arXiv subjects

Bo Song

Publications and source records attributed to Bo Song.

At least 37 records · Page 2Linked to original sources

arXiv2024

BadLabel: A Robust Perspective on Evaluating and Enhancing Label-noise Learning

Label-noise learning (LNL) aims to increase the model's generalization given training data with noisy labels. To facilitate practical LNL algorithms, researchers have proposed different label noise types, ranging from class-conditional to instance-dependent noises. In this paper, we introduce a novel label noise type called BadLabel, which can significantly degrade the performance of existing LNL algorithms by a large margin. BadLabel is crafted based on the label-flipping attack against standard classification, where specific samples are selected and their labels are flipped to other labels so that the loss values of clean and noisy labels become indistinguishable. To address the challenge posed by BadLabel, we further propose a robust LNL method that perturbs the labels in an adversarial manner at each epoch to make the loss values of clean and noisy labels again distinguishable. Once we select a small set of (mostly) clean labeled data, we can apply the techniques of semi-supervised learning to train the model accurately. Empirically, our experimental results demonstrate that existing LNL algorithms are vulnerable to the newly introduced BadLabel noise type, while our proposed robust LNL method can effectively improve the generalization performance of the model under various types of label noise. The new dataset of noisy labels and the source codes of robust LNL algorithms are available at https://github.com/zjfheart/BadLabels.

cs.LG↗

arXiv2024

Chirality tuning and reversing with resonant phase-change metasurfaces

Dynamic control of circular dichroism in photonic structures is critically important for compact spectrometers, stereoscopic displays, and information processing exploiting multiple degrees of freedom. Metasurfaces can help miniaturize chiral devices but only produce static and limited chiral responses. While external stimuli are able to tune resonances, their modulations are often weak, and reversing continuously the sign of circular dichroism is extremely challenging. Here, we demonstrate dynamically tunable chiral response of resonant metasurfaces supporting chiral bound states in the continuum combining them with phase-change materials. Phase transition between amorphous and crystalline phases allows to control chiral response and vary chirality rapidly from -0.947 to +0.958 backward and forward via chirality continuum. Our demonstrations underpin the rapid development of chiral photonics and its applications.

physics.optics↗

arXiv2023

CircleFormer: Circular Nuclei Detection in Whole Slide Images with Circle Queries and Attention

Both CNN-based and Transformer-based object detection with bounding box representation have been extensively studied in computer vision and medical image analysis, but circular object detection in medical images is still underexplored. Inspired by the recent anchor free CNN-based circular object detection method (CircleNet) for ball-shape glomeruli detection in renal pathology, in this paper, we present CircleFormer, a Transformer-based circular medical object detection with dynamic anchor circles. Specifically, queries with circle representation in Transformer decoder iteratively refine the circular object detection results, and a circle cross attention module is introduced to compute the similarity between circular queries and image features. A generalized circle IoU (gCIoU) is proposed to serve as a new regression loss of circular object detection as well. Moreover, our approach is easy to generalize to the segmentation task by adding a simple segmentation branch to CircleFormer. We evaluate our method in circular nuclei detection and segmentation on the public MoNuSeg dataset, and the experimental results show that our method achieves promising performance compared with the state-of-the-art approaches. The effectiveness of each component is validated via ablation studies as well. Our code is released at https://github.com/zhanghx-iim-ahu/CircleFormer.

cs.CV↗

arXiv2023

Assessing Vulnerabilities of Adversarial Learning Algorithm through Poisoning Attacks

Adversarial training (AT) is a robust learning algorithm that can defend against adversarial attacks in the inference phase and mitigate the side effects of corrupted data in the training phase. As such, it has become an indispensable component of many artificial intelligence (AI) systems. However, in high-stake AI applications, it is crucial to understand AT's vulnerabilities to ensure reliable deployment. In this paper, we investigate AT's susceptibility to poisoning attacks, a type of malicious attack that manipulates training data to compromise the performance of the trained model. Previous work has focused on poisoning attacks against standard training, but little research has been done on their effectiveness against AT. To fill this gap, we design and test effective poisoning attacks against AT. Specifically, we investigate and design clean-label poisoning attacks, allowing attackers to imperceptibly modify a small fraction of training data to control the algorithm's behavior on a specific target data point. Additionally, we propose the clean-label untargeted attack, enabling attackers can attach tiny stickers on training data to degrade the algorithm's performance on all test data, where the stickers could serve as a signal against unauthorized data collection. Our experiments demonstrate that AT can still be poisoned, highlighting the need for caution when using vanilla AT algorithms in security-related applications. The code is at https://github.com/zjfheart/Poison-adv-training.git.

