide theoretical guidance for the development of multi-peak near-IR dynamic integrated photonic devices

physics.optics↗

arXiv2022

Privacy for Free: How does Dataset Condensation Help Privacy?

To prevent unintentional data leakage, research community has resorted to data generators that can produce differentially private data for model training. However, for the sake of the data privacy, existing solutions suffer from either expensive training cost or poor generalization performance. Therefore, we raise the question whether training efficiency and privacy can be achieved simultaneously. In this work, we for the first time identify that dataset condensation (DC) which is originally designed for improving training efficiency is also a better solution to replace the traditional data generators for private data generation, thus providing privacy for free. To demonstrate the privacy benefit of DC, we build a connection between DC and differential privacy, and theoretically prove on linear feature extractors (and then extended to non-linear feature extractors) that the existence of one sample has limited impact ($O(m/n)$) on the parameter distribution of networks trained on $m$ samples synthesized from $n (n \gg m)$ raw samples by DC. We also empirically validate the visual privacy and membership privacy of DC-synthesized data by launching both the loss-based and the state-of-the-art likelihood-based membership inference attacks. We envision this work as a milestone for data-efficient and privacy-preserving machine learning.

cs.CR↗

arXiv2022

Towards Understanding and Harnessing the Effect of Image Transformation in Adversarial Detection

Deep neural networks (DNNs) are threatened by adversarial examples. Adversarial detection, which distinguishes adversarial images from benign images, is fundamental for robust DNN-based services. Image transformation is one of the most effective approaches to detect adversarial examples. During the last few years, a variety of image transformations have been studied and discussed to design reliable adversarial detectors. In this paper, we systematically synthesize the recent progress on adversarial detection via image transformations with a novel classification method. Then, we conduct extensive experiments to test the detection performance of image transformations against state-of-the-art adversarial attacks. Furthermore, we reveal that each individual transformation is not capable of detecting adversarial examples in a robust way, and propose a DNN-based approach referred to as \emph{AdvJudge}, which combines scores of 9 image transformations. Without knowing which individual scores are misleading or not misleading, AdvJudge can make the right judgment, and achieve a significant improvement in detection rate. Finally, we utilize an explainable AI tool to show the contribution of each image transformation to adversarial detection. Experimental results show that the contribution of image transformations to adversarial detection is significantly different, the combination of them can significantly improve the generic detection ability against state-of-the-art adversarial attacks.

cs.CV↗

arXiv2022

FaceMAE: Privacy-Preserving Face Recognition via Masked Autoencoders

Face recognition, as one of the most successful applications in artificial intelligence, has been widely used in security, administration, advertising, and healthcare. However, the privacy issues of public face datasets have attracted increasing attention in recent years. Previous works simply mask most areas of faces or synthesize samples using generative models to construct privacy-preserving face datasets, which overlooks the trade-off between privacy protection and data utility. In this paper, we propose a novel framework FaceMAE, where the face privacy and recognition performance are considered simultaneously. Firstly, randomly masked face images are used to train the reconstruction module in FaceMAE. We tailor the instance relation matching (IRM) module to minimize the distribution gap between real faces and FaceMAE reconstructed ones. During the deployment phase, we use trained FaceMAE to reconstruct images from masked faces of unseen identities without extra training. The risk of privacy leakage is measured based on face retrieval between reconstructed and original datasets. Experiments prove that the identities of reconstructed images are difficult to be retrieved. We also perform sufficient privacy-preserving face recognition on several public face datasets (i.e. CASIA-WebFace and WebFace260M). Compared to previous state of the arts, FaceMAE consistently \textbf{reduces at least 50\% error rate} on LFW, CFP-FP and AgeDB.

