Vision-Language Transformers (VLTs) have achieved remarkable success, but their computational costs pose a challenge due to the large number of input tokens and extensive model parameters. Existing VLT compression methods primarily rely on single-modality-based token pruning or coarse-grained weight pruning techniques. However, these methods face significant obstacles, such as ignoring the critical alignment of different modalities and lacking the flexibility to dynamically compress each layer for token pruning, exhibiting inevitable performance degradation due to coarse-grained weight pruning, and struggling with the simultaneous compression of both input tokens and model parameters. To address those limitations, we propose MADTP++, a novel approach that integrates custom-made token and weight pruning processes into a unified framework, achieving superior compression in both parameter counts and computational costs.

Multi-Modal Large Language Models (MLLMs) stand out in various tasks but still struggle with hallucinations. While recent training-free mitigation methods mostly introduce additional inference overhead via retrospection strategy and contrastive decoding, we propose attention reallocation (AttnReal) to mitigate hallucinations with nearly zero extra cost. Our approach is motivated by the key observations that, MLLM's unreasonable attention distribution causes features to be dominated by historical output tokens, which further contributes to hallucinated responses because of the distribution gap between different token types. Based on the observations, AttnReal recycles excessive attention from output tokens and reallocates it to visual tokens, which reduces MLLM's reliance on language priors and ensures the decoding process depends more on the visual inputs. More interestingly, we find that, by controlling the intensity of AttnReal, we can achieve a wide-range trade-off between the response faithfulness and overall performance. Comprehensive results from different benchmarks validate the effectiveness of AttnReal across six open-source MLLMs and three decoding strategies.

Taylor-Series-Expansion (TSE) is a mathematics theorem. It proves that the expansion of the first few finite Taylor Series is a good approximation of a nonlinear function in most cases. Inspired by the TSE theorem, a brand-new TSE-based vision transformer is designed. TSE-based vision transformer uses the shared first-order TSE transformer block’s weight (in analogy with the Taylor-Series first-order term), its finite multiple multiplications (in analogy with the Taylor-Series expanded high-order terms), and the corresponding learnable TSE coefficients to approximate the naive vision transformer. In this manner, the TSE-based vision model reduces the memory burden but keeps a similar accuracy as the naive counterpart. Derived from adding the Taylor skip mechanism in training, the TSE-based vision transformer has good dynamic expansion capability. Experiment results show TSE-based models can boost actual deployment latency by 1.30-1.36× on A100 GPU and 1.34-1.45× on AGX Orin with negligible accuracy degradation on ImageNet classification, COCO detection, and ADE20K segmentation benchmarking tasks. Moreover, TSE-based optimization is orthogonal to model compression. Combining with the state-of-the-art vision transformer compression method, it can boost actual deployment performance by 1.70-1.87× and 3.29-3.61× of latency and throughput on A100 GPU, and 1.67-1.74× and 2.76-2.94× improvement of latency and throughput on AGX Orin.

Residual networks have shown great success and become indispensable in recent deep neural network models. In this work, we aim to re-investigate the training process of residual networks from a novel social psychology perspective of loafing, and further propose a new training scheme as well as three improved strategies for boosting residual networks beyond their performance limits. Previous research has suggested that residual networks can be considered as ensembles of shallow networks, which implies that the final performance of a residual network is influenced by a group of subnetworks. We identify a previously overlooked problem that is analogous to social loafing, where subnetworks within a residual network are prone to exert less effort when working as part of a group compared to working alone. We define this problem as \textit{network loafing}. Similar to the decreased individual productivity and overall performance as demonstrated in society, network loafing inevitably causes sub-par performance. Inspired by solutions from social psychology, we first propose a novel training scheme called stimulative training, which randomly samples a residual subnetwork and calculates the KL divergence loss between the sampled subnetwork and the given residual network for extra supervision. In order to unleash the potential of stimulative training, we further propose three simple-yet-effective strategies, including a novel KL- loss that only aligns the network logits direction, random smaller inputs for subnetworks, and inter-stage sampling rules. Comprehensive experiments and analysis verify the effectiveness of stimulative training as well as its three improved strategies.

Class-incremental 3D object detection demands a 3D detector to locate and recognize novel categories in a stream fashion while preserving its base detection ability. However, existing methods require delicate 3D annotations for learning novel categories, resulting in significant labeling costs. To this end, we explore a label-efficient approach called Weakly Incremental 3D Detection (WI3D), which teaches a 3D detector to learn incrementally with off-the-shelf vision foundation models. We propose a novel dual-teaching framework incorporating both intra-modal and inter-modal knowledge from pseudo labels and feature space. Specifically, our framework features a class-agnostic pseudo-label refinement module, designed for the generation of high-quality 3D pseudo labels. This module is built on a lightweight transformer that models the spatial relationships between pseudo labels and their interactions with rich contextual information in point clouds. Additionally, we introduce a cross-modal knowledge transfer module to enhance the representation learning of novel classes, along with a reweighting knowledge distillation strategy that dynamically assesses and distills knowledge from previously learned categories. Extensive experiments show that our approach can efficiently learn novel concepts while preserving knowledge of base classes in WI3D scenarios, and surpass baseline approaches on both SUN-RGBD and ScanNet.

