arXiv subjects

Wang Zhen

Publications and source records attributed to Wang Zhen.

7 recordsLinked to original sources

arXiv2024

Diffusion Model from Scratch

Diffusion generative models are currently the most popular generative models. However, their underlying modeling process is quite complex, and starting directly with the seminal paper Denoising Diffusion Probability Model (DDPM) can be challenging. This paper aims to assist readers in building a foundational understanding of generative models by tracing the evolution from VAEs to DDPM through detailed mathematical derivations and a problem-oriented analytical approach. It also explores the core ideas and improvement strategies of current mainstream methodologies, providing guidance for undergraduate and graduate students interested in learning about diffusion models.

cs.CV↗

arXiv2022

Consistency of Implicit and Explicit Features Matters for Monocular 3D Object Detection

Low-cost autonomous agents including autonomous driving vehicles chiefly adopt monocular 3D object detection to perceive surrounding environment. This paper studies 3D intermediate representation methods which generate intermediate 3D features for subsequent tasks. For example, the 3D features can be taken as input for not only detection, but also end-to-end prediction and/or planning that require a bird's-eye-view feature representation. In the study, we found that in generating 3D representation previous methods do not maintain the consistency between the objects' implicit poses in the latent space, especially orientations, and the explicitly observed poses in the Euclidean space, which can substantially hurt model performance. To tackle this problem, we present a novel monocular detection method, the first one being aware of the poses to purposefully guarantee that they are consistent between the implicit and explicit features. Additionally, we introduce a local ray attention mechanism to efficiently transform image features to voxels at accurate 3D locations. Thirdly, we propose a handcrafted Gaussian positional encoding function, which outperforms the sinusoidal encoding function while retaining the benefit of being continuous. Results show that our method improves the state-of-the-art 3D intermediate representation method by 3.15%. We are ranked 1st among all the reported monocular methods on both 3D and BEV detection benchmark on KITTI leaderboard as of th result's submission time.

cs.CV↗

arXiv2019

Arbitrarily Weak Nonlinearity Can Destroy the Anderson Localization

Whether the Anderson localization can survive from the weak enough nonlinear interaction is still an open question. In this Letter, we study the effect of nonlinear interaction on disordered chain based on the wave turbulence theory. It is found that the equipartition time $T_{eq}$ is inversely proportional to the square of the nonlinearity strength $λ$, i.e., $T_{eq}\proptoλ^{-2}$, in thermodynamic limit. This result has two fundamentally important consequences. First, the Anderson localized modes can not survive from arbitrarily weak nonlinearity. Secondly, contrary to popular belief, disorder can lead to a more fast thermalization in the weak nonlinear region, which is due to the emergence of three-wave resonance.

cond-mat.stat-mech↗

arXiv2013

Microstructure and Structural Phase Transitions in Iron-Based Superconductors

Crystal structures and microstructural features, such as structural phase transitions, defect structures, chemical and structural inhomogeneities, are known to have profound effects on the physical properties of superconducting materials. Recently, a large amount of works on the structural properties of Fe-based high-Tc superconductors have been published. This review article will mainly focus on typical microstructural features in samples that have been well characterized by physical measurements. (a) Certain common structural features are discussed. In particular, crystal structural features for different superconducting families, local structural distortions in the Fe2Pn2 (Pn=P, As, Sb) or Fe2Ch2 (Ch=S, Se, Te) blocks, and structural transformations in the 122 system. (b) In FeTe(Se) (11-family), the superconductivity, chemical and structural inhomogeneities are investigated and discussed in correlation with superconductivity. (c) In K0.8Fe1.6+xSe2 system, we focus on typical compounds with emphasis on Fe-vacancy order and phase separations. The microstructural features in other superconducting materials are also briefly discussed.

cond-mat.supr-con↗

arXiv1996

System and Its Uncertainty Quanta: Theory of General System (I)

The concept of uncertainty quanta for a general system is introduced and applied to some important problems in physics and mathematics. EPR paradox gives new clue to the further understanding of particle correlation which turns out to be the nature of this world. Randomness in quantum mechanics, statistical physics and chaos is integrated. A picture for a new kind of mathematics is put forward.

quant-ph↗

arXiv1996

Where Has Entropy Gone: Theory of General System (II)

A pair of symmetric expressions for the second law of thermodynamics is put forward. The conservation and transfer of entropy is discussed and applied to problems like biology, culture and life itself. A new explanation is given to the cosmic expansion with the concept of diversity in this theory. The problem of contingency and necessity is also discussed.

quant-ph↗

arXiv1996

Quantum Cosmology: Theory of General System (III)

The concepts of the perfect system and degeneracy are introduced. A special symmetry is found which is related to the entropy invariant. The inversion relation of system is obtained which is used to give the oppsite direction of time to classical sencond law of thermodanymics. The nature of time is discussed together with causality relation. A new understanding of quantum mechanics is put forward which describes a new picture of the world.

quant-ph↗