arXiv subjects

Bin Zhang

Publications and source records attributed to Bin Zhang.

At least 127 records · Page 7Linked to original sources

arXiv2024

2DLIW-SLAM:2D LiDAR-Inertial-Wheel Odometry with Real-Time Loop Closure

Due to budgetary constraints, indoor navigation typically employs 2D LiDAR rather than 3D LiDAR. However, the utilization of 2D LiDAR in Simultaneous Localization And Mapping (SLAM) frequently encounters challenges related to motion degeneracy, particularly in geometrically similar environments. To address this problem, this paper proposes a robust, accurate, and multi-sensor-fused 2D LiDAR SLAM system specifically designed for indoor mobile robots. To commence, the original LiDAR data undergoes meticulous processing through point and line extraction. Leveraging the distinctive characteristics of indoor environments, line-line constraints are established to complement other sensor data effectively, thereby augmenting the overall robustness and precision of the system. Concurrently, a tightly-coupled front-end is created, integrating data from the 2D LiDAR, IMU, and wheel odometry, thus enabling real-time state estimation. Building upon this solid foundation, a novel global feature point matching-based loop closure detection algorithm is proposed. This algorithm proves highly effective in mitigating front-end accumulated errors and ultimately constructs a globally consistent map. The experimental results indicate that our system fully meets real-time requirements. When compared to Cartographer, our system not only exhibits lower trajectory errors but also demonstrates stronger robustness, particularly in degeneracy problem.

cs.RO↗

arXiv2024

PTDE: Personalized Training with Distilled Execution for Multi-Agent Reinforcement Learning

Centralized Training with Decentralized Execution (CTDE) has emerged as a widely adopted paradigm in multi-agent reinforcement learning, emphasizing the utilization of global information for learning an enhanced joint $Q$-function or centralized critic. In contrast, our investigation delves into harnessing global information to directly enhance individual $Q$-functions or individual actors. Notably, we discover that applying identical global information universally across all agents proves insufficient for optimal performance. Consequently, we advocate for the customization of global information tailored to each agent, creating agent-personalized global information to bolster overall performance. Furthermore, we introduce a novel paradigm named Personalized Training with Distilled Execution (PTDE), wherein agent-personalized global information is distilled into the agent's local information. This distilled information is then utilized during decentralized execution, resulting in minimal performance degradation. PTDE can be seamlessly integrated with state-of-the-art algorithms, leading to notable performance enhancements across diverse benchmarks, including the SMAC benchmark, Google Research Football (GRF) benchmark, and Learning to Rank (LTR) task.

cs.AI↗

arXiv2024

VN-Net: Vision-Numerical Fusion Graph Convolutional Network for Sparse Spatio-Temporal Meteorological Forecasting

Sparse meteorological forecasting is indispensable for fine-grained weather forecasting and deserves extensive attention. Recent studies have highlighted the potential of spatio-temporal graph convolutional networks (ST-GCNs) in predicting numerical data from ground weather stations. However, as one of the highest fidelity and lowest latency data, the application of the vision data from satellites in ST-GCNs remains unexplored. There are few studies to demonstrate the effectiveness of combining the above multi-modal data for sparse meteorological forecasting. Towards this objective, we introduce Vision-Numerical Fusion Graph Convolutional Network (VN-Net), which mainly utilizes: 1) Numerical-GCN (N-GCN) to adaptively model the static and dynamic patterns of spatio-temporal numerical data; 2) Vision-LSTM Network (V-LSTM) to capture multi-scale joint channel and spatial features from time series satellite images; 4) a GCN-based decoder to generate hourly predictions of specified meteorological factors. As far as we know, VN-Net is the first attempt to introduce GCN method to utilize multi-modal data for better handling sparse spatio-temporal meteorological forecasting. Our experiments on Weather2k dataset show VN-Net outperforms state-of-the-art by a significant margin on mean absolute error (MAE) and root mean square error (RMSE) for temperature, relative humidity, and visibility forecasting. Furthermore, we conduct interpretation analysis and design quantitative evaluation metrics to assess the impact of incorporating vision data.