3D dense captioning requires a model to translate its understanding of an input 3D scene into several captions associated with different object regions. Existing methods adopt a sophisticated "detect-then-describe" pipeline, which builds explicit relation modules upon a 3D detector with numerous hand-crafted components. While these methods have achieved initial success, the cascade pipeline tends to accumulate errors because of duplicated and inaccurate box estimations and messy 3D scenes. In this paper, we first propose Vote2Cap-DETR, a simple-yet-effective transformer framework that decouples the decoding process of caption generation and object localization through parallel decoding. Moreover, we argue that object localization and description generation require different levels of scene understanding, which could be challenging for a shared set of queries to capture. To this end, we propose an advanced version, Vote2Cap-DETR++, which decouples the queries into localization and caption queries to capture task-specific features. Additionally, we introduce the iterative spatial refinement strategy to vote queries for faster convergence and better localization performance. We also insert additional spatial information to the caption head for more accurate descriptions. Without bells and whistles, extensive experiments on two commonly used datasets, ScanRefer and Nr3D, demonstrate Vote2Cap-DETR and Vote2Cap-DETR++ surpass conventional "detect-then-describe" methods by a large margin.

Global channel pruning (GCP) aims to remove a subset of channels (filters) across different layers from a deep model without hurting the performance. Previous works focus on either single task model pruning or simply adapting it to multitask scenario, and still face the following problems when handling multitask pruning: 1) Due to the task mismatch, a well-pruned backbone for classification task focuses on preserving filters that can extract category-sensitive information, causing filters that may be useful for other tasks to be pruned during the backbone pruning stage; 2) For multitask predictions, different filters within or between layers are more closely related and interacted than that for single task prediction, making multitask pruning more difficult. Therefore, aiming at multitask model compression, we propose a Performance-Aware Global Channel Pruning (PAGCP) framework. We first theoretically present the objective for achieving superior GCP, by considering the joint saliency of filters from intra- and inter-layers. Then a sequentially greedy pruning strategy is proposed to optimize the objective, where a performance-aware oracle criterion is developed to evaluate sensitivity of filters to each task and preserve the globally most task-related filters. Experiments on several multitask datasets show that the proposed PAGCP can reduce the FLOPs and parameters by over 60% with minor performance drop, and achieves 1.2x∼3.3x acceleration on both cloud and mobile platforms.

Irreducible Cartesian tensors (ICTs) play a crucial role in the design of equivariant graph neural networks, as well as in theoretical chemistry and chemical physics. Meanwhile, the design space of available linear operations on tensors that preserve symmetry presents a significant challenge. The ICT decomposition and a basis of this equivariant space are difficult to obtain for high-rank tensors. After decades of research, Bonvicini (2024) has recently achieved an explicit ICT decomposition for n = 5 with factorial time/space complexity. In this work we, for the first time, obtain decomposition matrices for ICTs up to rank n = 9 with reduced and affordable complexity, by constructing what we call path matrices. The path matrices are obtained via performing chain-like contractions with Clebsch-Gordan matrices following the parentage scheme. We prove and leverage that the concatenation of path matrices is an orthonormal change-of-basis matrix between the Cartesian tensor product space and the spherical direct sum spaces. Furthermore, we identify a complete orthogonal basis for the equivariant space, rather than a spanning set (Pearce-Crump, 2023b), through this path matrices technique. Our method avoids the RREF algorithm and maintains a fully analytical derivation of each ICT decomposition matrix, thereby significantly improving the algorithm's speed to obtain arbitrary rank orthogonal ICT decomposition matrices and orthogonal equivariant bases. We further extend our result to the arbitrary tensor product and direct sum spaces, enabling free design between different spaces while keeping symmetry.

