arXiv subjects

Zheng Chen

Publications and source records attributed to Zheng Chen.

At least 145 records · Page 8Linked to original sources

arXiv2024

Energy-Efficient Federated Edge Learning with Streaming Data: A Lyapunov Optimization Approach

Federated learning (FL) has received significant attention in recent years for its advantages in efficient training of machine learning models across distributed clients without disclosing user-sensitive data. Specifically, in federated edge learning (FEEL) systems, the time-varying nature of wireless channels introduces inevitable system dynamics in the communication process, thereby affecting training latency and energy consumption. In this work, we further consider a streaming data scenario where new training data samples are randomly generated over time at edge devices. Our goal is to develop a dynamic scheduling and resource allocation algorithm to address the inherent randomness in data arrivals and resource availability under long-term energy constraints. To achieve this, we formulate a stochastic network optimization problem and use the Lyapunov drift-plus-penalty framework to obtain a dynamic resource management design. Our proposed algorithm makes adaptive decisions on device scheduling, computational capacity adjustment, and allocation of bandwidth and transmit power in every round. We provide convergence analysis for the considered setting with heterogeneous data and time-varying objective functions, which supports the rationale behind our proposed scheduling design. The effectiveness of our scheme is verified through simulation results, demonstrating improved learning performance and energy efficiency as compared to baseline schemes.

cs.LG↗

arXiv2024

Temporal Predictive Coding for Gradient Compression in Distributed Learning

This paper proposes a prediction-based gradient compression method for distributed learning with event-triggered communication. Our goal is to reduce the amount of information transmitted from the distributed agents to the parameter server by exploiting temporal correlation in the local gradients. We use a linear predictor that \textit{combines past gradients to form a prediction of the current gradient}, with coefficients that are optimized by solving a least-square problem. In each iteration, every agent transmits the predictor coefficients to the server such that the predicted local gradient can be computed. The difference between the true local gradient and the predicted one, termed the \textit{prediction residual, is only transmitted when its norm is above some threshold.} When this additional communication step is omitted, the server uses the prediction as the estimated gradient. This proposed design shows notable performance gains compared to existing methods in the literature, achieving convergence with reduced communication costs.

cs.IT↗

arXiv2024

Context-Generative Default Policy for Bounded Rational Agent

Bounded rational agents often make decisions by evaluating a finite selection of choices, typically derived from a reference point termed the $`$default policy,' based on previous experience. However, the inherent rigidity of the static default policy presents significant challenges for agents when operating in unknown environment, that are not included in agent's prior knowledge. In this work, we introduce a context-generative default policy that leverages the region observed by the robot to predict unobserved part of the environment, thereby enabling the robot to adaptively adjust its default policy based on both the actual observed map and the $\textit{imagined}$ unobserved map. Furthermore, the adaptive nature of the bounded rationality framework enables the robot to manage unreliable or incorrect imaginations by selectively sampling a few trajectories in the vicinity of the default policy. Our approach utilizes a diffusion model for map prediction and a sampling-based planning with B-spline trajectory optimization to generate the default policy. Extensive evaluations reveal that the context-generative policy outperforms the baseline methods in identifying and avoiding unseen obstacles. Additionally, real-world experiments conducted with the Crazyflie drones demonstrate the adaptability of our proposed method, even when acting in environments outside the domain of the training distribution.

cs.RO↗

arXiv2024

BernGraph: Probabilistic Graph Neural Networks for EHR-based Medication Recommendations

The medical community believes binary medical event outcomes in EHR data contain sufficient information for making a sensible recommendation. However, there are two challenges to effectively utilizing such data: (1) modeling the relationship between massive 0,1 event outcomes is difficult, even with expert knowledge; (2) in practice, learning can be stalled by the binary values since the equally important 0 entries propagate no learning signals. Currently, there is a large gap between the assumed sufficient information and the reality that no promising results have been shown by utilizing solely the binary data: visiting or secondary information is often necessary to reach acceptable performance. In this paper, we attempt to build the first successful binary EHR data-oriented drug recommendation system by tackling the two difficulties, making sensible drug recommendations solely using the binary EHR medical records. To this end, we take a statistical perspective to view the EHR data as a sample from its cohorts and transform them into continuous Bernoulli probabilities. The transformed entries not only model a deterministic binary event with a distribution but also allow reflecting \emph{event-event} relationship by conditional probability. A graph neural network is learned on top of the transformation. It captures event-event correlations while emphasizing \emph{event-to-patient} features. Extensive results demonstrate that the proposed method achieves state-of-the-art performance on large-scale databases, outperforming baseline methods that use secondary information by a large margin. The source code is available at \url{https://github.com/chenzRG/BEHRMecom}

cs.AI↗

arXiv2024

Anatomical Consistency Distillation and Inconsistency Synthesis for Brain Tumor Segmentation with Missing Modalities