cs.CV↗

arXiv2024

Time-resolved Spectral Properties of Fermi-GBM Bright Long Gamma-Ray Bursts

The prompt emission mechanism of gamma-ray bursts (GRBs) is still unclear, and the time-resolved spectral analysis of GRBs is a powerful tool for studying their underlying physical processes. We performed a detailed time-resolved spectral analysis of 78 bright long GRB samples detected by Fermi/Gamma-ray Burst Monitor (GBM). A total of 1490 spectra were obtained and their properties were studied using a typical Band-shape model. Firstly, the parameter distribution of the time-resolved spectrum given as follows: the low-energy spectral index $α\sim -0.72$, high-energy spectral index $β\sim -2.42$, the peak energy $E_{\rm p} \sim 221.69 \,\rm{keV}$, and the energy flux $F \sim 7.49\times 10^{-6} \rm{\, erg\,cm^{-2}\,s^{-1}}$. More than 80\% of the bursts exhibit the hardest low-energy spectral index $α_{\rm max}$ exceeding the synchrotron limit (-2/3). Secondly, the evolution patterns of $α$ and $E_{\rm p}$ were statistically analyzed. The results show that for multi-pulse GRBs the intensity-tracking pattern is more common than the hard-to-soft pattern in the evolution of both $E_{\rm p}$ and $α$. The hard-to-soft pattern is generally shown in single-pulse GRBs or in the initial pulse of multi-pulse GRBs. Finally, we found a significant positive correlation between $F$ and $E_{\rm p}$, with half of the samples exhibiting a positive correlation between $F$ and $α$. We discussed the spectral evolution of different radiation models. The diversity of spectral evolution patterns indicates that there may be more than one radiation mechanism occurring in the gamma-ray burst radiation process, including photospheric radiation and synchrotron radiation. However, it may also involve only one radiation mechanism, but more complicated physical details need to be considered.

astro-ph.HE↗

arXiv2024

Controlling Large Language Model-based Agents for Large-Scale Decision-Making: An Actor-Critic Approach

The remarkable progress in Large Language Models (LLMs) opens up new avenues for addressing planning and decision-making problems in Multi-Agent Systems (MAS). However, as the number of agents increases, the issues of hallucination in LLMs and coordination in MAS have become increasingly prominent. Additionally, the efficient utilization of tokens emerges as a critical consideration when employing LLMs to facilitate the interactions among a substantial number of agents. In this paper, we develop a modular framework called LLaMAC to mitigate these challenges. LLaMAC implements a value distribution encoding similar to that found in the human brain, utilizing internal and external feedback mechanisms to facilitate collaboration and iterative reasoning among its modules. Through evaluations involving system resource allocation and robot grid transportation, we demonstrate the considerable advantages afforded by our proposed approach.

cs.AI↗

arXiv2024

Modeling Spoof Noise by De-spoofing Diffusion and its Application in Face Anti-spoofing

Face anti-spoofing is crucial for ensuring the security and reliability of face recognition systems. Several existing face anti-spoofing methods utilize GAN-like networks to detect presentation attacks by estimating the noise pattern of a spoof image and recovering the corresponding genuine image. But GAN's limited face appearance space results in the denoised faces cannot cover the full data distribution of genuine faces, thereby undermining the generalization performance of such methods. In this work, we present a pioneering attempt to employ diffusion models to denoise a spoof image and restore the genuine image. The difference between these two images is considered as the spoof noise, which can serve as a discriminative cue for face anti-spoofing. We evaluate our proposed method on several intra-testing and inter-testing protocols, where the experimental results showcase the effectiveness of our method in achieving competitive performance in terms of both accuracy and generalization.

cs.CV↗

arXiv2024

Shape-Dependence of Spontaneous Photon Emission by Quantum Electron Wavepackets and the QED Origin of Bunched Electron Beam Superradiance

It has been shown that the spontaneous emission rate of photons by free electrons, unlike stimulated emission, is independent of the shape or modulation of the quantum electron wavefunction (QEW). Nevertheless, here we show that the quantum state of the emitted photons is non-classical and does depend on the QEW shape. This non-classicality originates from the shape dependent off-diagonal terms of the photon density matrix. This is manifested in the Wigner distribution function and would be observable experimentally through Homodyne detection techniques as a squeezing effect. Considering a scheme of electrons interaction with a single microcavity mode, we present a QED formulation of spontaneous emission by multiple modulated QEWs through a build-up process. Our findings indicate that in the case of a density modulated QEWs beam, the phase of the off-diagonal terms of the photon state emitted by the modulated QEWs is the harbinger of bunched beam superradiance, where the spontaneous emission is proportional to N_e^2. This observation offers a potential for enhancement of other quantum electron interactions with quantum systems by a modulated QEWs beam carrying coherence and quantum properties of the modulation.