In deep learning, different kinds of deep networks typically need different optimizers, which have to be chosen after multiple trials, making the training process inefficient. To relieve this issue and consistently improve the model training speed across deep networks, we propose the ADAptive Nesterov momentum algorithm, Adan for short. Adan first reformulates the vanilla Nesterov acceleration to develop a new Nesterov momentum estimation (NME) method, which avoids the extra overhead of computing gradient at the extrapolation point. Then Adan adopts NME to estimate the gradient’s first- and second-order moments in adaptive gradient algorithms for convergence acceleration. Besides, we prove that Adan finds an \epsilon-approximate first-order stationary point within O(\epsilon^{-3.5}) stochastic gradient complexity on the non-convex stochastic problems (e.g., deep learning problems), matching the best-known lower bound. Extensive experimental results show that Adan consistently surpasses the corresponding SoTA optimizers on vision, language, and RL tasks and sets new SoTAs for many popular networks and frameworks, e.g., ResNet, ConvNext, ViT, Swin, MAE, DETR, GPT-2, Transformer-XL, and BERT. More surprisingly, Adan can use half of the training cost (epochs) of SoTA optimizers to achieve higher or comparable performance on ViT, GPT-2, MAE, etc, and also shows great tolerance to a large range of minibatch size, e.g., from 1 k to 32 k.

Universal approximation capability, also referred to as universality, is an important property of deep neural networks,endowing them with the potency to accurately represent the under lying target function in learning tasks. In practice, the architecture of deep neural networks largely influences the performance of the models. However, most existing methodologies for designing neural architectures, such as the heuristic manual design or neural architecture search, ignore the universal approximation property,thus losing a potential safeguard about the performance. In this paper, we propose a unified framework to design the architectures of deep neural networks with a universality guarantee based on first-order optimization algorithms, where the forward pass is interpreted as the updates of an optimization algorithm. The (explicit or implicit) network is designed by replacing each gradient term in the algorithm with a learnable module similar to a two-layer network or its derivatives. Specifically, we explore the realm of width-bounded neural networks, a common practical scenario, showcasing their universality. Moreover, adding operations of normalization, downsampling, and upsampling does not hurt the universality. To the best of our knowledge, this is the first work that width-bounded networks with universal approximation guarantee can be designed in a principled way. Our framework can inspire a variety of neural architectures including some renowned structures such as ResNet and DenseNet, as well as novel innovations. The experimental results on image classification problems demonstrate that the newly inspired networks are competitive and surpass the baselines of ResNet, DenseNet, as well as the advanced ConvNeXt and ViT, testifying to the effectiveness of our framework.

AdamW modifies Adam by adding a decoupled weight decay to decay network weights per training iteration. For adaptive algorithms, this decoupled weight decay does not affect specific optimization steps, and differs from the widely used L2-regularizer which changes optimization steps via changing the first- and second-order gradient moments. Despite its great practical success, for AdamW, its convergence behavior and generalization improvement over Adam and L2-regularized Adam (L2-Adam) remain absent yet. To solve this issue, we prove the convergence of AdamW and justify its generalization advantages over Adam and L2-Adam. Specifically, AdamW provably converges but minimizes a dynamically regularized loss that combines vanilla loss and a dynamical regularization induced by decoupled weight decay, thus yielding different behaviors with Adam and L2-Adam. Moreover, on both general nonconvex problems and PL-conditioned problems, we establish stochastic gradient complexity of AdamW to find a stationary point. Such complexity is also applicable to Adam and L2-Adam, and improves their previously known complexity, especially for over-parameterized networks. Besides, we prove that AdamW enjoys smaller generalization errors than Adam and L2-Adam from the Bayesian posterior aspect. This result, for the first time, explicitly reveals the benefits of decoupled weight decay in AdamW. Experimental results validate our theory.

To reveal the mystery behind deep neural networks(DNNs), optimization may offer a good perspective. There are already some clues showing the strong connection between DNNs and optimization problems, e.g., under a mild condition, DNN's activation function is indeed a proximal operator. In this paper, we are committed to providing a unified optimization induced interpretability for a special class of networks--equilibrium models, i.e., neural networks defined by fixed point equations, which have become increasingly attractive recently. To this end, we first decompose DNNs into a new class of unit layer that is the proximal operator of an implicit convex function while keeping its output unchanged. Then, the equilibrium model of the unit layer can be derived, we name it Optimization Induced Equilibrium Networks (OptEq). The equilibrium point of OptEq can be theoretically connected to the solution of a convex optimization problem with explicit objectives. Based on this, we can flexibly introduce prior properties to the equilibrium points: 1)modifying the underlying convex problems explicitly so as to change the architectures of OptEq; and 2) merging the information into the fixed point iteration, which guarantees to choose the desired equilibrium point when the fixed point set is non-singleton. We show that OptEq outperforms previous implicit models even with fewer parameters.

