Searcharxiv⌕ Search

arXiv subjects

He Zhao

Publications and source records attributed to He Zhao.

At least 109 records · Page 6Linked to original sources
arXiv2023

NPCL: Neural Processes for Uncertainty-Aware Continual Learning

Continual learning (CL) aims to train deep neural networks efficiently on streaming data while limiting the forgetting caused by new tasks. However, learning transferable knowledge with less interference between tasks is difficult, and real-world deployment of CL models is limited by their inability to measure predictive uncertainties. To address these issues, we propose handling CL tasks with neural processes (NPs), a class of meta-learners that encode different tasks into probabilistic distributions over functions all while providing reliable uncertainty estimates. Specifically, we propose an NP-based CL approach (NPCL) with task-specific modules arranged in a hierarchical latent variable model. We tailor regularizers on the learned latent distributions to alleviate forgetting. The uncertainty estimation capabilities of the NPCL can also be used to handle the task head/module inference challenge in CL. Our experiments show that the NPCL outperforms previous CL approaches. We validate the effectiveness of uncertainty estimation in the NPCL for identifying novel data and evaluating instance-level model confidence. Code is available at \url{https://github.com/srvCodes/NPCL}.

cs.LG↗
arXiv2023

Beyond Unimodal: Generalising Neural Processes for Multimodal Uncertainty Estimation

Uncertainty estimation is an important research area to make deep neural networks (DNNs) more trustworthy. While extensive research on uncertainty estimation has been conducted with unimodal data, uncertainty estimation for multimodal data remains a challenge. Neural processes (NPs) have been demonstrated to be an effective uncertainty estimation method for unimodal data by providing the reliability of Gaussian processes with efficient and powerful DNNs. While NPs hold significant potential for multimodal uncertainty estimation, the adaptation of NPs for multimodal data has not been carefully studied. To bridge this gap, we propose Multimodal Neural Processes (MNPs) by generalising NPs for multimodal uncertainty estimation. Based on the framework of NPs, MNPs consist of several novel and principled mechanisms tailored to the characteristics of multimodal data. In extensive empirical evaluation, our method achieves state-of-the-art multimodal uncertainty estimation performance, showing its appealing robustness against noisy samples and reliability in out-of-distribution detection with faster computation time compared to the current state-of-the-art multimodal uncertainty estimation method.

cs.LG↗
arXiv2023

Cross-adversarial local distribution regularization for semi-supervised medical image segmentation

Medical semi-supervised segmentation is a technique where a model is trained to segment objects of interest in medical images with limited annotated data. Existing semi-supervised segmentation methods are usually based on the smoothness assumption. This assumption implies that the model output distributions of two similar data samples are encouraged to be invariant. In other words, the smoothness assumption states that similar samples (e.g., adding small perturbations to an image) should have similar outputs. In this paper, we introduce a novel cross-adversarial local distribution (Cross-ALD) regularization to further enhance the smoothness assumption for semi-supervised medical image segmentation task. We conducted comprehensive experiments that the Cross-ALD archives state-of-the-art performance against many recent methods on the public LA and ACDC datasets.

eess.IV↗
arXiv2023

Unidirectional coherent quasiparticles in the high-temperature rotational symmetry broken phase of AV3Sb5 kagome superconductors

Kagome metals AV3Sb5 (where the A can stand for K, Cs, or Rb) display a rich phase diagram of correlated electron states, including superconductivity and density waves. Within this landscape, recent experiments revealed signs of a transition below approximately 35 K attributed to an electronic nematic phase that spontaneously breaks rotational symmetry of the lattice. Here, we show that rotational symmetry breaking initiates universally at a high temperature in these materials, toward the 2 x 2 charge density wave transition temperature. We do this via spectroscopic-imaging scanning tunneling microscopy and study atomic-scale signatures of electronic symmetry breaking across several materials in the AV3Sb5 family: CsV3Sb5, KV3Sb5 and Sn-doped CsV3Sb5. Below a significantly lower temperature of about 30 K, we measure quantum interference of quasiparticles, a key signature for the formation of a coherent electronic state. These quasiparticles display a pronounced unidirectional feature in reciprocal space that strengthens as the superconducting state is approached. Our experiments reveal that high-temperature rotation symmetry breaking and the charge ordering states are separated from the superconducting ground state by an intermediate-temperature regime with coherent unidirectional quasiparticles. This picture is phenomenologically different compared to that in high-temperature superconductors, shedding light on the complex nature of rotation symmetry breaking in AV3Sb5 kagome superconductors.

