Searcharxiv⌕ Search

arXiv subjects

Yang Bai

Publications and source records attributed to Yang Bai.

At least 181 records · Page 10Linked to original sources

150 Years of Return Predictability Around the World: A Holistic View

Using new annual data of 16 developed countries across bond, equity, and housing markets, I study the return predictability using the payout-price ratios, i.e., coupon price, dividend price, and rent price. None of the 48 country-asset combinations shows consistent in-sample and out-of-sample performance with positive utility gain for the mean-variance investor. Only 3 (4/2) countries show positive economic gains in their equity (housing/bond) markets. The return predictability for the representative agents' risky asset portfolios and wealth portfolios is even weaker, suggesting that timing the investment return of a country using payout-price ratios will not make the investors better off. The predictive regressions based on the VAR analysis by Cochrane (2008, 2011) suggest that 14 (5) countries have predictable payout growth in the equity (housing) markets, ex., the dividend price predicts the dividend growth in the US. The VAR simulation using data from all the countries does not reject the null that the dividend growth is predictable. This paper presents firm evidence against the return predictability based on payout ratios.

q-fin.GN↗

Origin of nontopological soliton dark matter: solitosynthesis or phase transition

This work demonstrates that nontopological solitons with large global charges and masses, even above the Planck scale, can form in the early universe and dominate the dark matter abundance. In solitosynthesis, solitons prefer to grow as large as possible under equilibrium dynamics when an initial global charge asymmetry is present. Their abundance is set by when soliton formation via particle fusion freezes out, and their charges are set by the time it takes to accumulate free particles. This work improves the estimation of both quantities, and in particular shows that much larger-charged solitons form than previously thought. The results are estimated analytically and validated numerically by solving the coupled Boltzmann equations. Without solitosynthesis, phase transitions can still form solitons from particles left inside false-vacuum pockets and determine their present-day abundance and properties. Even with zero charge asymmetry, solitons formed in this way can have very large charges on account of statistical fluctuations in the numbers of (anti)particles inside each pocket.

hep-ph↗

Action Quality Assessment with Temporal Parsing Transformer

Action Quality Assessment(AQA) is important for action understanding and resolving the task poses unique challenges due to subtle visual differences. Existing state-of-the-art methods typically rely on the holistic video representations for score regression or ranking, which limits the generalization to capture fine-grained intra-class variation. To overcome the above limitation, we propose a temporal parsing transformer to decompose the holistic feature into temporal part-level representations. Specifically, we utilize a set of learnable queries to represent the atomic temporal patterns for a specific action. Our decoding process converts the frame representations to a fixed number of temporally ordered part representations. To obtain the quality score, we adopt the state-of-the-art contrastive regression based on the part representations. Since existing AQA datasets do not provide temporal part-level labels or partitions, we propose two novel loss functions on the cross attention responses of the decoder: a ranking loss to ensure the learnable queries to satisfy the temporal order in cross attention and a sparsity loss to encourage the part representations to be more discriminative. Extensive experiments show that our proposed method outperforms prior work on three public AQA benchmarks by a considerable margin.

cs.CV↗

Watermark Vaccine: Adversarial Attacks to Prevent Watermark Removal

As a common security tool, visible watermarking has been widely applied to protect copyrights of digital images. However, recent works have shown that visible watermarks can be removed by DNNs without damaging their host images. Such watermark-removal techniques pose a great threat to the ownership of images. Inspired by the vulnerability of DNNs on adversarial perturbations, we propose a novel defence mechanism by adversarial machine learning for good. From the perspective of the adversary, blind watermark-removal networks can be posed as our target models; then we actually optimize an imperceptible adversarial perturbation on the host images to proactively attack against watermark-removal networks, dubbed Watermark Vaccine. Specifically, two types of vaccines are proposed. Disrupting Watermark Vaccine (DWV) induces to ruin the host image along with watermark after passing through watermark-removal networks. In contrast, Inerasable Watermark Vaccine (IWV) works in another fashion of trying to keep the watermark not removed and still noticeable. Extensive experiments demonstrate the effectiveness of our DWV/IWV in preventing watermark removal, especially on various watermark removal networks.

cs.CV↗

Secure Network Function Computation for Linear Functions -- Part I: Source Security