Multi-modal Magnetic Resonance Imaging (MRI) is imperative for accurate brain tumor segmentation, offering indispensable complementary information. Nonetheless, the absence of modalities poses significant challenges in achieving precise segmentation. Recognizing the shared anatomical structures between mono-modal and multi-modal representations, it is noteworthy that mono-modal images typically exhibit limited features in specific regions and tissues. In response to this, we present Anatomical Consistency Distillation and Inconsistency Synthesis (ACDIS), a novel framework designed to transfer anatomical structures from multi-modal to mono-modal representations and synthesize modality-specific features. ACDIS consists of two main components: Anatomical Consistency Distillation (ACD) and Modality Feature Synthesis Block (MFSB). ACD incorporates the Anatomical Feature Enhancement Block (AFEB), meticulously mining anatomical information. Simultaneously, Anatomical Consistency ConsTraints (ACCT) are employed to facilitate the consistent knowledge transfer, i.e., the richness of information and the similarity in anatomical structure, ensuring precise alignment of structural features across mono-modality and multi-modality. Complementarily, MFSB produces modality-specific features to rectify anatomical inconsistencies, thereby compensating for missing information in the segmented features. Through validation on the BraTS2018 and BraTS2020 datasets, ACDIS substantiates its efficacy in the segmentation of brain tumors with missing MRI modalities.

eess.IV↗

arXiv2024

Carbon Footprint Accounting Driven by Large Language Models and Retrieval-augmented Generation

Carbon footprint accounting is crucial for quantifying greenhouse gas emissions and achieving carbon neutrality.The dynamic nature of processes, accounting rules, carbon-related policies, and energy supply structures necessitates real-time updates of CFA. Traditional life cycle assessment methods rely heavily on human expertise, making near-real-time updates challenging. This paper introduces a novel approach integrating large language models (LLMs) with retrieval-augmented generation technology to enhance the real-time, professional, and economical aspects of carbon footprint information retrieval and analysis. By leveraging LLMs' logical and language understanding abilities and RAG's efficient retrieval capabilities, the proposed method LLMs-RAG-CFA can retrieve more relevant professional information to assist LLMs, enhancing the model's generative abilities. This method offers broad professional coverage, efficient real-time carbon footprint information acquisition and accounting, and cost-effective automation without frequent LLMs' parameter updates. Experimental results across five industries(primary aluminum, lithium battery, photovoltaic, new energy vehicles, and transformers)demonstrate that the LLMs-RAG-CFA method outperforms traditional methods and other LLMs, achieving higher information retrieval rates and significantly lower information deviations and carbon footprint accounting deviations. The economically viable design utilizes RAG technology to balance real-time updates with cost-effectiveness, providing an efficient, reliable, and cost-saving solution for real-time carbon emission management, thereby enhancing environmental sustainability practices.

cs.IR↗

arXiv2024

A Physics-Informed Indirect Method for Trajectory Optimization

This work presents a Physics-Informed Indirect Method (PIIM) that propagates the dynamics of both states and co-states backward in time for trajectory optimization problems. In the case of a Time-Optimal Soft Landing Problem (TOSLP), based on the initial co-state vector normalization technique, we show that the initial guess of the mass co-state and the numerical factor can be eliminated from the shooting procedure. As a result, the initial guess of the unknown co-states can be constrained to lie on a unit 3-D hypersphere. Then, using the PIIM allows one to exploit the physical significance of the optimal control law, which further narrows down the solution space to a unit 3-D octant sphere. Meanwhile, the analytical estimations of the fuel consumption and final time are provided. Additionally, a usually overlooked issue that results in an infeasible solution with a negative final time, is fixed by a simple remedy strategy. Consequently, the reduced solution space becomes sufficiently small to ensure fast, robust, and guaranteed convergence for the TOSLP. Then, we extend the PIIM to solve the Fuel-Optimal Soft Landing Problem (FOSLP) with a homotopy approach. The numerical simulations show that compared with the conventional indirect method with a success rate of 89.35%, it takes a shorter time for the proposed method to find the feasible solution to the FOSLP with a success rate of 100%.