cs.CV↗

arXiv2022

CAFE: Learning to Condense Dataset by Aligning Features

Dataset condensation aims at reducing the network training effort through condensing a cumbersome training set into a compact synthetic one. State-of-the-art approaches largely rely on learning the synthetic data by matching the gradients between the real and synthetic data batches. Despite the intuitive motivation and promising results, such gradient-based methods, by nature, easily overfit to a biased set of samples that produce dominant gradients, and thus lack global supervision of data distribution. In this paper, we propose a novel scheme to Condense dataset by Aligning FEatures (CAFE), which explicitly attempts to preserve the real-feature distribution as well as the discriminant power of the resulting synthetic set, lending itself to strong generalization capability to various architectures. At the heart of our approach is an effective strategy to align features from the real and synthetic data across various scales, while accounting for the classification of real samples. Our scheme is further backed up by a novel dynamic bi-level optimization, which adaptively adjusts parameter updates to prevent over-/under-fitting. We validate the proposed CAFE across various datasets, and demonstrate that it generally outperforms the state of the art: on the SVHN dataset, for example, the performance gain is up to 11%. Extensive experiments and analyses verify the effectiveness and necessity of proposed designs.

cs.CV↗

arXiv2022

Nonreciprocal Infrared Absorption via Resonant Magneto-optical Coupling to InAs

Nonreciprocal elements are a vital building block of electrical and optical systems. In the infrared regime, there is a particular interest in structures that break reciprocity because their thermal absorptive (and emissive) properties should not obey the Kirchhoff thermal radiation law. In this work, we break time-reversal symmetry and reciprocity in n-type doped magneto-optic InAs with a static magnetic field where light coupling is mediated by a guided-mode-resonator (GMR) structure whose resonant frequency coincides with the epsilon-near-zero (ENZ) resonance of the doped InAs. Using this structure, we observe the nonreciprocal absorptive behavior as a function of magnetic field and scattering angle in the infrared. Accounting for resonant and nonresonant optical scattering, we reliably model experimental results that break reciprocal absorption relations in the infrared. The ability to design such nonreciprocal absorbers opens an avenue to explore devices with unequal absorptivity and emissivity in specific channels.

physics.optics↗

arXiv2022

Resonant control of elastic collisions between $^{23}$Na$^{40}$K molecules and $^{40}$K atoms

We have demonstrated the resonant control of the elastic scattering cross sections in the vicinity of Feshbach resonances between $^{23}$Na$^{40}$K molecules and $^{40}$K atoms by studying the thermalization between them. The elastic scattering cross sections vary by more than two orders of magnitude close to the resonance, and can be well described by an asymmetric Fano profile. The parameters that characterize the magnetically tunable s-wave scattering length are determined from the elastic scattering cross sections. The observation of resonantly controlled elastic scattering cross sections opens up the possibility to study strongly interacting atom-molecule mixtures and improve our understanding of the complex atom-molecule Feshbach resonances.

physics.atom-ph↗

arXiv2022

The Central 1000 au of a Pre-stellar Core Revealed with ALMA. II. Almost Complete Freeze-out

Pre-stellar cores represent the initial conditions in the process of star and planet formation. Their low temperatures ($<$10 K) allow the formation of thick icy dust mantles, which will be partially preserved in the future protoplanetary disks, ultimately affecting the chemical composition of planetary systems. Previous observations have shown that carbon- and oxygen-bearing species, in particular CO, are heavily depleted in pre-stellar cores due to the efficient molecular freeze-out onto the surface of cold dust grains. However, N-bearing species such as NH$_3$ and, in particular, its deuterated isotopologues, appear to maintain high abundances where CO molecules are mainly in solid phase. Thanks to ALMA, we present here the first clear observational evidence of NH$_2$D freeze-out toward the L1544 pre-stellar core, suggestive of the presence of a"complete-depletion zone" within a $\simeq$1800 au radius, in agreement with astrochemical pre-stellar core model predictions. Our state-of-the-art chemical model coupled with a non-LTE radiative transfer code demonstrates that NH$_2$D becomes mainly incorporated in icy mantles in the central 2000 au and starts freezing-out already at $\simeq$7000 au. Radiative transfer effects within the pre-stellar core cause the NH$_2$D(1$_{11}$-1$_{01}$) emission to appear centrally concentrated, with a flattened distribution within the central $\simeq$3000 au, unlike the 1.3 mm dust continuum emission which shows a clear peak within the central $\simeq$1800 au. This prevented NH$_2$D freeze-out to be detected in previous observations, where the central 1000 au cannot be spatially resolved.