In this paper, we put forward secure network function computation over a directed acyclic network. In such a network, a sink node is required to compute with zero error a target function of which the inputs are generated as source messages at multiple source nodes, while a wiretapper, who can access any one but not more than one wiretap set in a given collection of wiretap sets, is not allowed to obtain any information about a security function of the source messages. The secure computing capacity for the above model is defined as the maximum average number of times that the target function can be securely computed with zero error at the sink node with the given collection of wiretap sets and security function for one use of the network. The characterization of this capacity is in general overwhelmingly difficult. In the current paper, we consider securely computing linear functions with a wiretapper who can eavesdrop any subset of edges up to a certain size r, referred to as the security level, with the security function being the identity function. We first prove an upper bound on the secure computing capacity, which is applicable to arbitrary network topologies and arbitrary security levels. When the security level r is equal to 0, our upper bound reduces to the computing capacity without security consideration. We discover the surprising fact that for some models, there is no penalty on the secure computing capacity compared with the computing capacity without security consideration. We further obtain an equivalent expression of the upper bound by using a graph-theoretic approach, and accordingly we develop an efficient approach for computing this bound. Furthermore, we present a construction of linear function-computing secure network codes and obtain a lower bound on the secure computing capacity.

cs.IT↗

IceCube at the Frontier of Macroscopic Dark Matter Direct Detection

For a class of macroscopic dark matter models, inelastic scattering of dark matter off a nucleus can generate electromagnetic signatures with GeV-scale energy. The IceCube detector, with its kilometer-scale size, is ideal for directly detecting such inelastic scattering. Based on the slow particle trigger for the DeepCore detector, we perform a detailed signal and background simulation to estimate the discovery potential. For order 1 GeV deposited energy in each interaction, we find that IceCube can probe the dark matter masses up to one gram.

hep-ph↗

Residual Local Feature Network for Efficient Super-Resolution

Deep learning based approaches has achieved great performance in single image super-resolution (SISR). However, recent advances in efficient super-resolution focus on reducing the number of parameters and FLOPs, and they aggregate more powerful features by improving feature utilization through complex layer connection strategies. These structures may not be necessary to achieve higher running speed, which makes them difficult to be deployed to resource-constrained devices. In this work, we propose a novel Residual Local Feature Network (RLFN). The main idea is using three convolutional layers for residual local feature learning to simplify feature aggregation, which achieves a good trade-off between model performance and inference time. Moreover, we revisit the popular contrastive loss and observe that the selection of intermediate features of its feature extractor has great influence on the performance. Besides, we propose a novel multi-stage warm-start training strategy. In each stage, the pre-trained weights from previous stages are utilized to improve the model performance. Combined with the improved contrastive loss and training strategy, the proposed RLFN outperforms all the state-of-the-art efficient image SR models in terms of runtime while maintaining both PSNR and SSIM for SR. In addition, we won the first place in the runtime track of the NTIRE 2022 efficient super-resolution challenge. Code will be available at https://github.com/fyan111/RLFN.

cs.CV↗

Stochastic threshold in cell size control

Classic models of cell size control consider cells divide while reaching a threshold, e.g. size, age, or size extension. The molecular basis of the threshold involves multiple layers of regulation as well as gene noises. In this work, we study cell cycle as first-passage problem with stochastic threshold and discover such stochasticity affects the inter-division statistics, which bewilders the criteria to distinguish the types of size control models. The analytic results show the autocorrelation in the threshold can drive a sizer model to the adder-like and even timer-like inter-division statistics, which is supported by simulations. Following the picture that the autocorrelation in the threshold can propagate to the inter-division statistics, we further show that the adder model can be driven to the timer-like one by positive autocorrelated threshold, and even to the sizer-like one when the threshold is negatively autocorrelated. This work highlights the importance to examine gene noise in size control.

cond-mat.stat-mech↗

Adaptive Frequency Learning in Two-branch Face Forgery Detection

Face forgery has attracted increasing attention in recent applications of computer vision. Existing detection techniques using the two-branch framework benefit a lot from a frequency perspective, yet are restricted by their fixed frequency decomposition and transform. In this paper, we propose to Adaptively learn Frequency information in the two-branch Detection framework, dubbed AFD. To be specific, we automatically learn decomposition in the frequency domain by introducing heterogeneity constraints, and propose an attention-based module to adaptively incorporate frequency features into spatial clues. Then we liberate our network from the fixed frequency transforms, and achieve better performance with our data- and task-dependent transform layers. Extensive experiments show that AFD generally outperforms.