Finite time-vertex graph signals (FTVGS) provide an efficient representation for capturing spatio-temporal correlations across multiple data sources on irregular structures. Although sampling and reconstruction of FTVGS with known spectral support have been extensively studied, the case of unknown spectral support requires further investigation. Existing random sampling methods may extract samples from any vertex at any time, but such strategies are not friendly in practice, where sampling is typically limited to a subset of vertices and moments. To address this requirement, we propose a subset random sampling scheme for FTVGS. Specifically, we first randomly select a subset of rows and columns to form a submatrix, followed by random sampling within that submatrix. In theory, we provide sufficient conditions for reconstructing the original FTVGS with high probability. Additionally, we introduce a reconstruction framework incorporating low-rank, sparsity, and smoothness priors (LSSP), and verify the feasibility of the reconstruction and the effectiveness of the framework through experiments.

Time-vertex graph signal (TVGS) models describe time-varying data with irregular structures. The bandlimitedness in the joint time-vertex Fourier spectral domain reflects smoothness in both temporal and graph topology. In this paper, we study the critical sampling of three types of TVGS including continuous-time signals, infinite-length sequences, and finite-length sequences in the time domain for each vertex on the graph. For a jointly bandlimited TVGS, we prove a lower bound on sampling density or sampling ratio, which depends on the measure of the spectral support in the joint time-vertex Fourier spectral domain. We also provide a lower bound on the sampling density or sampling ratio of each vertex on sampling sets for perfect recovery. To demonstrate that critical sampling is achievable, we propose the sampling and reconstruction procedures for the different types of TVGS. Finally, we show how the proposed sampling schemes can be applied to numerical as well as real datasets.

The theory of sampling and recovery of bandlimited graph signals has been extensively studied. However, in many cases, the observation of a signal is quite coarse. For example, users only provide simple comments such as “like” or “dislike” for a product on an e-commerce platform. This is a particular scenario where only the sign information of a graph signal can be measured. In this paper, we are interested in how to sample based on sign information in an online manner, by which the direction of the original graph signal can be estimated. The online signed sampling problem of a graph signal can be formulated as a Markov decision process in a finite horizon. Unfortunately, it is intractable for large size graphs. We propose a low-complexity greedy signed sampling algorithm (GSS) as well as a stopping criterion. Meanwhile, we prove that the objective function is adaptive monotonic and adaptive submodular, so that the performance is close enough to the global optimum with a lower bound. Finally, we demonstrate the effectiveness of the GSS algorithm by both synthesis and realworld data.

In this paper, we consider the problem of estimating the angular parameters, i.e., the nominal angle-of-arrivals (AoAs) and angular spreads, of incoherently distributed sources using the phased-array equipped with a single RF chain. We first derive the approximate Fourier series of the received power. The coefficients can be expressed in closed form with the angular parameters. In the case of single source, this finding directly suggests the design of the low-complexity algorithm that performs spatial sampling and discrete Fourier transform to estimate the Fourier series coefficients, from which the nominal AoA and angular spread can be obtained successively. In the case of multiple sources, we focus on one source at one time, and the multiples sources are handled one by one. Based on the Fourier series expression, the power fitting approach is proposed to build the nonlinear least-squares problem. Then, the semi-exhaustive search algorithm is developed to find the solution, which gives the angular parameters of the target source. Additionally, the approximate Cramer-Rao bound is derived as benchmark. The numerical results demonstrate that in certain cases, the proposed methods can even outperform the existing method that uses fully-digital array.

Interactive behavior-aware planning benefits from the hierarchical learning process when adapting to dense traffic. However, the difficulty in the Intelligent Transportation System (ITS) is that the autonomous vehicle fails to execute real-time response due to hardly perceiving dynamic objects beyond the visual range. This problem can be tackled by vehicle-roadcloud cooperation that synchronously collects global perception information and makes strategic policy for deployment. Here we propose a hierarchical edge-decision framework, which addresses real-time motion skill that distills from analogical reasoning of spatial-temporal events. The first step is establishing the goal-conditioned motion library from the centralized edge-cloud perception, to compose the belief-based best response with collision avoidance. In addition, a novel perspective of latent space is presented to promote motion rehearsal in the cloud, which could generate prior credible trajectories based on the policy distillation procedure of extracting informative action from thoroughly exploring changing events. Moreover, the twostage hierarchy decision is developed to boost the efficiency of advanced policy modification, through evaluating the hierarchical judgment matrices considering conditional criteria, thereby constituting an optimum auto-driving motion with vehicle-roadcloud collaborative system. Extensive validation on challenging autonomous driving scenarios outperforms, demonstrating that our edge-decision method significantly promotes adaption to the complex time-varying environment in ITS system in a smooth and sustainable manner.