cs.CR↗

arXiv2023

Observing the two-dimensional Bose glass in an optical quasicrystal

The combined effect of disorder and interactions is central to the richness of condensed matter physics and can lead to novel quantum states such as the Bose glass phase in disordered bosonic systems. Here, we report on the experimental realisation of the two-dimensional Bose glass using ultra-cold atoms in an eight-fold symmetric quasicrystalline optical lattice. By probing the coherence properties of the system, we observe a Bose glass to superfluid transition and map out the phase diagram in the weakly interacting regime. Moreover, we reveal the non-ergodic nature of the Bose glass by probing the capability to restore coherence. Our observations are in good agreement with recent quantum Monte Carlo predictions and pave the way for experimentally testing the connection between the Bose glass, many-body localisation, and glassy dynamics more generally.

cond-mat.quant-gas↗

arXiv2022

Relay Hindsight Experience Replay: Self-Guided Continual Reinforcement Learning for Sequential Object Manipulation Tasks with Sparse Rewards

Exploration with sparse rewards remains a challenging research problem in reinforcement learning (RL). Especially for sequential object manipulation tasks, the RL agent always receives negative rewards until completing all sub-tasks, which results in low exploration efficiency. To solve these tasks efficiently, we propose a novel self-guided continual RL framework, RelayHER (RHER). RHER first decomposes a sequential task into new sub-tasks with increasing complexity and ensures that the simplest sub-task can be learned quickly by utilizing Hindsight Experience Replay (HER). Secondly, we design a multi-goal & multi-task network to learn these sub-tasks simultaneously. Finally, we propose a Self-Guided Exploration Strategy (SGES). With SGES, the learned sub-task policy will guide the agent to the states that are helpful to learn more complex sub-task with HER. By this self-guided exploration and relay policy learning, RHER can solve these sequential tasks efficiently stage by stage. The experimental results show that RHER significantly outperforms vanilla-HER in sample-efficiency on five singleobject and five complex multi-object manipulation tasks (e.g., Push, Insert, ObstaclePush, Stack, TStack, etc.). The proposed RHER has also been applied to learn a contact-rich push task on a physical robot from scratch, and the success rate reached 10/10 with only 250 episodes.

cs.RO↗

arXiv2021

Soft-Elasticity Optimises Dissipation in 3D-Printed Liquid Crystal Elastomers

Soft-elasticity in monodomain liquid crystal elastomers (LCEs) is promising for impact-absorbing applications where strain energy is ideally absorbed at constant stress. Conventionally, compressive and impact studies on LCEs have not been performed given the notorious difficulty synthesizing sufficiently large monodomain devices. Here we demonstrate 3D printing bulk ($>cm^3$) monodomain LCE devices using direct ink writing and study their compressive soft-elasticity over 8 decades of strain rate. At quasi-static rates, the monodomain soft-elastic LCE dissipated 45% of strain energy while comparator materials dissipated less than 20%. At strain rates up to $3000~s^{-1}$, our soft-elastic monodomain LCE consistently performed closest to an ideal-impact absorber. Drop testing reveals soft-elasticity as a likely mechanism for effectively reducing the severity of impacts -- with soft elastic LCEs offering a Gadd Severity Index 40% lower than a comparable isotropic elastomer. Lastly, we demonstrate tailoring deformation and buckling behavior in monodomain LCEs via the printed director orientation.

cond-mat.mtrl-sci↗

arXiv2021

Realizing discontinuous quantum phase transitions in a strongly-correlated driven optical lattice

Discontinuous quantum phase transitions and the associated metastability play central roles in diverse areas of physics ranging from ferromagnetism to false vacuum decay in the early universe. Using strongly-interacting ultracold atoms in an optical lattice, we realize a driven many-body system whose quantum phase transition can be tuned from continuous to discontinuous. Resonant shaking of a one-dimensional optical lattice hybridizes the lowest two Bloch bands, driving a novel transition from a Mott insulator to a $π$-superfluid, i.e., a superfluid state with staggered phase order. For weak shaking amplitudes, this transition is discontinuous (first-order) and the system can remain frozen in a metastable state, whereas for strong shaking, it undergoes a continuous transition toward a $π$-superfluid. Our observations of this metastability and hysteresis are in good quantitative agreement with numerical simulations and pave the way for exploring the crucial role of quantum fluctuations in discontinuous transitions.

cond-mat.quant-gas↗

arXiv2021

A Neural Lip-Sync Framework for Synthesizing Photorealistic Virtual News Anchors

Lip sync has emerged as a promising technique for generating mouth movements from audio signals. However, synthesizing a high-resolution and photorealistic virtual news anchor is still challenging. Lack of natural appearance, visual consistency, and processing efficiency are the main problems with existing methods. This paper presents a novel lip-sync framework specially designed for producing high-fidelity virtual news anchors. A pair of Temporal Convolutional Networks are used to learn the cross-modal sequential mapping from audio signals to mouth movements, followed by a neural rendering network that translates the synthetic facial map into a high-resolution and photorealistic appearance. This fully trainable framework provides end-to-end processing that outperforms traditional graphics-based methods in many low-delay applications. Experiments also show the framework has advantages over modern neural-based methods in both visual appearance and efficiency.