cond-mat.str-el↗
arXiv2023

Multiscale Memory Comparator Transformer for Few-Shot Video Segmentation

Few-shot video segmentation is the task of delineating a specific novel class in a query video using few labelled support images. Typical approaches compare support and query features while limiting comparisons to a single feature layer and thereby ignore potentially valuable information. We present a meta-learned Multiscale Memory Comparator (MMC) for few-shot video segmentation that combines information across scales within a transformer decoder. Typical multiscale transformer decoders for segmentation tasks learn a compressed representation, their queries, through information exchange across scales. Unlike previous work, we instead preserve the detailed feature maps during across scale information exchange via a multiscale memory transformer decoding to reduce confusion between the background and novel class. Integral to the approach, we investigate multiple forms of information exchange across scales in different tasks and provide insights with empirical evidence on which to use in each task. The overall comparisons among query and support features benefit from both rich semantics and precise localization. We demonstrate our approach primarily on few-shot video object segmentation and an adapted version on the fully supervised counterpart. In all cases, our approach outperforms the baseline and yields state-of-the-art performance. Our code is publicly available at https://github.com/MSiam/MMC-MultiscaleMemory.

cs.CV↗
arXiv2023

SPar: estimating stellar parameters from multi-band photometries with empirical stellar libraries

Modern large-scale photometric surveys have provided us with multi-band photometries of billions of stars. Determining the stellar atmospheric parameters, such as the effective temperature (\teff) and metallicities (\feh), absolute magnitudes ($M_{G}$), distances ($d$) and reddening values (\ebr) is fundamental to study the stellar populations, structure, kinematics and chemistry of the Galaxy. This work constructed an empirical stellar library which maps the stellar parameters to multi-band photometries from a dataset with Gaia parallaxes, LAMOST atmospheric parameters, and optical to near-infrared photometry from several photometric surveys. Based on the stellar library, we developed a new algorithm, SPar (\textbf{S}tellar \textbf{P}arameters from multib\textbf{a}nd photomet\textbf{r}y), which fits the multi-band stellar photometries to derive the stellar parameters (\teff, \feh, $M_G$, $d$ and \ebr) of the individual stars. The algorithm is applied to the multi-band photometric measurements of a sample of stars selected from the SMSS survey, which have stellar parameters derived from the spectroscopic surveys. The stellar parameters derived from multi-band photometries by our algorithm are in good agreement with those from the spectroscopic surveys. The typical differences between our results and the literature values are 170\,K for \teff, 0.23\,dex for \feh, 0.13\,mag for $M_G$ and 0.05\,mag for \ebr. The algorithm proved to be robust and effective and will be applied to the data of future large-scale photometric surveys such as the Mephisto and CSST surveys.

astro-ph.SR↗
arXiv2023

Smectic Pair Density Wave Order in EuRbFe4As4

The pair density wave (PDW) is a novel superconducting state in which Cooper pairs carry center-of-mass momentum in equilibrium, leading to the breaking of translational symmetry. Experimental evidence for such a state exists in high magnetic field and in some materials that feature density wave orders that explicitly break translational symmetry. However, evidence for a zero-field PDW state that exists independent of other spatially ordered states has so far been elusive. Here, we show that such a state exists in the iron pnictide superconductor EuRbFe4As4 (Eu-1144), a material that features coexisting superconductivity (Tc ~ 37K) and magnetism (Tm ~ 15 K). We show from the Spectroscopic Imaging Scanning Tunneling Microscopy (SI-STM) measurements that the superconducting gap at low temperature has long-range, unidirectional spatial modulations with an incommensurate period of ~8 unit cells. Upon raising the temperature above Tm, the modulated superconductor disappears, but a uniform superconducting gap survives to Tc. When an external magnetic field is applied, gap modulations disappear inside the vortex halo. The SI-STM and bulk measurements show the absence of other density wave orders, showing that the PDW state is a primary, zero-field superconducting state in this compound. Both four-fold rotational symmetry and translation symmetry are recovered above Tm, indicating that the PDW is a smectic order.

