Searcharxiv⌕ Search

arXiv subjects

Thanh Nguyen

Publications and source records attributed to Thanh Nguyen.

At least 73 records · Page 4Linked to original sources
arXiv2022

LAD: A Hybrid Deep Learning System for Benign Paroxysmal Positional Vertigo Disorders Diagnostic

Herein, we introduce "Look and Diagnose" (LAD), a hybrid deep learning-based system that aims to support doctors in the medical field in diagnosing effectively the Benign Paroxysmal Positional Vertigo (BPPV) disorder. Given the body postures of the patient in the Dix-Hallpike and lateral head turns test, the visual information of both eyes is captured and fed into LAD for analyzing and classifying into one of six possible disorders the patient might be suffering from. The proposed system consists of two streams: (1) an RNN-based stream that takes raw RGB images of both eyes to extract visual features and optical flow of each eye followed by ternary classification to determine left/right posterior canal (PC) or other; and (2) pupil detector stream that detects the pupil when it is classified as Non-PC and classifies the direction and strength of the beating to categorize the Non-PC types into the remaining four classes: Geotropic BPPV (left and right) and Apogeotropic BPPV (left and right). Experimental results show that with the patient's body postures, the system can accurately classify given BPPV disorder into the six types of disorders with an accuracy of 91% on the validation set. The proposed method can successfully classify disorders with an accuracy of 93% for the Posterior Canal disorder and 95% for the Geotropic and Apogeotropic disorder, paving a potential direction for research with the medical data.

cs.CV↗
arXiv2022

Machine learning spectral indicators of topology

Topological materials discovery has emerged as an important frontier in condensed matter physics. While theoretical classification frameworks have been used to identify thousands of candidate topological materials, experimental determination of materials' topology often poses significant technical challenges. X-ray absorption spectroscopy (XAS) is a widely-used materials characterization technique sensitive to atoms' local symmetry and chemical bonding, which are intimately linked to band topology by the theory of topological quantum chemistry (TQC). Moreover, as a local structural probe, XAS is known to have high quantitative agreement between experiment and calculation, suggesting that insights from computational spectra can effectively inform experiments. In this work, we leverage computed X-ray absorption near-edge structure (XANES) spectra of more than 10,000 inorganic materials to train a neural network (NN) classifier that predicts topological class directly from XANES signatures, achieving F$_1$ scores of 89% and 93% for topological and trivial classes, respectively. Additionally, we obtain consistent classifications using corresponding experimental and computational XANES spectra for a small number of measured compounds. Given the simplicity of the XAS setup and its compatibility with multimodal sample environments, the proposed machine learning-augmented XAS topological indicator has the potential to discover broader categories of topological materials, such as non-cleavable compounds and amorphous materials, and may further inform field-driven phenomena in situ, such as magnetic field-driven topological phase transitions.

cond-mat.dis-nn↗
arXiv2022

Few-shot Object Counting and Detection

We tackle a new task of few-shot object counting and detection. Given a few exemplar bounding boxes of a target object class, we seek to count and detect all objects of the target class. This task shares the same supervision as the few-shot object counting but additionally outputs the object bounding boxes along with the total object count. To address this challenging problem, we introduce a novel two-stage training strategy and a novel uncertainty-aware few-shot object detector: Counting-DETR. The former is aimed at generating pseudo ground-truth bounding boxes to train the latter. The latter leverages the pseudo ground-truth provided by the former but takes the necessary steps to account for the imperfection of pseudo ground-truth. To validate the performance of our method on the new task, we introduce two new datasets named FSCD-147 and FSCD-LVIS. Both datasets contain images with complex scenes, multiple object classes per image, and a huge variation in object shapes, sizes, and appearance. Our proposed approach outperforms very strong baselines adapted from few-shot object counting and few-shot object detection with a large margin in both counting and detection metrics. The code and models are available at https://github.com/VinAIResearch/Counting-DETR.

cs.CV↗
arXiv2022

Hydrogen abstraction reactions in formic and thioformic acid isomers by hydrogen and deuterium atoms. Insights on isomerism and deuteration

The isomerism of molecules in the interstellar medium and the mechanisms behind it are essential questions in the chemistry of organic molecules in space. In particular, for the simple formic and thioformic acids, the low temperatures found in molecular clouds indicate that cis-trans isomerization in the gas-phase must be impeded. Reactions happening on top of interstellar dust grains may explain the isomer interconversion at low temperatures. We studied the isomerization processes of formic and thioformic acid susceptible to happen on the surface of interstellar dust grains and initiated by H abstraction reactions. Similarly, deuterium enrichment of the acids can occur by the same mechanism. Our objective is to shed light on both topics to increase our understanding of key precursors of organic molecules in space.