cs.CV↗

arXiv2020

Evidence for Bosonization in a three-dimensional gas of SU($N$) fermions

Blurring the boundary between bosons and fermions lies at the heart of a wide range of intriguing quantum phenomena in multiple disciplines, ranging from condensed matter physics and atomic, molecular and optical physics to high energy physics. One such example is a multi-component Fermi gas with SU($N$) symmetry that is expected to behave like spinless bosons in the large $N$ limit, where the large number of internal states weakens constraints from the Pauli exclusion principle. However, bosonization in SU($N$) fermions has never been established in high dimensions where exact solutions are absent. Here, we report direct evidence for bosonization in a SU($N$) fermionic ytterbium gas with tunable $N$ in three dimensions (3D). We measure contacts, the central quantity controlling dilute quantum gases, from the momentum distribution, and find that the contact per spin approaches a constant with a 1/$N$ scaling in the low fugacity regime consistent with our theoretical prediction. This scaling signifies the vanishing role of the fermionic statistics in thermodynamics, and allows us to verify bosonization through measuring a single physical quantity. Our work delivers a highly controllable quantum simulator to exchange the bosonic and fermionic statistics through tuning the internal degrees of freedom in any generic dimensions. It also suggests a new route towards exploring multi-component quantum systems and their underlying symmetries with contacts.

cond-mat.quant-gas↗

arXiv2020

Designable ductility of a nano-network from few-layered graphene bonded with benzene: A molecular dynamics study

In nanoscale, motion operation of a nano-objective is usually realized by displacement load, which put forwards high requirement for ductility of material. Since pristine graphene has low ductility, once the stretching strain exceeds its critical value, it breaks in brittle style and loses ability to bear the external load quickly. Herein, to improve the ductility, a corrugated sandwich carbon nano-network model based on few-layered graphene is proposed, in which the two surface layers are bonded with several corrugated core layers via benzene molecules. Effects of factors such as the geometry of the carbon network, temperature, and strain rate, on the ductility are evaluated by molecular dynamics simulations. Conclusions are drawn for potential application of the new two-dimensional material with designable ductility.

physics.comp-ph↗

arXiv2020

Welding few-layered graphene ribbon via penetration of high-speed fullerenes

Two-dimensional materials are popular in design of nanodevices. To improve bending stiffness of few-layered graphene ribbon, the relative sliding between neighboring layers should be avoided by method such as bonding them together. In this study, the new bonds between the graphene layers in ribbon are formed by ballistic C60 fullerenes from both surfaces of the ribbon. Collision effects are evaluated through molecular dynamics simulations. The formation of the new two dimensional carbon material is demonstrated by breakage and generation of bonds between atoms from graphene and/or fullerenes. Note that the incident velocity of the fullerenes had an interval, in which both graphene and fullerenes will break and further be connected by new covalent bonds at the impact area.

cond-mat.mtrl-sci↗

arXiv2020

Balance Between Efficient and Effective Learning: Dense2Sparse Reward Shaping for Robot Manipulation with Environment Uncertainty

Efficient and effective learning is one of the ultimate goals of the deep reinforcement learning (DRL), although the compromise has been made in most of the time, especially for the application of robot manipulations. Learning is always expensive for robot manipulation tasks and the learning effectiveness could be affected by the system uncertainty. In order to solve above challenges, in this study, we proposed a simple but powerful reward shaping method, namely Dense2Sparse. It combines the advantage of fast convergence of dense reward and the noise isolation of the sparse reward, to achieve a balance between learning efficiency and effectiveness, which makes it suitable for robot manipulation tasks. We evaluated our Dense2Sparse method with a series of ablation experiments using the state representation model with system uncertainty. The experiment results show that the Dense2Sparse method obtained higher expected reward compared with the ones using standalone dense reward or sparse reward, and it also has a superior tolerance of system uncertainty.

cs.LG↗

arXiv2020

Effective statistical fringe removal algorithm for high-sensitivity imaging of ultracold atoms

High-sensitivity imaging of ultracold atoms is often challenging when interference patterns are imprinted on the imaging light. Such image noises result in low signal-to-noise ratio and limit the capability to extract subtle physical quantities. Here we demonstrate an advanced fringe removal algorithm for absorption imaging of ultracold atoms, which efficiently suppresses unwanted fringe patterns using a small number of sample images without taking additional reference images. The protocol is based on an image decomposition and projection method with an extended image basis. We apply this scheme to raw absorption images of degenerate Fermi gases for the measurement of atomic density fluctuations and temperatures. The quantitative analysis shows that image noises can be efficiently removed with only tens of reference images, which manifests the efficiency of our protocol. Our algorithm would be of particular interest for the quantum emulation experiments in which several physical parameters need to be scanned within a limited time duration.