math.OC↗

arXiv2024

Spatially covariant gravity with nonmetricity

Scalar fields play an important role in constructing modified gravity theories. In the case of a single scalar field with timelike gradient, the corresponding Lagrangian in the unitary gauge takes the form of spatially covariant gravity (SCG), which is proved useful in analyzing and extending the generally covariant theories. In this work, we apply the SCG method to the scalar-nonmetricity theory, of which the Lagrangian is built of the nonmetricity tensor and a scalar field. We derive the 3+1 decomposition of the geometric quantities and especially covariant derivatives of the scalar field up to the third order in the presence of a nonvanishing nonmetricity tensor. By fixing the unitary gauge, the resulting Lagrangian takes the form of a SCG with nonmetricity, in which all the ingredients are spatial tensors. We then exhaust the scalar monomials of SCG with nonmetricity up to $d=3$ with $d$ the total number of derivatives. Since the disformation tensor plays as an auxiliary variable, we take the Lagrangian with $d=2$ as an example to show that after solving the disformation tensor, we can get an effective SCG theory for the metric variables but with modified coefficients. Our results provides a novel approach to extending the scalar-nonmetricity theory.

gr-qc↗

arXiv2024

Visual-Geometry GP-based Navigable Space for Autonomous Navigation

Autonomous navigation in unknown environments is challenging and demands the consideration of both geometric and semantic information in order to parse the navigability of the environment. In this work, we propose a novel space modeling framework, Visual-Geometry Sparse Gaussian Process (VG-SGP), that simultaneously considers semantics and geometry of the scene. Our proposed approach can overcome the limitation of visual planners that fail to recognize geometry associated with the semantic and the geometric planners that completely overlook the semantic information which is very critical in real-world navigation. The proposed method leverages dual Sparse Gaussian Processes in an integrated manner; the first is trained to forecast geometrically navigable spaces while the second predicts the semantically navigable areas. This integrated model is able to pinpoint the overlapping (geometric and semantic) navigable space. The simulation and real-world experiments demonstrate that the ability of the proposed VG-SGP model, coupled with our innovative navigation strategy, outperforms models solely reliant on visual or geometric navigation algorithms, highlighting a superior adaptive behavior.

cs.RO↗

arXiv2024

Fredformer: Frequency Debiased Transformer for Time Series Forecasting

The Transformer model has shown leading performance in time series forecasting. Nevertheless, in some complex scenarios, it tends to learn low-frequency features in the data and overlook high-frequency features, showing a frequency bias. This bias prevents the model from accurately capturing important high-frequency data features. In this paper, we undertook empirical analyses to understand this bias and discovered that frequency bias results from the model disproportionately focusing on frequency features with higher energy. Based on our analysis, we formulate this bias and propose Fredformer, a Transformer-based framework designed to mitigate frequency bias by learning features equally across different frequency bands. This approach prevents the model from overlooking lower amplitude features important for accurate forecasting. Extensive experiments show the effectiveness of our proposed approach, which can outperform other baselines in different real-world time-series datasets. Furthermore, we introduce a lightweight variant of the Fredformer with an attention matrix approximation, which achieves comparable performance but with much fewer parameters and lower computation costs. The code is available at: https://github.com/chenzRG/Fredformer

cs.LG↗

arXiv2024

Self-Supervised Position Debiasing for Large Language Models

Fine-tuning has been demonstrated to be an effective method to improve the domain performance of large language models (LLMs). However, LLMs might fit the dataset bias and shortcuts for prediction, leading to poor generation performance. Previous works have proven that LLMs are prone to exhibit position bias, i.e., leveraging information positioned at the beginning or end, or specific positional cues within the input. Existing debiasing methods for LLMs require external bias knowledge or annotated non-biased samples, which is lacking for position debiasing and impractical in reality. In this work, we propose a self-supervised position debiasing (SOD) framework to mitigate position bias for LLMs. SOD leverages unsupervised responses from pre-trained LLMs for debiasing without relying on any external knowledge. To improve the quality of unsupervised responses, we propose an objective alignment (OAM) module to prune these responses. Experiments on eight datasets and five tasks show that SOD consistently outperforms existing methods in mitigating three types of position biases. Besides, SOD achieves this by sacrificing only a small performance on biased samples, which is general and effective. To facilitate the reproducibility of the results, we share the code of all methods and datasets on https://github.com/LZKSKY/SOD.