astro-ph.GA↗
arXiv2022

Elucidating proximity magnetism through polarized neutron reflectometry and machine learning

Polarized neutron reflectometry is a powerful technique to interrogate the structures of multilayered magnetic materials with depth sensitivity and nanometer resolution. However, reflectometry profiles often inhabit a complicated objective function landscape using traditional fitting methods, posing a significant challenge to parameter retrieval. In this work, we develop a data-driven framework to recover the sample parameters from polarized neutron reflectometry data with minimal user intervention. We train a variational autoencoder to map reflectometry profiles with moderate experimental noise to an interpretable, low-dimensional space from which sample parameters can be extracted with high resolution. We apply our method to recover the scattering length density profiles of the topological insulator-ferromagnetic insulator heterostructure Bi$_2$Se$_3$/EuS exhibiting proximity magnetism, in good agreement with the results of conventional fitting. We further analyze a more challenging reflectometry profile of the topological insulator-antiferromagnet heterostructure (Bi,Sb)$_2$Te$_3$/Cr$_2$O$_3$ and identify possible interfacial proximity magnetism in this material. We anticipate the framework developed here can be applied to resolve hidden interfacial phenomena in a broad range of layered systems.

cond-mat.mtrl-sci↗
arXiv2022

Learning to Attack with Fewer Pixels: A Probabilistic Post-hoc Framework for Refining Arbitrary Dense Adversarial Attacks

Deep neural network image classifiers are reported to be susceptible to adversarial evasion attacks, which use carefully crafted images created to mislead a classifier. Many adversarial attacks belong to the category of dense attacks, which generate adversarial examples by perturbing all the pixels of a natural image. To generate sparse perturbations, sparse attacks have been recently developed, which are usually independent attacks derived by modifying a dense attack's algorithm with sparsity regularisations, resulting in reduced attack efficiency. In this paper, we aim to tackle this task from a different perspective. We select the most effective perturbations from the ones generated from a dense attack, based on the fact we find that a considerable amount of the perturbations on an image generated by dense attacks may contribute little to attacking a classifier. Accordingly, we propose a probabilistic post-hoc framework that refines given dense attacks by significantly reducing the number of perturbed pixels but keeping their attack power, trained with mutual information maximisation. Given an arbitrary dense attack, the proposed model enjoys appealing compatibility for making its adversarial images more realistic and less detectable with fewer perturbations. Moreover, our framework performs adversarial attacks much faster than existing sparse attacks.

cs.CV↗
arXiv2022

Panoramic mapping of phonon transport from ultrafast electron diffraction and machine learning

One central challenge in understanding phonon thermal transport is a lack of experimental tools to investigate mode-based transport information. Although recent advances in computation lead to mode-based information, it is hindered by unknown defects in bulk region and at interfaces. Here we present a framework that can reveal microscopic phonon transport information in heterostructures, integrating state-of-the-art ultrafast electron diffraction (UED) with advanced scientific machine learning. Taking advantage of the dual temporal and reciprocal-space resolution in UED, we are able to reliably recover the frequency-dependent interfacial transmittance with possible extension to frequency-dependent relaxation times of the heterostructure. This enables a direct reconstruction of real-space, real-time, frequency-resolved phonon dynamics across an interface. Our work provides a new pathway to experimentally probe phonon transport mechanisms with unprecedented details.

cond-mat.mtrl-sci↗
arXiv2022

Electronic properties of correlated kagomé metals AV$_3$Sb$_5$(A = K, Rb, Cs): A perspective

Following the discovery of a new family of kagomé prototypical materials with structure AV$_3$Sb$_5$ (A = K, Rb, Cs), there has been heightened interest in studying the correlation-driven electronic phenomena in these kagomé lattice systems. The study of these materials has gone beyond magneto-transport measurements to reveal exciting features such as Dirac bands, anomalous Hall effect, bulk superconductivity with $T_c$ $\sim$ 0.9 K-2.5 K, and the observation of charge density wave instabilities, suggesting an intertwining of topological physics and new quantum orders. Moreover, very recent works on numerous types of experiments have appeared further examining the unconventional superconductivity and the exotic electronic states found within these kagomé materials. Theories on the strong interactions that play a role in these systems have been proposed to shed light on the nature of these topological charge density waves. In this brief review, we summarize these recent experimental findings and theoretical proposals, and envision the materials as new platforms to study the interplay between topological physics and strongly-correlated electronic systems.