Federated learning has provided a new approach to coordinating a group of clients to train a machine learning model collaboratively, which can be easily embedded into Unmanned Aerial Vehicle (UAV) swarms. Compared with the terrestrial wireless networks, the UAV swarm faces more precarious communication conditions, rendering synchronous aggregation no longer tenable. Additionally, the data collected from UAVs tend to be heterogeneous due to different deployment regions or requirements. To overcome these restrictions, this article has proposed a novel two-stage Asynchronous Federated Learning scheme for the UAV swarm. Initially, the convergence property of both convex and non-convex models trained by the proposed scheme is analyzed. In the pre-training stage, we modeled the learning process as a cooperative game with demonstrated monotonicity and submodularity. Furthermore, the Shapley Value is imported to quantify data values of UAVs, and the upper bound of its estimation error rate is derived. In the training stage, a new concept named Network Age of Updates (AoU) is proposed to address the fairness issue, quantifying the model’s generalization capability with data value consideration, and a sequential UAV selection scheduling is performed through the AoU minimization by Whittle Index method. Finally, the system performance is validated through both theoretical analysis and simulations.

Nowadays, the Unmanned Aerial Vehicle (UAV) has emerged as a powerful platform for diversified application development. In traffic areas specifically for the much-anticipated Intelligent Transportation System (ITS), UAVs can coordinate with the ground internet of things infrastructure, such as the Road Side Unit (RSU), to perform 3D data collection and processing, improving traffic safety and transportation efficiency. In ITS, decision-making is crucial and relies heavily on proper and timely processing and transmission of the massive amounts of data generated by ubiquitous sensors, especially in emergency disposal scenarios. In this paper, the latency incurred from RSU-aided data offloading and UAV-aided data relay plays a decisive role, characterized by both the average and risk performance metric, Conditional Value at Risk (CVaR), for quantifying the risk incurred from the latency violation over a certain threshold in probability. Specifically, the peak Age of Information (AoI) with its distribution is adopted to underlying the CVaR analysis in relay latency. Besides, the matching algorithm to coordinate the mutual transmission between UAVs and RSUs is proposed, achieving low computation complexity and minimum risk for both sides. The system performance is validated through analysis, simulation, and field experiments.

We consider a Cognitive Radio based Internet of Things (CR-IoT) system consisting of multiple Primary IoT Devices (PDs) and Secondary IoT Devices (SDs). We focus on the impact of the IoT devices’ heterogeneous traffic pattern on the energy efficiency and on the age of information (AoI) performance, and aim to maximize the sum of energy efficiency of the SDs under their individual constraints of the average age of information (AoI), via optimally deciding the transmit power and channel allocation. We first derive the closed-form expressions of the energy efficiency and the average AoI, and subsequently explore their convexity and monotonicity to the transmit power. Following these haracterizations, an optimal transmit power optimization algorithm (TPOA) is proposed for the SDs, and the channel allocation problem is reformulated as a matching game based on the utilities calculated from TPOA. To improve the system performance, we introduce the virtual price charged in each channel. Taking advantage of the correlation between two adjacent matchings, we propose a semi-distributed, low-complexity adaptive matching algorithm to construct an -stable matching, and the effectiveness of the proposed matching algorithm is verified through Monte Carlo simulations

We consider a wireless network with a source periodically generating time-sensitive information and transmitting it to a destination via one of N non-stationary orthogonal wireless channels. The goal of the scheduling policy is to keep the information at the destination fresh, which is captured by the Age of Information (AoI) metric. While obtaining an analytical and accurate AoI performance characterization in non-stationary wireless channels is usually intractable, we thereby resort to multi-armed bandits (MAB) to solve this problem, where the non-stationary channels and AoI as taken as arms and rewards, respectively. We consider three special non-stationary channels in which the lower bound on the AoI regret achievable by any policy is derived, respectively. In addition, the upper bound of Exp3.S, Active Arm Elimination (AAE) and Cumulative Sum Upper Confidence Bound (CUSUM-UCB) policy for the corresponding three settings are presented. Furthermore, the variants of AAE and CUSUM-UCB are proposed and verified more effectively than their original policies via simulations.

In this letter, we investigate a reconfigurable intelligent surface (RIS) assisted downlink multi-user multipleinput single-output (MISO) system. We derive the approximations of average rate of user equipments (UEs) under proper assumptions, and reveal the relations between average rates and system parameters. Furthermore, a RIS configuration algorithm is proposed to improve the average sum rate with low complexity by exploiting the approximation of average rate and statistical channel state information (CSI). The rationality of the assumptions, the tightness of the approximations and the effectiveness of the proposed algorithm are validated by numerical results.