cs.CL↗

arXiv2024

Nonlinear Optimal Guidance with Constraints on Impact Time and Impact Angle

This paper aims to address the nonlinear optimal guidance problem with impact-time and impact-angle constraints, which is fundamentally important for multiple pursuers to collaboratively achieve a target. Addressing such a guidance problem is equivalent to solving a nonlinear minimum-effort control problem in real time. To this end, the Pontryagain's maximum principle is employed to convert extremal trajectories as the solutions of a parameterized differential system. The geometric property for the solution of the parameterized system is analyzed, leading to an additional optimality condition. By incorporating this optimality condition and the usual disconjugacy condition into the parameterized system, the dataset for optimal trajectories can be generated by propagating the parameterized system without using any optimization methods. In addition, a scaling invariance property is found for the solutions of the parameterized system. As a consequence of this scaling invariance property, a simple feedforward neural network trained by the solution of the parameterized system, selected at any fixed time, can be used to generate the nonlinear optimal guidance within milliseconds. Finally, numerical examples are presented, showing that the nonlinear optimal guidance command generated by the trained network can not only ensure the expected impact angle and impact time are precisely met but also requires less control effort compared with existing guidance methods.

math.OC↗

arXiv2024

DmADs-Net: Dense multiscale attention and depth-supervised network for medical image segmentation

Deep learning has made important contributions to the development of medical image segmentation. Convolutional neural networks, as a crucial branch, have attracted strong attention from researchers. Through the tireless efforts of numerous researchers, convolutional neural networks have yielded numerous outstanding algorithms for processing medical images. The ideas and architectures of these algorithms have also provided important inspiration for the development of later technologies.Through extensive experimentation, we have found that currently mainstream deep learning algorithms are not always able to achieve ideal results when processing complex datasets and different types of datasets. These networks still have room for improvement in lesion localization and feature extraction. Therefore, we have created the Dense Multiscale Attention and Depth-Supervised Network (DmADs-Net).We use ResNet for feature extraction at different depths and create a Multi-scale Convolutional Feature Attention Block to improve the network's attention to weak feature information. The Local Feature Attention Block is created to enable enhanced local feature attention for high-level semantic information. In addition, in the feature fusion phase, a Feature Refinement and Fusion Block is created to enhance the fusion of different semantic information.We validated the performance of the network using five datasets of varying sizes and types. Results from comparative experiments show that DmADs-Net outperformed mainstream networks. Ablation experiments further demonstrated the effectiveness of the created modules and the rationality of the network architecture.

eess.IV↗

arXiv2024

NARUTO: Neural Active Reconstruction from Uncertain Target Observations

We present NARUTO, a neural active reconstruction system that combines a hybrid neural representation with uncertainty learning, enabling high-fidelity surface reconstruction. Our approach leverages a multi-resolution hash-grid as the mapping backbone, chosen for its exceptional convergence speed and capacity to capture high-frequency local features.The centerpiece of our work is the incorporation of an uncertainty learning module that dynamically quantifies reconstruction uncertainty while actively reconstructing the environment. By harnessing learned uncertainty, we propose a novel uncertainty aggregation strategy for goal searching and efficient path planning. Our system autonomously explores by targeting uncertain observations and reconstructs environments with remarkable completeness and fidelity. We also demonstrate the utility of this uncertainty-aware approach by enhancing SOTA neural SLAM systems through an active ray sampling strategy. Extensive evaluations of NARUTO in various environments, using an indoor scene simulator, confirm its superior performance and state-of-the-art status in active reconstruction, as evidenced by its impressive results on benchmark datasets like Replica and MP3D.

cs.CV↗

arXiv2024

Fuel-optimal powered descent guidance for lunar pinpoint landing using neural networks