astro-ph.SR↗

arXiv2022

FAUST III. Misaligned rotations of the envelope, outflow, and disks in the multiple protostellar system of VLA 1623$-$2417

We report a study of the low-mass Class-0 multiple system VLA 1623AB in the Ophiuchus star-forming region, using H$^{13}$CO$^+$ ($J=3-2$), CS ($J=5-4$), and CCH ($N=3-2$) lines as part of the ALMA Large Program FAUST. The analysis of the velocity fields revealed the rotation motion in the envelope and the velocity gradients in the outflows (about 2000 au down to 50 au). We further investigated the rotation of the circum-binary VLA 1623A disk as well as the VLA 1623B disk. We found that the minor axis of the circum-binary disk of VLA 1623A is misaligned by about 12 degrees with respect to the large-scale outflow and the rotation axis of the envelope. In contrast, the minor axis of the circum-binary disk is parallel to the large-scale magnetic field according to previous dust polarization observations, suggesting that the misalignment may be caused by the different directions of the envelope rotation and the magnetic field. If the velocity gradient of the outflow is caused by rotation, the outflow has a constant angular momentum and the launching radius is estimated to be $5-16$ au, although it cannot be ruled out that the velocity gradient is driven by entrainments of the two high-velocity outflows. Furthermore, we detected for the first time a velocity gradient associated with rotation toward the VLA 16293B disk. The velocity gradient is opposite to the one from the large-scale envelope, outflow, and circum-binary disk. The origin of its opposite gradient is also discussed.

astro-ph.GA↗

arXiv2021

Makeup216: Logo Recognition with Adversarial Attention Representations

One of the challenges of logo recognition lies in the diversity of forms, such as symbols, texts or a combination of both; further, logos tend to be extremely concise in design while similar in appearance, suggesting the difficulty of learning discriminative representations. To investigate the variety and representation of logo, we introduced Makeup216, the largest and most complex logo dataset in the field of makeup, captured from the real world. It comprises of 216 logos and 157 brands, including 10,019 images and 37,018 annotated logo objects. In addition, we found that the marginal background around the pure logo can provide a important context information and proposed an adversarial attention representation framework (AAR) to attend on the logo subject and auxiliary marginal background separately, which can be combined for better representation. Our proposed framework achieved competitive results on Makeup216 and another large-scale open logo dataset, which could provide fresh thinking for logo recognition. The dataset of Makeup216 and the code of the proposed framework will be released soon.

cs.CV↗

arXiv2021

One-dimensional Deep Low-rank and Sparse Network for Accelerated MRI

Deep learning has shown astonishing performance in accelerated magnetic resonance imaging (MRI). Most state-of-the-art deep learning reconstructions adopt the powerful convolutional neural network and perform 2D convolution since many magnetic resonance images or their corresponding k-space are in 2D. In this work, we present a new approach that explores the 1D convolution, making the deep network much easier to be trained and generalized. We further integrate the 1D convolution into the proposed deep network, named as One-dimensional Deep Low-rank and Sparse network (ODLS), which unrolls the iteration procedure of a low-rank and sparse reconstruction model. Extensive results on in vivo knee and brain datasets demonstrate that, the proposed ODLS is very suitable for the case of limited training subjects and provides improved reconstruction performance than state-of-the-art methods both visually and quantitatively. Additionally, ODLS also shows nice robustness to different undersampling scenarios and some mismatches between the training and test data. In summary, our work demonstrates that the 1D deep learning scheme is memory-efficient and robust in fast MRI.

eess.IV↗

arXiv2021

Magnetic Feshbach resonances in collisions of $^{23}$Na$^{40}$K with $^{40}$K

We present measurements of more than 80 magnetic Feshbach resonances in collisions of ultracold $^{23}$Na$^{40}$K with $^{40}$K. We assign quantum numbers to a group of low-field resonances and show that they are due to long-range states of the triatomic complex in which the quantum numbers of the separated atom and molecule are approximately preserved. The resonant states are not members of chaotic bath of short-range states. Similar resonances are expected to be a common feature of alkali-metal diatom + atom systems.