In this paper, we consider a cooperative device-todevice (D2D) communication system, where the D2D transmitters (DTs) act as relays to assist the densified cellular network users (CUs) for transmission quality of service (QoS) improvement. The proposed system achieves a win-win situation, i.e. improving the spectrum efficiency of the CUs that cannot meet their rate requirement while providing spectrum access for D2D pairs. Unlike previous works, to reduce the overhead, we design a novel two-timescale resource allocation scheme, in which the pairing between CUs and D2D pairs is decided at a long timescale and transmission time for CU and D2D pair is determined at a short timescale. Specifically, to characterize the long-term payoff of each potential CU-D2D pair, we investigate the optimal cooperation policy to decide the transmission time based on the instantaneous channel state information (CSI). We prove that the optimal policy is a threshold policy which can be achieved via binary search. Since CUs and D2D pairs are self-interested, they are paired only when they agree to cooperate mutually. Therefore, to study the cooperation behaviors of CUs and D2D pairs, we formulate the pairing problem as a matching game, based on the long-term payoff achieved by the optimal cooperation policy of each possible pairing. Furthermore, unlike most previous matching models in D2D networks, we allow transfer between CUs and D2D pairs to improve the performance. To solve the pairing problem, a distributed algorithm is proposed, which converges to an -stable matching. We show that there is a trade-off between the optimality and the computational complexity of the algorithm. We also analyze the algorithm in terms of the robustness to the unilateral deviation of D2D pairs. Finally, the simulation results verify the efficiency of the proposed matching algorithm.

While Large Language Models (LLMs) demonstrate strong performance across domains, their long-context capabilities are limited by transient neural activa tions causing information decay and unstructured feed-forward network (FFN) weights leading to semantic fragmentation. Inspired by the brain’s working mem ory and cortical modularity, we propose PaceLLM, featuring two innovations: (1) a Persistent Activity (PA) Mechanism that mimics prefrontal cortex (PFC) neurons’ persistent firing by introducing an activation-level memory bank to dynamically retrieve, reuse, and update critical FFN states, addressing contextual decay; and (2) Cortical Expert (CE) Clustering that emulates task-adaptive neural specialization to reorganize FFN weights into semantic modules, establishing cross-token depen dencies and mitigating fragmentation. Extensive evaluations show that PaceLLM achieves 6% improvement on LongBench’s Multi-document QA and 12.5–17.5% performance gains on ∞-Bench tasks, while extending measurable context length to 200K tokens in Needle-In-A-Haystack (NIAH) tests. This work pioneers brain inspired LLM optimization and is complementary to other works. Besides, it can be generalized to any model and enhance their long-context performance and interpretability without structural overhauls.

Multimodal Large Language Models (MLLMs) have shownremarkable capabilities in video content understand ing but still struggle with fine-grained motion comprehen sion. To comprehensively assess the motion understand ing ability of existing MLLMs, we introduce FAVOR-Bench, comprising 1,776 videos with structured manual annota tions of various motions. Our benchmark includes both close-ended and open-ended tasks. For close-ended evalu ation, we carefully design 8,184 multiple-choice question answer pairs spanning six distinct sub-tasks. For open ended evaluation, we develop both a novel cost-efficient LLM-free and a GPT-assisted caption assessment method, where the former can enhance benchmarking interpretabil ity and reproducibility. Comprehensive experiments with 21 state-of-the-art MLLMs reveal significant limitations in their ability to comprehend and describe detailed temporal dynamics in video motions. To alleviate this limitation, we further build FAVOR-Train, a dataset consisting of 17,152 videos with fine-grained motion annotations. The results of finetuning Qwen2.5-VL on FAVOR-Train yield consis tent improvements on motion-related tasks of TVBench, Mo tionBench and our FAVOR-Bench. Comprehensive assess ment results demonstrate that the proposed FAVOR-Bench and FAVOR-Train provide valuable tools to the community for developing more powerful video understanding models.

The scientific research paradigm is undergoing a profound transformation owing to the development of Artificial Intelligence (AI). Recent works demonstrate that various AI-assisted research methods can largely improve research efficiency by improving data analysis, accelerating computation, and fostering novel idea generation. To further move towards the ultimate goal (i.e., automatic scientific research), in this paper, we introduce Dolphin, a closed-loop LLM-driven framework to enhance the automation level of scientific research. Dolphin first generates novel ideas based on feedback from previous experiments and relevant papers ranked by the topic and task attributes. Then, the generated ideas can be implemented using a code template refined and debugged with the designed exception-traceback-guided local code structure. Finally, Dolphin automatically analyzes the results of each idea and feeds the results back to the next round of idea generation. Experiments are conducted on the benchmark datasets of different topics and a subset of MLE-bench. Results show that Dolphin can continuously improve the performance of the input topic in a loop. We highlight that Dolphin can automatically propose methods that are comparable to the state-of-the-art in some tasks such as 3D point classification.