This paper presents a Neural Networks (NNs) based approach for designing the Fuel-Optimal Powered Descent Guidance (FOPDG) for lunar pinpoint landing. According to Pontryagin's Minimum Principle, the optimality conditions are first derived. To generate the dataset of optimal trajectories for training NNs, we formulate a parameterized system, which allows for generating each optimal trajectory by a simple propagation without using any optimization method. Then, a dataset containing the optimal state and optimal thrust vector pairs can be readily collected. Since it is challenging for NNs to approximate bang-bang (or discontinuous) type of optimal thrust magnitude, we introduce a regularisation function to the switching function so that the regularized switching function approximated by a simple NN can be used to represent the optimal thrust magnitude. Meanwhile, another two well-trained NNs are used to predict the thrust steering angle and time of flight given a flight state. Finally, numerical simulations show that the proposed method is capable of generating the FOPDG that steers the lunar lander to the desired landing site with acceptable landing errors.

math.OC↗

arXiv2024

Fuel-Optimal Trajectory Planning for Lunar Vertical Landing

In this paper, we consider a trajectory planning problem arising from a lunar vertical landing with minimum fuel consumption. The vertical landing requirement is written as a final steering angle constraint, and a nonnegative regularization term is proposed to modify the cost functional. In this way, the final steering angle constraint will be inherently satisfied according to Pontryagin's Minimum Principle. As a result, the modified optimal steering angle has to be determined by solving a transcendental equation. To this end, a transforming procedure is employed, which allows for finding the desired optimal steering angle by a simple bisection method. Consequently, the modified optimal control problem can be solved by the indirect shooting method. Finally, some numerical examples are presented to demonstrate and verify the developments of the paper.

math.OC↗

arXiv2024

Position-Aware Parameter Efficient Fine-Tuning Approach for Reducing Positional Bias in LLMs

Recent advances in large language models (LLMs) have enhanced their ability to process long input contexts. This development is particularly crucial for tasks that involve retrieving knowledge from an external datastore, which can result in long inputs. However, recent studies show a positional bias in LLMs, demonstrating varying performance depending on the location of useful information within the input sequence. In this study, we conduct extensive experiments to investigate the root causes of positional bias. Our findings indicate that the primary contributor to LLM positional bias stems from the inherent positional preferences of different models. We demonstrate that merely employing prompt-based solutions is inadequate for overcoming the positional preferences. To address this positional bias issue of a pre-trained LLM, we developed a Position-Aware Parameter Efficient Fine-Tuning (PAPEFT) approach which is composed of a data augmentation technique and a parameter efficient adapter, enhancing a uniform attention distribution across the input context. Our experiments demonstrate that the proposed approach effectively reduces positional bias, improving LLMs' effectiveness in handling long context sequences for various tasks that require externally retrieved knowledge.

cs.CL↗

arXiv2024

Point-DETR3D: Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection

Training high-accuracy 3D detectors necessitates massive labeled 3D annotations with 7 degree-of-freedom, which is laborious and time-consuming. Therefore, the form of point annotations is proposed to offer significant prospects for practical applications in 3D detection, which is not only more accessible and less expensive but also provides strong spatial information for object localization. In this paper, we empirically discover that it is non-trivial to merely adapt Point-DETR to its 3D form, encountering two main bottlenecks: 1) it fails to encode strong 3D prior into the model, and 2) it generates low-quality pseudo labels in distant regions due to the extreme sparsity of LiDAR points. To overcome these challenges, we introduce Point-DETR3D, a teacher-student framework for weakly semi-supervised 3D detection, designed to fully capitalize on point-wise supervision within a constrained instance-wise annotation budget.Different from Point-DETR which encodes 3D positional information solely through a point encoder, we propose an explicit positional query initialization strategy to enhance the positional prior. Considering the low quality of pseudo labels at distant regions produced by the teacher model, we enhance the detector's perception by incorporating dense imagery data through a novel Cross-Modal Deformable RoI Fusion (D-RoI).Moreover, an innovative point-guided self-supervised learning technique is proposed to allow for fully exploiting point priors, even in student models.Extensive experiments on representative nuScenes dataset demonstrate our Point-DETR3D obtains significant improvements compared to previous works. Notably, with only 5% of labeled data, Point-DETR3D achieves over 90% performance of its fully supervised counterpart.

cs.CV↗