cs.CV↗

More Than Reading Comprehension: A Survey on Datasets and Metrics of Textual Question Answering

Textual Question Answering (QA) aims to provide precise answers to user's questions in natural language using unstructured data. One of the most popular approaches to this goal is machine reading comprehension(MRC). In recent years, many novel datasets and evaluation metrics based on classical MRC tasks have been proposed for broader textual QA tasks. In this paper, we survey 47 recent textual QA benchmark datasets and propose a new taxonomy from an application point of view. In addition, We summarize 8 evaluation metrics of textual QA tasks. Finally, we discuss current trends in constructing textual QA benchmarks and suggest directions for future work.

cs.CL↗

Hierarchical nature of hydrogen-based direct reduction of iron oxides

Fossil-free ironmaking is indispensable for reducing massive anthropogenic CO2 emissions in the steel industry. Hydrogen-based direct reduction (HyDR) is among the most attractive solutions for green ironmaking, with high technology readiness. The underlying mechanisms governing this process are characterized by a complex interaction of several chemical (phase transformations), physical (transport), and mechanical (stresses) phenomena. Their interplay leads to rich microstructures, characterized by a hierarchy of defects ranging across several orders of magnitude in length, including vacancies, dislocations, internal interfaces, and free surfaces in the form of cracks and pores. These defects can all act as reaction, nucleation, and diffusion sites, shaping the overall reduction kinetics. A clear understanding of the roles and interactions of these dynamically-evolving nano-/microstructure features is missing. Gaining better insights in these effects could enable improved access to the microstructure-based design of more efficient HyDR methods, with potentially high impact on the urgently needed decarbonization in the steel industry.

cond-mat.mtrl-sci↗

Improving Adversarial Robustness via Channel-wise Activation Suppressing

The study of adversarial examples and their activation has attracted significant attention for secure and robust learning with deep neural networks (DNNs). Different from existing works, in this paper, we highlight two new characteristics of adversarial examples from the channel-wise activation perspective: 1) the activation magnitudes of adversarial examples are higher than that of natural examples; and 2) the channels are activated more uniformly by adversarial examples than natural examples. We find that the state-of-the-art defense adversarial training has addressed the first issue of high activation magnitudes via training on adversarial examples, while the second issue of uniform activation remains. This motivates us to suppress redundant activation from being activated by adversarial perturbations via a Channel-wise Activation Suppressing (CAS) strategy. We show that CAS can train a model that inherently suppresses adversarial activation, and can be easily applied to existing defense methods to further improve their robustness. Our work provides a simple but generic training strategy for robustifying the intermediate layer activation of DNNs.

cs.LG↗

Moses: Efficient Exploitation of Cross-device Transferable Features for Tensor Program Optimization

Achieving efficient execution of machine learning models has attracted significant attention recently. To generate tensor programs efficiently, a key component of DNN compilers is the cost model that can predict the performance of each configuration on specific devices. However, due to the rapid emergence of hardware platforms, it is increasingly labor-intensive to train domain-specific predictors for every new platform. Besides, current design of cost models cannot provide transferable features between different hardware accelerators efficiently and effectively. In this paper, we propose Moses, a simple and efficient design based on the lottery ticket hypothesis, which fully takes advantage of the features transferable to the target device via domain adaptation. Compared with state-of-the-art approaches, Moses achieves up to 1.53X efficiency gain in the search stage and 1.41X inference speedup on challenging DNN benchmarks.

cs.LG↗

Nucleus Capture by Macroscopic Dark Matter

For a class of macroscopic dark matter with a large interaction strength with Standard Model particles, a nucleus could be captured by the dense, heavy dark matter as it traverses ordinary material. The radiated photon carries most of the binding energy and is a characteristic signature for dark matter detection. We develop analytic formulas and present numerical results for this radiative capture process in the low energy, non-dipole limit. Large-volume neutrino detectors like NO$ν$A, JUNO, DUNE and Super(Hyper)-K may detect multi-hit or single-hit radiative capture events and can search for dark matter up to one gram in mass.

hep-ph↗

Automated Customization of On-Thing Inference for Quality-of-Experience Enhancement

