arXiv subjects

Chen Bai

Publications and source records attributed to Chen Bai.

30 records · Page 2Linked to original sources

arXiv2025

NavigScene: Bridging Local Perception and Global Navigation for Beyond-Visual-Range Autonomous Driving

Autonomous driving systems have made significant advances in Q&A, perception, prediction, and planning based on local visual information, yet they struggle to incorporate broader navigational context that human drivers routinely utilize. We address this critical gap between local sensor data and global navigation information by proposing NavigScene, an auxiliary navigation-guided natural language dataset that simulates a human-like driving environment within autonomous driving systems. Moreover, we develop three complementary paradigms to leverage NavigScene: (1) Navigation-guided Reasoning, which enhances vision-language models by incorporating navigation context into the prompting approach; (2) Navigation-guided Preference Optimization, a reinforcement learning method that extends Direct Preference Optimization to improve vision-language model responses by establishing preferences for navigation-relevant summarized information; and (3) Navigation-guided Vision-Language-Action model, which integrates navigation guidance and vision-language models with conventional driving models through feature fusion. Extensive experiments demonstrate that our approaches significantly improve performance across perception, prediction, planning, and question-answering tasks by enabling reasoning capabilities beyond visual range and improving generalization to diverse driving scenarios. This work represents a significant step toward more comprehensive autonomous driving systems capable of navigating complex, unfamiliar environments with greater reliability and safety.

cs.RO↗

arXiv2024

Entanglement dynamics in 2d HCFTs on the curved background: the case of q-Möbius Hamiltonian

We will explore the dynamical property of non-equilibrium phenomena induced by two-dimensional holographic conformal field theory (2d holographic CFT) Hamiltonian on the curved spacetime by studying the time dependence of the entanglement entropy and mutual information. Here, holographic CFT is the CFT having the gravity dual. We will start from the boundary and thermofield double states, evolve the systems in Euclidean time with the Hamiltonian on the curved background, and then evolve them in real-time with the same Hamiltonian. We found that the early- and late-time entanglement structure depends on the curved background, while the entanglement growth does not, and is linear. Furthermore, in the gravity dual for the thermofield double state, this entanglement growth is due to the linear growth of the wormhole, while in the one for the boundary state, it is due to the in-falling of the end of the world brane to the black hole. We discussed the low temperature system can be regarded as the dynamical system induced by the multi-joining quenches. We also discussed the effective description of the high temperature system, called line tension picture.

hep-th↗

arXiv2024

On the Two-sided Permutation Inversion Problem

In the permutation inversion problem, the task is to find the preimage of some challenge value, given oracle access to the permutation. This is a fundamental problem in query complexity, and appears in many contexts, particularly cryptography. In this work, we examine the setting in which the oracle allows for quantum queries to both the forward and the inverse direction of the permutation -- except that the challenge value cannot be submitted to the latter. Within that setting, we consider two options for the inversion algorithm: whether it can get quantum advice about the permutation, and whether it must produce the entire preimage (search) or only the first bit (decision). We prove several theorems connecting the hardness of the resulting variations of the inversion problem, and establish a number of lower bounds. Our results indicate that, perhaps surprisingly, the inversion problem does not become significantly easier when the adversary is granted oracle access to the inverse, provided it cannot query the challenge itself.

quant-ph↗

arXiv2024

GNNavigator: Towards Adaptive Training of Graph Neural Networks via Automatic Guideline Exploration

Graph Neural Networks (GNNs) succeed significantly in many applications recently. However, balancing GNNs training runtime cost, memory consumption, and attainable accuracy for various applications is non-trivial. Previous training methodologies suffer from inferior adaptability and lack a unified training optimization solution. To address the problem, this work proposes GNNavigator, an adaptive GNN training configuration optimization framework. GNNavigator meets diverse GNN application requirements due to our unified software-hardware co-abstraction, proposed GNNs training performance model, and practical design space exploration solution. Experimental results show that GNNavigator can achieve up to 3.1x speedup and 44.9% peak memory reduction with comparable accuracy to state-of-the-art approaches.