cond-mat.supr-con↗
arXiv2023

Transformed Distribution Matching for Missing Value Imputation

We study the problem of imputing missing values in a dataset, which has important applications in many domains. The key to missing value imputation is to capture the data distribution with incomplete samples and impute the missing values accordingly. In this paper, by leveraging the fact that any two batches of data with missing values come from the same data distribution, we propose to impute the missing values of two batches of samples by transforming them into a latent space through deep invertible functions and matching them distributionally. To learn the transformations and impute the missing values simultaneously, a simple and well-motivated algorithm is proposed. Our algorithm has fewer hyperparameters to fine-tune and generates high-quality imputations regardless of how missing values are generated. Extensive experiments over a large number of datasets and competing benchmark algorithms show that our method achieves state-of-the-art performance.

cs.LG↗
arXiv2023

Vector Quantized Wasserstein Auto-Encoder

Learning deep discrete latent presentations offers a promise of better symbolic and summarized abstractions that are more useful to subsequent downstream tasks. Inspired by the seminal Vector Quantized Variational Auto-Encoder (VQ-VAE), most of work in learning deep discrete representations has mainly focused on improving the original VQ-VAE form and none of them has studied learning deep discrete representations from the generative viewpoint. In this work, we study learning deep discrete representations from the generative viewpoint. Specifically, we endow discrete distributions over sequences of codewords and learn a deterministic decoder that transports the distribution over the sequences of codewords to the data distribution via minimizing a WS distance between them. We develop further theories to connect it with the clustering viewpoint of WS distance, allowing us to have a better and more controllable clustering solution. Finally, we empirically evaluate our method on several well-known benchmarks, where it achieves better qualitative and quantitative performances than the other VQ-VAE variants in terms of the codebook utilization and image reconstruction/generation.

cs.LG↗
arXiv2023

Generating Adversarial Examples with Task Oriented Multi-Objective Optimization

Deep learning models, even the-state-of-the-art ones, are highly vulnerable to adversarial examples. Adversarial training is one of the most efficient methods to improve the model's robustness. The key factor for the success of adversarial training is the capability to generate qualified and divergent adversarial examples which satisfy some objectives/goals (e.g., finding adversarial examples that maximize the model losses for simultaneously attacking multiple models). Therefore, multi-objective optimization (MOO) is a natural tool for adversarial example generation to achieve multiple objectives/goals simultaneously. However, we observe that a naive application of MOO tends to maximize all objectives/goals equally, without caring if an objective/goal has been achieved yet. This leads to useless effort to further improve the goal-achieved tasks, while putting less focus on the goal-unachieved tasks. In this paper, we propose \emph{Task Oriented MOO} to address this issue, in the context where we can explicitly define the goal achievement for a task. Our principle is to only maintain the goal-achieved tasks, while letting the optimizer spend more effort on improving the goal-unachieved tasks. We conduct comprehensive experiments for our Task Oriented MOO on various adversarial example generation schemes. The experimental results firmly demonstrate the merit of our proposed approach. Our code is available at \url{https://github.com/tuananhbui89/TAMOO}.

cs.LG↗
arXiv2023

Feature-based Learning for Diverse and Privacy-Preserving Counterfactual Explanations

Interpretable machine learning seeks to understand the reasoning process of complex black-box systems that are long notorious for lack of explainability. One flourishing approach is through counterfactual explanations, which provide suggestions on what a user can do to alter an outcome. Not only must a counterfactual example counter the original prediction from the black-box classifier but it should also satisfy various constraints for practical applications. Diversity is one of the critical constraints that however remains less discussed. While diverse counterfactuals are ideal, it is computationally challenging to simultaneously address some other constraints. Furthermore, there is a growing privacy concern over the released counterfactual data. To this end, we propose a feature-based learning framework that effectively handles the counterfactual constraints and contributes itself to the limited pool of private explanation models. We demonstrate the flexibility and effectiveness of our method in generating diverse counterfactuals of actionability and plausibility. Our counterfactual engine is more efficient than counterparts of the same capacity while yielding the lowest re-identification risks.