The rapid improvement of deep learning models with the integration of the physical world has dramatically improved embodied AI capabilities. Meanwhile, the powerful embodied AI models and their scales place an increasing burden on deployment efficiency. The efficiency issue is more apparent on embodied AI platforms than on data centers because they have more limited computational resources and memory bandwidth. Meanwhile, most embodied AI scenarios, like autonomous driving and robotics, are more sensitive to fast responses. Theoretically, the traditional model compression techniques can help embodied AI models with more efficient computation, lower memory and energy consumption, and reduced latency. Because the embodied AI models are expected to interact with the physical world, the corresponding compressed models are also expected to resist natural corruption caused by real-world events such as noise, blur, weather conditions, and even adversarial corruption. This paper explores the novel paradigm to boost the efficiency of the embodied AI models and the robust compression boundary. The efficacy of our method has been proven to find the optimal balance between accuracy, efficiency, and robustness in real-world conditions.

Vision-language model (VLM) is one of the most important models for multi-modal tasks. Real industrial applications often meet the challenge of adapting VLMs to different sce narios, such as varying hardware platforms or performance requirements. Traditional methods involve training or fine tuning to adapt multiple unique VLMs or using model com pression techniques to create multiple compact models. These approaches are complex and resource-intensive. This paper introduces a novel paradigm called Once-Tuning Multiple-Variants (OTMV). OTMV requires only a single tuning process to inject dynamic weight expansion capacity into the original VLM structure. This tuned VLM can then be expanded into multiple variants tailored for different sce narios in inference. The tuning mechanism of OTMV is in spired by the mathematical series expansion theorem, which helps to reduce the parameter size and memory require ments while maintaining accuracy for VLM. Experiment re sults show that OTMV-tuned models achieve comparable accuracy to baseline VLMs across various visual-language tasks. The experiments also demonstrate the dynamic ex pansion capability of OTMV-tuned VLMs, outperforming traditional model compression and adaptation techniques in terms of accuracy and efficiency.

Upcycled Mixture-of-Experts (MoE) models have shown great potential in various tasks by converting the original Feed-Forward Network (FFN) layers in pre-trained dense models into MoE layers. However, these models still suffer from significant parameter inefficiency due to the introduction of multiple experts. In this work, we propose a novel DeRS (Decompose, Replace, and Synthesis) paradigm to overcome this shortcoming, which is motivated by our observations about the unique redundancy mechanisms of upcycled MoE experts. Specifically, DeRS decomposes the experts into one expert-shared base weight and multiple expert-specific delta weights, and subsequently represents these delta weights in lightweight forms. Our proposed DeRS paradigm can be applied to enhance parameter efficiency in two different scenarios, including: 1) DeRS Compression for inference stage, using sparsification or quantization to compress vanilla upcycled MoE models; and 2) DeRS Upcycling for training stage, employing lightweight sparse or low-rank matrixes to efficiently upcycle dense models into MoE models. Extensive experiments across three different tasks show that the proposed methods can achieve extreme parameter efficiency while maintaining the performance for both training and compression of upcycled MoE models.

Iterative-based methods have become mainstream in stereo matching due to their high performance. However, these methods heavily rely on labeled data and face challenges with unlabeled real-world data. To this end, we propose a consistency-aware self-training framework for iterative-based stereo matching for the first time, leveraging real-world unlabeled data in a teacher-student manner. We first observe that regions with larger errors tend to exhibit more pronounced oscillation characteristics during model this http URL on this, we introduce a novel consistency-aware soft filtering module to evaluate the reliability of teacher-predicted pseudo-labels, which consists of a multi-resolution prediction consistency filter and an iterative prediction consistency filter to assess the prediction fluctuations of multiple resolutions and iterative optimization respectively. Further, we introduce a consistency-aware soft-weighted loss to adjust the weight of pseudo-labels accordingly, relieving the error accumulation and performance degradation problem due to incorrect pseudo-labels. Extensive experiments demonstrate that our method can improve the performance of various iterative-based stereo matching approaches in various scenarios. In particular, our method can achieve further enhancements over the current SOTA methods on several benchmark datasets.

Reconstructing 3D scenes from multiple viewpoints is a fundamental task in stereo vision. Recently, advances in generalizable 3D Gaussian Splatting have enabled high-quality novel view synthesis for unseen scenes from sparse input views by feed-forward predicting per-pixel Gaussian parameters without extra optimiza tion. However, existing methods typically generate single-scale 3D Gaussians, which lack representation of both large-scale structure and texture details, re sulting in mislocation and artefacts. In this paper, we propose a novel frame work, HiSplat, which introduces a hierarchical manner in generalizable 3D Gaus sian Splatting to construct hierarchical 3D Gaussians via a coarse-to-fine strategy. Specifically, HiSplat generates large coarse-grained Gaussians to capture large scale structures, followed by fine-grained Gaussians to enhance delicate texture details. To promote inter-scale interactions, we propose an Error Aware Module for Gaussian compensation and a Modulating Fusion Module for Gaussian repair. Our method achieves joint optimization of hierarchical representations, allowing for novel view synthesis using only two-view reference images. Comprehensive experiments on various datasets demonstrate that HiSplat significantly enhances reconstruction quality and cross-dataset generalization compared to prior single scale methods. The corresponding ablation study and analysis of different-scale 3D Gaussians reveal the mechanism behind the effectiveness.