cond-mat.str-el↗
arXiv2022

Topological Signatures in Nodal Semimetals through Neutron Scattering

Topological nodal semimetals are known to host a variety of fascinating electronic properties due to the topological protection of the band-touching nodes. Neutron scattering, despite its power in probing elementary excitations, has not been routinely applied to topological semimetals, mainly due to the lack of an explicit connection between the neutron response and the signature of topology. In this work, we theoretically investigate the role that neutron scattering can play to unveil the topological nodal features: a large magnetic neutron response with spectral non-analyticity can be generated solely from the nodal bands. A new formula for the dynamical structure factor for generic topological nodal metals is derived. For Weyl semimetals, we show that the locations of Weyl nodes, the Fermi velocities and the signature of chiral anomaly can all leave hallmark neutron spectral responses. Our work offers a neutron-based avenue towards probing bulk topological materials.

cond-mat.mtrl-sci↗
arXiv2021

Successive H-atom addition to solid OCS on compact amorphous solid water

Carbonyl sulfide (OCS) is an abundant sulfur (S)-bearing species in the interstellar medium. It is present not only in the gas phase, but also on interstellar grains as a solid; therefore, OCS very likely undergoes physicochemical processes on icy surfaces at very low temperatures. The present study experimentally and computationally investigates the reaction of solid OCS with hydrogen (H) atoms on amorphous solid water at low temperatures. The results show that the addition of H to OCS proceeds via quantum tunneling, and further addition of H leads to the formation of carbon monoxide (CO), hydrogen sulfide (H2S), formaldehyde (H2CO), methanol (CH3OH) and thioformic acid (HC(O)SH). These experimental results are explained by our quantum chemical calculations, which demonstrate that the initial addition of H to the S atom of OCS is the most predominant, leading to the formation of OCS-H radicals. Once the formed OCS-H radical is stabilized on ice, further addition of H to the S atom yields CO and H2S, while that to the C atom yields HC(O)SH. We have also confirmed, in a separate experiment, the HC(O)SH formation by the HCO reactions with the SH radicals. The present results would have an important implication for the recent detection of HC(O)SH toward G+0.693-0.027.

astro-ph.GA↗
arXiv2021

Sample-efficient Reinforcement Learning Representation Learning with Curiosity Contrastive Forward Dynamics Model

Developing an agent in reinforcement learning (RL) that is capable of performing complex control tasks directly from high-dimensional observation such as raw pixels is yet a challenge as efforts are made towards improving sample efficiency and generalization. This paper considers a learning framework for Curiosity Contrastive Forward Dynamics Model (CCFDM) in achieving a more sample-efficient RL based directly on raw pixels. CCFDM incorporates a forward dynamics model (FDM) and performs contrastive learning to train its deep convolutional neural network-based image encoder (IE) to extract conducive spatial and temporal information for achieving a more sample efficiency for RL. In addition, during training, CCFDM provides intrinsic rewards, produced based on FDM prediction error, encourages the curiosity of the RL agent to improve exploration. The diverge and less-repetitive observations provide by both our exploration strategy and data augmentation available in contrastive learning improve not only the sample efficiency but also the generalization. Performance of existing model-free RL methods such as Soft Actor-Critic built on top of CCFDM outperforms prior state-of-the-art pixel-based RL methods on the DeepMind Control Suite benchmark.

cs.LG↗
arXiv2021

Signature of Many-Body Localization of Phonons in Strongly Disordered Superlattices

Many-body localization (MBL) has attracted significant attention due to its immunity to thermalization, role in logarithmic entanglement entropy growth, and opportunities to reach exotic quantum orders. However, experimental realization of MBL in solid-state systems has remained challenging. Here we report evidence of a possible phonon MBL phase in disordered GaAs/AlAs superlattices. Through grazing-incidence inelastic X-ray scattering, we observe a strong deviation of the phonon population from equilibrium in samples doped with ErAs nanodots at low temperature, signaling a departure from thermalization. This behavior occurs within finite phonon energy and wavevector windows, suggesting a localization-thermalization crossover. We support our observation by proposing a theoretical model for the effective phonon Hamiltonian in disordered superlattices, and showing that it can be mapped exactly to a disordered 1D Bose-Hubbard model with a known MBL phase. Our work provides momentum-resolved experimental evidence of phonon localization, extending the scope of MBL to disordered solid-state systems.

cond-mat.mtrl-sci↗
arXiv2021

Robust MAML: Prioritization task buffer with adaptive learning process for model-agnostic meta-learning