cs.AI↗
arXiv2023

Small Fermi pockets intertwined with charge stripes and pair density wave order in a kagome superconductor

The kagome superconductor family AV3Sb5 (A=Cs, K, Rb) emerged as an exciting platform to study exotic Fermi surface instabilities. Here we use spectroscopic-imaging scanning tunneling microscopy (SI-STM) and angle-resolved photoemission spectroscopy (ARPES) to reveal how the surprising cascade of higher and lower-dimensional density waves in CsV3Sb5 is intimately tied to a set of small reconstructed Fermi pockets. ARPES measurements visualize the formation of these pockets generated by a 3D charge density wave transition. The pockets are connected by dispersive q* wave vectors observed in Fourier transforms of STM differential conductance maps. As the additional 1D charge order emerges at a lower temperature, q* wave vectors become substantially renormalized, signaling further reconstruction of the Fermi pockets. Remarkably, in the superconducting state, the superconducting gap modulations give rise to an in-plane Cooper pair-density-wave at the same q* wave vectors. Our work demonstrates the intrinsic origin of the charge-stripes and the pair-density-wave in CsV3Sb5 and their relationship to the Fermi pockets. These experiments uncover a unique scenario of how Fermi pockets generated by a parent charge density wave state can provide a favorable platform for the emergence of additional density waves.

cond-mat.supr-con↗
arXiv2023

Extinction of Taurus, Orion, Perseus and California Molecular Clouds Based on the LAMOST, 2MASS and Gaia surveys I: Three-dimensional Extinction and Structure

The three-dimensional extinction and structure are studied for the Taurus, Orion, Perseus and California molecular clouds based on the LAMOST spectroscopy. Stellar color excess is calculated with the intrinsic color index derived from the atmospheric parameters in the LAMOST DR8 catalog and the observed color index in the Gaia EDR3 and the 2MASS PSC. In combination with the distance from the Gaia EDR3 parallax, the three-dimensional dust extinction maps are retrieved in the color excesses $E_{\rm{G_{BP},G_{RP}}}$ and $E_{\rm{J,K_{S}}}$ with an uncertainty of $\sim$0.03mag and $\sim$0.07mag respectively. The extinction maps successfully separate the clouds that overlap in the sky area and manifest the structure of the individual cloud. Meanwhile, a bow-like structure is found with a distance range from 175pc to 250pc, half of which is a part of the Per-Tau Shell in similar coordinates and distance while the other half is not. Three low-extinction rings are additionally discovered and briefly discussed.

astro-ph.SR↗
arXiv2022

The Pristine Inner Galaxy Survey (PIGS) VI: Different vertical distributions between two DIBs at 442.8 nm and 862.1 nm

Although diffuse interstellar bands (DIBs) were discovered over 100 years ago, for most of them, their origins are still unknown. Investigation on the correlations between different DIBs is an important way to study the behavior and distributions of their carriers. Based on stacking thousands of spectra from the Pristine Inner Galaxy Survey, we study the correlations between two DIBs at 442.8 nm ($λ$442.8) and 862.1 nm ($λ$862.1), as well as the dust grains, in a range of latitude spanning ${\sim}22^{\circ}$ ($4^{\circ}\,{<}\,|b|\,{<}\,15^{\circ}$) toward the Galactic center ($|\ell|\,{<}\,11^{\circ}$). Tight linear intensity correlations can be found between $λ$442.8, $λ$862.1, and dust grains for $|b|\,{<}\,12^{\circ}$ or $E(B-V)\,{>}\,0.3$ mag. For $|b|\,{>}\,12^{\circ}$, $λ$442.8 and $λ$862.1 present larger relative strength with respect to the dust grains. A systematic variation of the relative strength between $λ$442.8 and $λ$862.1 with $|b|$ and $E(B-V)$ concludes that the two DIBs do not share a common carrier. Furthermore, the carrier of $λ$862.1 is more abundant at high latitudes than that of $λ$442.8. This work can be treated as an example showing the significance and potentials to the DIB research covering a large latitude range.

astro-ph.GA↗
arXiv2022

Rotation of the dislocation grid in multilayer FeSe films and visualization of electronic nematic domains via orbital-selective tunneling