quant-ph↗

arXiv2024

Online Boosting Adaptive Learning under Concept Drift for Multistream Classification

Multistream classification poses significant challenges due to the necessity for rapid adaptation in dynamic streaming processes with concept drift. Despite the growing research outcomes in this area, there has been a notable oversight regarding the temporal dynamic relationships between these streams, leading to the issue of negative transfer arising from irrelevant data. In this paper, we propose a novel Online Boosting Adaptive Learning (OBAL) method that effectively addresses this limitation by adaptively learning the dynamic correlation among different streams. Specifically, OBAL operates in a dual-phase mechanism, in the first of which we design an Adaptive COvariate Shift Adaptation (AdaCOSA) algorithm to construct an initialized ensemble model using archived data from various source streams, thus mitigating the covariate shift while learning the dynamic correlations via an adaptive re-weighting strategy. During the online process, we employ a Gaussian Mixture Model-based weighting mechanism, which is seamlessly integrated with the acquired correlations via AdaCOSA to effectively handle asynchronous drift. This approach significantly improves the predictive performance and stability of the target stream. We conduct comprehensive experiments on several synthetic and real-world data streams, encompassing various drifting scenarios and types. The results clearly demonstrate that OBAL achieves remarkable advancements in addressing multistream classification problems by effectively leveraging positive knowledge derived from multiple sources.

cs.LG↗

arXiv2023

PDiT: Interleaving Perception and Decision-making Transformers for Deep Reinforcement Learning

Designing better deep networks and better reinforcement learning (RL) algorithms are both important for deep RL. This work studies the former. Specifically, the Perception and Decision-making Interleaving Transformer (PDiT) network is proposed, which cascades two Transformers in a very natural way: the perceiving one focuses on \emph{the environmental perception} by processing the observation at the patch level, whereas the deciding one pays attention to \emph{the decision-making} by conditioning on the history of the desired returns, the perceiver's outputs, and the actions. Such a network design is generally applicable to a lot of deep RL settings, e.g., both the online and offline RL algorithms under environments with either image observations, proprioception observations, or hybrid image-language observations. Extensive experiments show that PDiT can not only achieve superior performance than strong baselines in different settings but also extract explainable feature representations. Our code is available at \url{https://github.com/maohangyu/PDiT}.

cs.LG↗

arXiv2023

Inducing Stackelberg Equilibrium through Spatio-Temporal Sequential Decision-Making in Multi-Agent Reinforcement Learning

In multi-agent reinforcement learning (MARL), self-interested agents attempt to establish equilibrium and achieve coordination depending on game structure. However, existing MARL approaches are mostly bound by the simultaneous actions of all agents in the Markov game (MG) framework, and few works consider the formation of equilibrium strategies via asynchronous action coordination. In view of the advantages of Stackelberg equilibrium (SE) over Nash equilibrium, we construct a spatio-temporal sequential decision-making structure derived from the MG and propose an N-level policy model based on a conditional hypernetwork shared by all agents. This approach allows for asymmetric training with symmetric execution, with each agent responding optimally conditioned on the decisions made by superior agents. Agents can learn heterogeneous SE policies while still maintaining parameter sharing, which leads to reduced cost for learning and storage and enhanced scalability as the number of agents increases. Experiments demonstrate that our method effectively converges to the SE policies in repeated matrix game scenarios, and performs admirably in immensely complex settings including cooperative tasks and mixed tasks.

cs.MA↗

arXiv2023

TPTU-v2: Boosting Task Planning and Tool Usage of Large Language Model-based Agents in Real-world Systems