Model agnostic meta-learning (MAML) is a popular state-of-the-art meta-learning algorithm that provides good weight initialization of a model given a variety of learning tasks. The model initialized by provided weight can be fine-tuned to an unseen task despite only using a small amount of samples and within a few adaptation steps. MAML is simple and versatile but requires costly learning rate tuning and careful design of the task distribution which affects its scalability and generalization. This paper proposes a more robust MAML based on an adaptive learning scheme and a prioritization task buffer(PTB) referred to as Robust MAML (RMAML) for improving scalability of training process and alleviating the problem of distribution mismatch. RMAML uses gradient-based hyper-parameter optimization to automatically find the optimal learning rate and uses the PTB to gradually adjust train-ing task distribution toward testing task distribution over the course of training. Experimental results on meta reinforcement learning environments demonstrate a substantial performance gain as well as being less sensitive to hyper-parameter choice and robust to distribution mismatch.

cs.LG↗
arXiv2021

Ensemble Learning based on Classifier Prediction Confidence and Comprehensive Learning Particle Swarm Optimisation for polyp localisation

Colorectal cancer (CRC) is the first cause of death in many countries. CRC originates from a small clump of cells on the lining of the colon called polyps, which over time might grow and become malignant. Early detection and removal of polyps are therefore necessary for the prevention of colon cancer. In this paper, we introduce an ensemble of medical polyp segmentation algorithms. Based on an observation that different segmentation algorithms will perform well on different subsets of examples because of the nature and size of training sets they have been exposed to and because of method-intrinsic factors, we propose to measure the confidence in the prediction of each algorithm and then use an associate threshold to determine whether the confidence is acceptable or not. An algorithm is selected for the ensemble if the confidence is below its associate threshold. The optimal threshold for each segmentation algorithm is found by using Comprehensive Learning Particle Swarm Optimization (CLPSO), a swarm intelligence algorithm. The Dice coefficient, a popular performance metric for image segmentation, is used as the fitness criteria. Experimental results on two polyp segmentation datasets MICCAI2015 and Kvasir-SEG confirm that our ensemble achieves better results compared to some well-known segmentation algorithms.

cs.CV↗
arXiv2021

Probabilistic Dipole Inversion for Adaptive Quantitative Susceptibility Mapping

A learning-based posterior distribution estimation method, Probabilistic Dipole Inversion (PDI), is proposed to solve the quantitative susceptibility mapping (QSM) inverse problem in MRI with uncertainty estimation. In PDI, a deep convolutional neural network (CNN) is used to represent the multivariate Gaussian distribution as the approximate posterior distribution of susceptibility given the input measured field. Such CNN is first trained on healthy subjects via posterior density estimation, where the training dataset contains samples from the true posterior distribution. Domain adaptations are then deployed on patient datasets with new pathologies not included in pre-training, where PDI updates the pre-trained CNN's weights in an unsupervised fashion by minimizing the Kullback-Leibler divergence between the approximate posterior distribution represented by CNN and the true posterior distribution from the likelihood distribution of a known physical model and pre-defined prior distribution. Based on our experiments, PDI provides additional uncertainty estimation compared to the conventional MAP approach, while addressing the potential issue of the pre-trained CNN when test data deviates from training.

eess.IV↗
arXiv2021

NeRD: Neural Representation of Distribution for Medical Image Segmentation

We introduce Neural Representation of Distribution (NeRD) technique, a module for convolutional neural networks (CNNs) that can estimate the feature distribution by optimizing an underlying function mapping image coordinates to the feature distribution. Using NeRD, we propose an end-to-end deep learning model for medical image segmentation that can compensate the negative impact of feature distribution shifting issue caused by commonly used network operations such as padding and pooling. An implicit function is used to represent the parameter space of the feature distribution by querying the image coordinate. With NeRD, the impact of issues such as over-segmenting and missing have been reduced, and experimental results on the challenging white matter lesion segmentation and left atrial segmentation verify the effectiveness of the proposed method. The code is available via https://github.com/tinymilky/NeRD.

eess.IV↗
arXiv2021

Machine Learning on Neutron and X-Ray Scattering

Neutron and X-ray scattering represent two state-of-the-art materials characterization techniques that measure materials' structural and dynamical properties with high precision. These techniques play critical roles in understanding a wide variety of materials systems, from catalysis to polymers, nanomaterials to macromolecules, and energy materials to quantum materials. In recent years, neutron and X-ray scattering have received a significant boost due to the development and increased application of machine learning to materials problems. This article reviews the recent progress in applying machine learning techniques to augment various neutron and X-ray scattering techniques. We highlight the integration of machine learning methods into the typical workflow of scattering experiments. We focus on scattering problems that faced challenge with traditional methods but addressable using machine learning, such as leveraging the knowledge of simple materials to model more complicated systems, learning with limited data or incomplete labels, identifying meaningful spectra and materials' representations for learning tasks, mitigating spectral noise, and many others. We present an outlook on a few emerging roles machine learning may play in broad types of scattering and spectroscopic problems in the foreseeable future.

cond-mat.mtrl-sci↗