cs.LG↗

arXiv2024

Anything in Any Scene: Photorealistic Video Object Insertion

Realistic video simulation has shown significant potential across diverse applications, from virtual reality to film production. This is particularly true for scenarios where capturing videos in real-world settings is either impractical or expensive. Existing approaches in video simulation often fail to accurately model the lighting environment, represent the object geometry, or achieve high levels of photorealism. In this paper, we propose Anything in Any Scene, a novel and generic framework for realistic video simulation that seamlessly inserts any object into an existing dynamic video with a strong emphasis on physical realism. Our proposed general framework encompasses three key processes: 1) integrating a realistic object into a given scene video with proper placement to ensure geometric realism; 2) estimating the sky and environmental lighting distribution and simulating realistic shadows to enhance the light realism; 3) employing a style transfer network that refines the final video output to maximize photorealism. We experimentally demonstrate that Anything in Any Scene framework produces simulated videos of great geometric realism, lighting realism, and photorealism. By significantly mitigating the challenges associated with video data generation, our framework offers an efficient and cost-effective solution for acquiring high-quality videos. Furthermore, its applications extend well beyond video data augmentation, showing promising potential in virtual reality, video editing, and various other video-centric applications. Please check our project website https://anythinginanyscene.github.io for access to our project code and more high-resolution video results.

cs.CV↗

arXiv2023

SoC-Tuner: An Importance-guided Exploration Framework for DNN-targeting SoC Design

Designing a system-on-chip (SoC) for deep neural network (DNN) acceleration requires balancing multiple metrics such as latency, power, and area. However, most existing methods ignore the interactions among different SoC components and rely on inaccurate and error-prone evaluation tools, leading to inferior SoC design. In this paper, we present SoC-Tuner, a DNN-targeting exploration framework to find the Pareto optimal set of SoC configurations efficiently. Our framework constructs a thorough SoC design space of all components and divides the exploration into three phases. We propose an importance-based analysis to prune the design space, a sampling algorithm to select the most representative initialization points, and an information-guided multi-objective optimization method to balance multiple design metrics of SoC design. We validate our framework with the actual very-large-scale-integration (VLSI) flow on various DNN benchmarks and show that it outperforms previous methods. To the best of our knowledge, this is the first work to construct an exploration framework of SoCs for DNN acceleration.

cs.AR↗

arXiv2023

Energy-efficient Integrated Sensing and Communication System and DNLFM Waveform

Integrated sensing and communication (ISAC) is a key enabler of 6G. Unlike communication radio links, the sensing signal requires to experience round trips from many scatters. Therefore, sensing is more power-sensitive and faces a severer multi-target interference. In this paper, the ISAC system employs dedicated sensing signals, which can be reused as the communication reference signal. This paper proposes to add time-frequency matched windows at both the transmitting and receiving sides, which avoids mismatch loss and increases energy efficiency. Discrete non-linear frequency modulation (DNLFM) is further proposed to achieve both time-domain constant modulus and frequency-domain arbitrary windowing weights. DNLFM uses very few Newton iterations and a simple geometrically-equivalent method to generate, which greatly reduces the complex numerical integral in the conventional method. Moreover, the spatial-domain matched window is proposed to achieve low sidelobes. The simulation results show that the proposed methods gain a higher energy efficiency than conventional methods.

eess.SP↗

arXiv2023

Dynamic Structural Brain Network Construction by Hierarchical Prototype Embedding GCN using T1-MRI

Constructing structural brain networks using T1-weighted magnetic resonance imaging (T1-MRI) presents a significant challenge due to the lack of direct regional connectivity information. Current methods with T1-MRI rely on predefined regions or isolated pretrained location modules to obtain atrophic regions, which neglects individual specificity. Besides, existing methods capture global structural context only on the whole-image-level, which weaken correlation between regions and the hierarchical distribution nature of brain connectivity.We hereby propose a novel dynamic structural brain network construction method based on T1-MRI, which can dynamically localize critical regions and constrain the hierarchical distribution among them for constructing dynamic structural brain network. Specifically, we first cluster spatially-correlated channel and generate several critical brain regions as prototypes. Further, we introduce a contrastive loss function to constrain the prototypes distribution, which embed the hierarchical brain semantic structure into the latent space. Self-attention and GCN are then used to dynamically construct hierarchical correlations of critical regions for brain network and explore the correlation, respectively. Our method is evaluated on ADNI-1 and ADNI-2 databases for mild cognitive impairment (MCI) conversion prediction, and acheive the state-of-the-art (SOTA) performance. Our source code is available at http://github.com/*******.