physics.atom-ph↗

arXiv2021

GreedyFool: Multi-Factor Imperceptibility and Its Application to Designing a Black-box Adversarial Attack

Adversarial examples are well-designed input samples, in which perturbations are imperceptible to the human eyes, but easily mislead the output of deep neural networks (DNNs). Existing works synthesize adversarial examples by leveraging simple metrics to penalize perturbations, that lack sufficient consideration of the human visual system (HVS), which produces noticeable artifacts. To explore why the perturbations are visible, this paper summarizes four primary factors affecting the perceptibility of human eyes. Based on this investigation, we design a multi-factor metric MulFactorLoss for measuring the perceptual loss between benign examples and adversarial ones. In order to test the imperceptibility of the multi-factor metric, we propose a novel black-box adversarial attack that is referred to as GreedyFool. GreedyFool applies differential evolution to evaluate the effects of perturbed pixels on the confidence of a target DNN, and introduces greedy approximation to automatically generate adversarial perturbations. We conduct extensive experiments on the ImageNet and CIFRA-10 datasets and a comprehensive user study with 60 participants. The experimental results demonstrate that MulFactorLoss is a more imperceptible metric than the existing pixelwise metrics, and GreedyFool achieves a 100% success rate in a black-box manner.

cs.LG↗

arXiv2021

Feature-Filter: Detecting Adversarial Examples through Filtering off Recessive Features

Deep neural networks (DNNs) are under threat from adversarial example attacks. The adversary can easily change the outputs of DNNs by adding small well-designed perturbations to inputs. Adversarial example detection is a fundamental work for robust DNNs-based service. Adversarial examples show the difference between humans and DNNs in image recognition. From a human-centric perspective, image features could be divided into dominant features that are comprehensible to humans, and recessive features that are incomprehensible to humans, yet are exploited by DNNs. In this paper, we reveal that imperceptible adversarial examples are the product of recessive features misleading neural networks, and an adversarial attack is essentially a kind of method to enrich these recessive features in the image. The imperceptibility of the adversarial examples indicates that the perturbations enrich recessive features, yet hardly affect dominant features. Therefore, adversarial examples are sensitive to filtering off recessive features, while benign examples are immune to such operation. Inspired by this idea, we propose a label-only adversarial detection approach that is referred to as feature-filter. Feature-filter utilizes discrete cosine transform to approximately separate recessive features from dominant features, and gets a mutant image that is filtered off recessive features. By only comparing DNN's prediction labels on the input and its mutant, feature-filter can real-time detect imperceptible adversarial examples at high accuracy and few false positives.

cs.LG↗

arXiv2021

Accretion Flows or Outflow Cavities? Uncovering the Gas Dynamics around Lupus 3-MMS

Understanding how material accretes onto the rotationally supported disk from the surrounding envelope of gas and dust in the youngest protostellar systems is important for describing how disks are formed. Magnetohydrodynamic simulations of magnetized, turbulent disk formation usually show spiral-like streams of material (accretion flows) connecting the envelope to the disk. However, accretion flows in these early stages of protostellar formation still remain poorly characterized due to their low intensity and possibly some extended structures are disregarded as being part of the outflow cavity. We use ALMA archival data of a young Class 0 protostar, Lupus 3-MMS, to uncover four extended accretion flow-like structures in C$^{18}$O that follow the edges of the outflows. We make various types of position-velocity cuts to compare with the outflows and find the extended structures are not consistent with the outflow emission, but rather more consistent with a simple infall model. We then use a dendrogram algorithm to isolate five sub-structures in position-position-velocity space. Four out of the five sub-structures fit well ($>$95%) with our simple infall model, with specific angular momenta between $2.7-6.9\times10^{-4}\,$km$\,$s$^{-1}\,$pc and mass-infall rates of $0.5-1.1\times10^{-6}\,M_{\odot}\,$yr$^{-1}$. Better characterization of the physical structure in the supposed "outflow-cavities" is important to disentangle the true outflow cavities and accretion flows.

astro-ph.GA↗