The rapid uptake of intelligent applications is pushing deep learning (DL) capabilities to Internet-of-Things (IoT). Despite the emergence of new tools for embedding deep neural networks (DNNs) into IoT devices, providing satisfactory Quality of Experience (QoE) to users is still challenging due to the heterogeneity in DNN architectures, IoT devices, and user preferences. This paper studies automated customization for DL inference on IoT devices (termed as on-thing inference), and our goal is to enhance user QoE by configuring the on-thing inference with an appropriate DNN for users under different usage scenarios. The core of our method is a DNN selection module that learns user QoE patterns on-the-fly and identifies the best-fit DNN for on-thing inference with the learned knowledge. It leverages a novel online learning algorithm, NeuralUCB, that has excellent generalization ability for handling various user QoE patterns. We also embed the knowledge transfer technique in NeuralUCB to expedite the learning process. However, NeuralUCB frequently solicits QoE ratings from users, which incurs non-negligible inconvenience. To address this problem, we design feedback solicitation schemes to reduce the number of QoE solicitations while maintaining the learning efficiency of NeuralUCB. A pragmatic problem, aggregated QoE, is further investigated to improve the practicality of our framework. We conduct experiments on both synthetic and real-world data. The results indicate that our method efficiently learns the user QoE pattern with few solicitations and provides drastic QoE enhancement for IoT devices.

cs.LG↗

Muon g-2 from Millicharged Hidden Confining Sector

We provide a novel explanation to the muon $g-2$ excess with new physics contributions at the two-loop level. In this scenario, light millicharged particles are introduced to modify the photon vacuum polarization that contributes to muon $g-2$ at one additional loop. The muon $g-2$ excess can be explained with the millicharged particle mass $m_χ$ around 10 MeV and the product of the multiplicity factor and millicharge squared of $N_χ\varepsilon^2 \sim 10^{-3}$. The minimal model faces severe constraints from direct searches at fixed-target experiments and astrophysical observables. However, if the millicharged particles are also charged under a hidden confining gauge group $SU(N_χ)$ with a confinement scale of MeV, hidden-sector hadrons are unstable and can decay into neutrinos, which makes this scenario consistent with existing constraints. This explanation can be well tested at low-energy lepton colliders such as BESIII and Belle II as well as other proposed fixed-target experiments.

hep-ph↗

Clustering Effect of (Linearized) Adversarial Robust Models

Adversarial robustness has received increasing attention along with the study of adversarial examples. So far, existing works show that robust models not only obtain robustness against various adversarial attacks but also boost the performance in some downstream tasks. However, the underlying mechanism of adversarial robustness is still not clear. In this paper, we interpret adversarial robustness from the perspective of linear components, and find that there exist some statistical properties for comprehensively robust models. Specifically, robust models show obvious hierarchical clustering effect on their linearized sub-networks, when removing or replacing all non-linear components (e.g., batch normalization, maximum pooling, or activation layers). Based on these observations, we propose a novel understanding of adversarial robustness and apply it on more tasks including domain adaption and robustness boosting. Experimental evaluations demonstrate the rationality and superiority of our proposed clustering strategy.

cs.LG↗

Chemo-Mechanical Phase-Field Modeling of Iron Oxide Reduction with Hydrogen

The reduction of iron ore with carbon-carriers is one of the largest sources of greenhouse gas emissions in the industry, motivating global activities to replace the coke-based blast furnace reduction by hydrogen-based direct reduction (HyDR). Iron oxide reduction with hydrogen has been widely investigated both experimentally and theoretically. The process includes multiple types of chemical reactions, solid state and defect-mediated diffusion (by oxygen and hydrogen species), several phase transformations, as well as massive volume shrinkage and mechanical stress buildup. In this work, a chemo-mechanically coupled phase-field (PF) model has been developed to explore the interplay between phase transformation, chemical reaction, species diffusion, large elasto-plastic deformation and microstructure evolution. Energetic constitutive relations of the model are based on the system free energy which is calibrated with the help of a thermodynamic database. The model has been first applied to the classical core-shell (wüstite-iron) structure. Simulations show that the phase transformation from wüstite to alpha-iron can result in high stress and rapidly decelerating reaction kinetics. Mechanical stresses can contribute elastic energy to the system, making phase transformation difficult. Thus slow reaction kinetics and low metallization are observed. However, if the stress becomes comparatively high, it can shift the shape of the free energy from a double-well to a single-well case and speed up the transformation. The model has been further applied to simulate an actual iron oxide specimen with its complex microstructure, characterized by electron microscopy. The simulation results show that isolated pores in the microstructure are filled with water vapor during reduction, an effect which influences the local reaction atmosphere and dynamics.

cond-mat.mtrl-sci↗