Understanding the interplay of structural and electronic symmetry breaking in Fe-based high temperature superconductors remains of high interest. In this work we grow strain-patterned multilayer FeSe thin films in a range of thicknesses using molecular beam epitaxy. We study the formation of electronic nematic domains and spatially-varying strain using scanning tunneling microscopy and spectroscopy. We directly visualize the formation of edge dislocations that give rise to a two-dimensional edge dislocation network in the films. Interestingly, we observe a 45 degree in-plane rotation of the dislocation network as a function of film thickness, yielding antisymmetric strain along different directions. This results in different coupling ratios between electronic nematic domains and antisymmetric strain. Lastly, we are able to distinguish between different orthogonal nematic domains by revealing a small energy-dependent difference in differential conductance maps between the two regions. This could be explained by orbital-selective tip-sample tunneling. Our observations bring new insights into the dislocation network formation in epitaxial thin films and provide another nanoscale tool to explore electronic nematicity in Fe-based superconductors.

cond-mat.supr-con↗
arXiv2022

Plethora of tunable Weyl fermions in kagome magnet Fe3Sn2 thin films

Interplay of magnetism and electronic band topology in unconventional magnets enables the creation and fine control of novel electronic phenomena. In this work, we use scanning tunneling microscopy and spectroscopy to study thin films of a prototypical kagome magnet Fe3Sn2. Our experiments reveal an unusually large number of densely-spaced spectroscopic features straddling the Fermi level. These are consistent with signatures of low-energy Weyl fermions and associated topological Fermi arc surface states predicted by theory. By measuring their response as a function of magnetic field, we discover a pronounced evolution in energy tied to the magnetization direction. Electron scattering and interference imaging further demonstrates the tunable nature of a subset of related electronic states. Our experiments provide the first visualization of how in-situ spin reorientation drives changes in the electronic density of states of the Weyl fermion band structure. Combined with previous reports of massive Dirac fermions, flat bands and electronic nematicity, our work establishes Fe3Sn2 as a unique platform that harbors an extraordinarily wide array of topological and correlated electron phenomena.

cond-mat.mtrl-sci↗
arXiv2022

Adaptive Distribution Calibration for Few-Shot Learning with Hierarchical Optimal Transport

Few-shot classification aims to learn a classifier to recognize unseen classes during training, where the learned model can easily become over-fitted based on the biased distribution formed by only a few training examples. A recent solution to this problem is calibrating the distribution of these few sample classes by transferring statistics from the base classes with sufficient examples, where how to decide the transfer weights from base classes to novel classes is the key. However, principled approaches for learning the transfer weights have not been carefully studied. To this end, we propose a novel distribution calibration method by learning the adaptive weight matrix between novel samples and base classes, which is built upon a hierarchical Optimal Transport (H-OT) framework. By minimizing the high-level OT distance between novel samples and base classes, we can view the learned transport plan as the adaptive weight information for transferring the statistics of base classes. The learning of the cost function between a base class and novel class in the high-level OT leads to the introduction of the low-level OT, which considers the weights of all the data samples in the base class. Experimental results on standard benchmarks demonstrate that our proposed plug-and-play model outperforms competing approaches and owns desired cross-domain generalization ability, indicating the effectiveness of the learned adaptive weights.

cs.LG↗
arXiv2022

Uncertainty Estimation for Multi-view Data: The Power of Seeing the Whole Picture

Uncertainty estimation is essential to make neural networks trustworthy in real-world applications. Extensive research efforts have been made to quantify and reduce predictive uncertainty. However, most existing works are designed for unimodal data, whereas multi-view uncertainty estimation has not been sufficiently investigated. Therefore, we propose a new multi-view classification framework for better uncertainty estimation and out-of-domain sample detection, where we associate each view with an uncertainty-aware classifier and combine the predictions of all the views in a principled way. The experimental results with real-world datasets demonstrate that our proposed approach is an accurate, reliable, and well-calibrated classifier, which predominantly outperforms the multi-view baselines tested in terms of expected calibration error, robustness to noise, and accuracy for the in-domain sample classification and the out-of-domain sample detection tasks.

cs.LG↗