cs.CV↗

arXiv2022

3DG-STFM: 3D Geometric Guided Student-Teacher Feature Matching

We tackle the essential task of finding dense visual correspondences between a pair of images. This is a challenging problem due to various factors such as poor texture, repetitive patterns, illumination variation, and motion blur in practical scenarios. In contrast to methods that use dense correspondence ground-truths as direct supervision for local feature matching training, we train 3DG-STFM: a multi-modal matching model (Teacher) to enforce the depth consistency under 3D dense correspondence supervision and transfer the knowledge to 2D unimodal matching model (Student). Both teacher and student models consist of two transformer-based matching modules that obtain dense correspondences in a coarse-to-fine manner. The teacher model guides the student model to learn RGB-induced depth information for the matching purpose on both coarse and fine branches. We also evaluate 3DG-STFM on a model compression task. To the best of our knowledge, 3DG-STFM is the first student-teacher learning method for the local feature matching task. The experiments show that our method outperforms state-of-the-art methods on indoor and outdoor camera pose estimations, and homography estimation problems. Code is available at: https://github.com/Ryan-prime/3DG-STFM.

cs.CV↗

arXiv2022

Toward the nonequilibrium thermodynamic analog of complexity and the Jarzynski identity

The Jarzynski identity can describe small-scale nonequilibrium systems through stochastic thermodynamics. The identity considers fluctuating trajectories in a phase space. The complexity geometry frames the discussions on quantum computational complexity using the method of Riemannian geometry, which builds a bridge between optimal quantum circuits and classical geodesics in the space of unitary operators. Complexity geometry enables the application of the methods of classical physics to deal with pure quantum problems. By combining the two frameworks, i.e., the Jarzynski identity and complexity geometry, we derived a complexity analog of the Jarzynski identity using the complexity geometry. We considered a set of geodesics in the space of unitary operators instead of the trajectories in a phase space. The obtained complexity version of the Jarzynski identity strengthened the evidence for the existence of a well-defined resource theory of uncomplexity and presented an extensive discussion on the second law of complexity. Furthermore, analogous to the thermodynamic fluctuation-dissipation theorem, we proposed a version of the fluctuation-dissipation theorem for the complexity. Although this study does not focus on holographic fluctuations, we found that the results are surprisingly suitable for capturing their information. The results obtained using nonequilibrium methods may contribute to understand the nature of the complexity and study the features of the holographic fluctuations.

hep-th↗

arXiv2021

Post-Quantum Security of the Even-Mansour Cipher

The Even-Mansour cipher is a simple method for constructing a (keyed) pseudorandom permutation $E$ from a public random permutation~$P:\{0,1\}^n \rightarrow \{0,1\}^n$. It is secure against classical attacks, with optimal attacks requiring $q_E$ queries to $E$ and $q_P$ queries to $P$ such that $q_E \cdot q_P \approx 2^n$. If the attacker is given \emph{quantum} access to both $E$ and $P$, however, the cipher is completely insecure, with attacks using $q_E, q_P = O(n)$ queries known. In any plausible real-world setting, however, a quantum attacker would have only \emph{classical} access to the keyed permutation~$E$ implemented by honest parties, even while retaining quantum access to~$P$. Attacks in this setting with $q_E \cdot q_P^2 \approx 2^n$ are known, showing that security degrades as compared to the purely classical case, but leaving open the question as to whether the Even-Mansour cipher can still be proven secure in this natural, "post-quantum" setting. We resolve this question, showing that any attack in that setting requires $q_E \cdot q^2_P + q_P \cdot q_E^2 \approx 2^n$. Our results apply to both the two-key and single-key variants of Even-Mansour. Along the way, we establish several generalizations of results from prior work on quantum-query lower bounds that may be of independent interest.

quant-ph↗

arXiv2019

Retrieval of non-sparse object through scattering media beyond the memory effect

Optical imaging through scattering media is a commonly confronted with the problem of reconstruction of complex objects and optical memory effect. To solve the problem, here, we propose a novel configuration based on the combination of ptychography and shower-curtain effect, which enables the retrieval of non-sparse samples through scattering media beyond the memory effect. Furthermore, by virtue of the shower-curtain effect, the proposed imaging system is insensitive to dynamic scattering media. Results from the retrieval of hair follicle section demonstrate the effectiveness and feasibility of the proposed method. The field of view is improved to 2.64mm. This present technique will be a potential approach for imaging through deep biological tissue.

eess.IV↗