cond-mat.quant-gas↗

arXiv2019

Collective excitations in two-dimensional SU($N$) Fermi gases with tunable spin

We measure collective excitations of a harmonically trapped two-dimensional (2D) SU($N$) Fermi gas of $^{173}$Yb confined to a stack of layers formed by a one-dimensional optical lattice. Quadrupole and breathing modes are excited and monitored in the collisionless regime $\lvert\ln(k_F a_{2D})\rvert\gg 1$ with tunable spin. We observe that the quadrupole mode frequency decreases with increasing number of spin components due to the amplification of the interaction effect by $N$ in agreement with a theoretical prediction based on 2D kinetic equations. The breathing mode frequency, however, is measured to be twice the dipole oscillation frequency regardless of $N$. We also follow the evolution of collective excitations in the dimensional crossover from two to three dimensions and characterize the damping rate of quadrupole and breathing modes for tunable SU($N$) fermions, both of which reveal the enhanced inter-particle collisions for larger spin. Our result paves the way to investigate the collective property of 2D SU($N$) Fermi liquid with enlarged spin.

cond-mat.quant-gas↗

arXiv2019

Cell vibron polariton in the myelin sheath of nerve

Polaritons are arousing tremendous interests in physics and material sciences for their unique and amazing properties, especially including the condensation, lasing without inversion and even room-temperature superfluidity. Herein, we propose a cell vibron polariton (cell-VP): a collectively coherent mode of a photon and all phospholipid molecules in a myelin sheath which is a nervous cell majorly consisting of the phospholipid molecules. Cell-VP can be resonantly self-confined in the myelin sheath under physiological conditions. The observations benefit from the specifically compact, ordered and polar thin-film structure of the sheath, and the relatively strong coupling of the mid-infrared photon with the vibrons of phospholipid tails in the myelin. The underlying physics is revealed to be the collectively coherent superposition of the photon and vibrons, the polariton induced significant enhancement of myelin permittivity, and the resonance of the polariton with the sheath cell. The captured cell-VPs in myelin sheaths may provide a promising way for super-efficient consumption of extra-weak bioenergy and even directly serve for quantum information in the nervous system. These findings further the understanding of neuroscience on the cellular level from the view of quantum mechanics.

physics.gen-ph↗

arXiv2019

Observation of nodal-line semimetal with ultracold fermions in an optical lattice

Observation of topological phases beyond two-dimension (2D) has been an open challenge for ultracold atoms. Here, we realize for the first time a 3D spin-orbit coupled nodal-line semimetal in an optical lattice and observe the bulk line nodes with ultracold fermions. The realized topological semimetal exhibits an emergent magnetic group symmetry. This allows to detect the nodal lines by effectively reconstructing the 3D topological band from a series of measurements of integrated spin textures, which precisely render spin textures on the parameter-tuned magnetic-group-symmetric planes. The detection technique can be generally applied to explore 3D topological states of similar symmetries. Furthermore, we observe the band inversion lines from topological quench dynamics, which are bulk counterparts of Fermi arc states and connect the Dirac points, reconfirming the realized topological band. Our results demonstrate the first approach to effectively observe 3D band topology, and open the way to probe exotic topological physics for ultracold atoms in high dimensions.

cond-mat.quant-gas↗

arXiv2019

Correlated Anomaly Detection from Large Streaming Data

Correlated anomaly detection (CAD) from streaming data is a type of group anomaly detection and an essential task in useful real-time data mining applications like botnet detection, financial event detection, industrial process monitor, etc. The primary approach for this type of detection in previous researches is based on principal score (PS) of divided batches or sliding windows by computing top eigenvalues of the correlation matrix, e.g. the Lanczos algorithm. However, this paper brings up the phenomenon of principal score degeneration for large data set, and then mathematically and practically prove current PS-based methods are likely to fail for CAD on large-scale streaming data even if the number of correlated anomalies grows with the data size at a reasonable rate; in reality, anomalies tend to be the minority of the data, and this issue can be more serious. We propose a framework with two novel randomized algorithms rPS and gPS for better detection of correlated anomalies from large streaming data of various correlation strength. The experiment shows high and balanced recall and estimated accuracy of our framework for anomaly detection from a large server log data set and a U.S. stock daily price data set in comparison to direct principal score evaluation and some other recent group anomaly detection algorithms. Moreover, our techniques significantly improve the computation efficiency and scalability for principal score calculation.

cs.LG↗