Large Language Models (LLMs) have demonstrated proficiency in addressing tasks that necessitate a combination of task planning and the usage of external tools that require a blend of task planning and the utilization of external tools, such as APIs. However, real-world complex systems present three prevalent challenges concerning task planning and tool usage: (1) The real system usually has a vast array of APIs, so it is impossible to feed the descriptions of all APIs to the prompt of LLMs as the token length is limited; (2) the real system is designed for handling complex tasks, and the base LLMs can hardly plan a correct sub-task order and API-calling order for such tasks; (3) Similar semantics and functionalities among APIs in real systems create challenges for both LLMs and even humans in distinguishing between them. In response, this paper introduces a comprehensive framework aimed at enhancing the Task Planning and Tool Usage (TPTU) abilities of LLM-based agents operating within real-world systems. Our framework comprises three key components designed to address these challenges: (1) the API Retriever selects the most pertinent APIs for the user task among the extensive array available; (2) LLM Finetuner tunes a base LLM so that the finetuned LLM can be more capable for task planning and API calling; (3) the Demo Selector adaptively retrieves different demonstrations related to hard-to-distinguish APIs, which is further used for in-context learning to boost the final performance. We validate our methods using a real-world commercial system as well as an open-sourced academic dataset, and the outcomes clearly showcase the efficacy of each individual component as well as the integrated framework.

cs.AI↗

arXiv2023

DeepSpeed4Science Initiative: Enabling Large-Scale Scientific Discovery through Sophisticated AI System Technologies

In the upcoming decade, deep learning may revolutionize the natural sciences, enhancing our capacity to model and predict natural occurrences. This could herald a new era of scientific exploration, bringing significant advancements across sectors from drug development to renewable energy. To answer this call, we present DeepSpeed4Science initiative (deepspeed4science.ai) which aims to build unique capabilities through AI system technology innovations to help domain experts to unlock today's biggest science mysteries. By leveraging DeepSpeed's current technology pillars (training, inference and compression) as base technology enablers, DeepSpeed4Science will create a new set of AI system technologies tailored for accelerating scientific discoveries by addressing their unique complexity beyond the common technical approaches used for accelerating generic large language models (LLMs). In this paper, we showcase the early progress we made with DeepSpeed4Science in addressing two of the critical system challenges in structural biology research.

cs.AI↗

arXiv2023

Leader-Follower Formation Control of Perturbed Nonholonomic Agents along Parametric Curves with Directed Communication

In this paper, we propose a novel formation controller for nonholonomic agents to form general parametric curves. First, we derive a unified parametric representation for both open and closed curves. Then, a leader-follower formation controller is designed to form the parametric curves. We consider directed communications and constant input disturbances rejection in the controller design. Rigorous Lyapunov-based stability analysis proves the asymptotic stability of the proposed controller. Detailed numerical simulations and experimental studies are conducted to verify the performance of the proposed method.

cs.MA↗

arXiv2023

VITS-Based Singing Voice Conversion Leveraging Whisper and multi-scale F0 Modeling

This paper introduces the T23 team's system submitted to the Singing Voice Conversion Challenge 2023. Following the recognition-synthesis framework, our singing conversion model is based on VITS, incorporating four key modules: a prior encoder, a posterior encoder, a decoder, and a parallel bank of transposed convolutions (PBTC) module. We particularly leverage Whisper, a powerful pre-trained ASR model, to extract bottleneck features (BNF) as the input of the prior encoder. Before BNF extraction, we perform pitch perturbation to the source signal to remove speaker timbre, which effectively avoids the leakage of the source speaker timbre to the target. Moreover, the PBTC module extracts multi-scale F0 as the auxiliary input to the prior encoder, thereby capturing better pitch variations of singing. We design a three-stage training strategy to better adapt the base model to the target speaker with limited target speaker data. Official challenge results show that our system has superior performance in naturalness, ranking 1st and 2nd respectively in Task 1 and 2. Further ablation justifies the effectiveness of our system design.

eess.AS↗

arXiv2023

Automated Bioinformatics Analysis via AutoBA

With the fast-growing and evolving omics data, the demand for streamlined and adaptable tools to handle the analysis continues to grow. In response to this need, we introduce Auto Bioinformatics Analysis (AutoBA), an autonomous AI agent based on a large language model designed explicitly for conventional omics data analysis. AutoBA simplifies the analytical process by requiring minimal user input while delivering detailed step-by-step plans for various bioinformatics tasks. Through rigorous validation by expert bioinformaticians, AutoBA's robustness and adaptability are affirmed across a diverse range of omics analysis cases, including whole genome sequencing (WGS), RNA sequencing (RNA-seq), single-cell RNA-seq, ChIP-seq, and spatial transcriptomics. AutoBA's unique capacity to self-design analysis processes based on input data variations further underscores its versatility. Compared with online bioinformatic services, AutoBA deploys the analysis locally, preserving data privacy. Moreover, different from the predefined pipeline, AutoBA has adaptability in sync with emerging bioinformatics tools. Overall, AutoBA represents a convenient tool, offering robustness and adaptability for complex omics data analysis.

q-bio.GN↗

arXiv2023

Quantum Dynamical Approach to Predicting the Optical Pumping Threshold for Lasing in Organic Materials

We present a quantum dynamic study on organic lasing phenomena, which is a challenging issue in organic optoelectronics. Previously, phenomenological method has achieved success in describing experimental observation. However, it cannot directly bridge the laser threshold with molecular electronic structure parameters and cavity parameters. Quantum dynamics method for describing organic lasing and obtaining laser threshold is highly expected. In this Letter, we first propose a microscopic model suitable for describing the lasing dynamics of organic molecular system and we apply the time-dependent wave-packet diffusion (TDWPD) to reveal the microscopic quantum dynamical process for the optical pumped lasing behavior. Lasing threshold is obtained from the onset of output as a function of optical input pumping. We predict that the lasing threshold has an optimal value as function of the cavity volume and depends linearly on the intracavity photon leakage rate. The structure-property relationships between molecular electronic structure parameters (including the energy of molecular excited state, the transition dipole and the organization energy) and the laser threshold obtained through numerical calculations are in qualitative agreement the experimental results, which also confirms the reliability of our approach. This work is beneficial to understanding the mechanism of organic laser and optimizing the design of organic laser materials. TOC

physics.chem-ph↗

arXiv2023

Individual and Averaged Power Density Spectra of X-ray bursts from SGR J1935+2154: Quasiperiodic Oscillation Search and Slopes

The study of quasi-periodic oscillations (QPOs) and power density spectra (PDS) continuum properties can help shed light on the still illusive emission physics of magnetars and as a window into the interiors of neutron stars using asteroseismology. In this work, we employ a Bayesian method to search for the QPOs in the hundreds of X-ray bursts from SGR J1935+2154 observed by {\it Insight}-HXMT, GECAM and Fermi/GBM from July 2014 to January 2022. Although no definitive QPO signal (significance $>3σ$) is detected in individual bursts or the averaged periodogram of the bursts grouped by duration, we identify several bursts exhibiting possible QPO at $\sim$ 40 Hz, which is consistent with that reported in the X-ray burst associated with FRB 200428. We investigate the PDS continuum properties and find that the distribution of the PDS slope in the simple power-law model peaks $\sim$ 2.5, which is consistent with other magnetars but higher than 5/3 commonly seen in gamma-ray bursts. Besides, the distribution of the break frequency in the broken power-law model peaks at $\sim$ 60 Hz. Finally, we report that the power-law index of PDS has an anti-correlation and power-law dependence on the burst duration as well as the minimum variation timescale.

astro-ph.HE↗

arXiv2023

The Minimum Variation Timescales of X-ray bursts from SGR J1935+2154

The minimum variation timescale (MVT) of soft gamma-ray repeaters can be an important probe to estimate the emission region in pulsar-like models, as well as the Lorentz factor and radius of the possible relativistic jet in gamma-ray burst (GRB)-like models, thus revealing their progenitors and physical mechanisms. In this work, we systematically study the MVTs of hundreds of X-ray bursts (XRBs) from SGR J1935+2154 observed by {\it Insight}-HXMT, GECAM and Fermi/GBM from July 2014 to Jan 2022 through the Bayesian Block algorithm. We find that the MVTs peak at $\sim$ 2 ms, corresponding to a light travel time size of about 600 km, which supports the magnetospheric origin in pulsar-like models. The shock radius and the Lorentz factor of the jet are also constrained in GRB-like models. Interestingly, the MVT of the XRB associated with FRB 200428 is $\sim$ 70 ms, which is longer than that of most bursts and implies its special radiation mechanism. Besides, the median of MVTs is 7 ms, shorter than the median MVTs of 40 ms and 480 ms for short GRBs or long GRBs, respectively. However, the MVT is independent of duration, similar to GRBs. Finally, we investigate the energy dependence of MVT and suggest that there is a marginal evidence for a power-law relationship like GRBs but the rate of variation is at least about an order of magnitude smaller. These features may provide an approach to identify bursts with a magnetar origin.

astro-ph.HE↗