Diffusion models have recently achieved great success in the synthesis of high-quality images and videos. However, the existing denoising techniques in diffusion models are commonly based on step-by-step noise predictions, which suffers from high computation cost, resulting in a prohibitive latency for interactive applications. In this paper, we propose AdaptiveDiffusion to relieve this bottleneck by adaptively reducing the noise prediction steps during the denoising process. Our method considers the potential of skipping as many noise prediction steps as possible while keeping the final denoised results identical to the original full-step ones. Specifically, the skipping strategy is guided by the third-order latent difference that indicates the stability between timesteps during the denoising process, which benefits the reusing of previous noise prediction results. Extensive experiments on image and video diffusion models demonstrate that our method can significantly speed up the denoising process while generating identical results to the original process, achieving up to an average 2~5x speedup without quality degradation.

The polygon mesh representation of 3D data exhibits great flexibility, fast rendering speed, and storage efficiency, which is widely preferred in various applications. However, given its unstructured graph representation, the direct generation of high-fidelity 3D meshes is challenging. Fortunately, with a pre-defined ordering strategy, 3D meshes can be represented as sequences, and the generation process can be seamlessly treated as an auto-regressive problem. In this paper, we validate the Neural Coordinate Field (NeurCF), an explicit coordinate representation with implicit neural embeddings, is a simple-yet-effective representation for large scale sequential mesh modeling. After that, we present MeshXL, a family of generative pre-trained auto-regressive models, which addresses the process of 3D mesh generation with modern large language model approaches. Extensive experiments show that MeshXL is able to generate high-quality 3D meshes, and can also serve as foundation models for various down-stream applications.

Recently, 3D understanding has become popular to facilitate autonomous agents to perform further decisionmaking. However, existing 3D datasets and methods are often limited to specific tasks. On the other hand, recent progress in Large Language Models (LLMs) and Multimodal Language Models (MLMs) have demonstrated exceptional general language and imagery tasking performance. Therefore, it is interesting to unlock MLM's potential to be 3D generalist for wider tasks. However, current MLMs' research has been less focused on 3D tasks due to a lack of large-scale 3D instruction-following datasets. In this work, we introduce a comprehensive 3D instructionfollowing dataset called M3DBench, which possesses the following characteristics: 1) It supports general multimodal instructions interleaved with text, images, 3D objects, and other visual prompts. 2) It unifies diverse 3D tasks at both region and scene levels, covering a variety of fundamental abilities in real-world 3D environments. 3) It is a large-scale 3D instruction-following dataset with over 320k instruction-response pairs. Furthermore, we establish a new benchmark for assessing the performance of large models in understanding multi-modal 3D prompts. Extensive experiments demonstrate the effectiveness of our dataset and baseline, supporting general 3D-centric tasks, which can inspire future research.

Recent advancements in language models have demonstrated their adeptness in conducting multi-turn dialogues and retaining conversational context. However, this proficiency remains largely unexplored in other multimodal gen erative models, particularly in human motion models. By integrating multi-turn conversations in controlling continuous virtual human movements, generative human motion models can achieve an intuitive and step-by-step process of human task execution for humanoid robotics, game agents, or other embodied systems. In this work, we present MotionChain, a conversational human motion controller to generate continuous and long-term human motion through multimodal prompts. Specifically, MotionChain consists of multi-modal tokenizers that transform var ious data types such as text, image, and motion, into discrete tokens, coupled with a Vision-Motion-aware Language model. By leveraging large-scale language, vision-language, and vision-motion data to assist motion-related generation tasks, MotionChain thus comprehends each instruction in multi-turn conversation and generates human motions followed by these prompts. Extensive experiments vali date the efficacy of MotionChain, demonstrating state-of-the-art performance in conversational motion generation, as well as more intuitive manners of controlling and interacting with virtual humans.

DomainAdaptation (DA)hasbeenwidely explored and made significant progress on cross-domain 3D tasks recently. Despite being ef fective, existing works fail to deal with rapidly changing domains due to the unpredictable test time scenarios and meanwhile fast response time requirement. Thus, we explore a new task named test-time domain adaptive 3D object detection and propose Reg-TTA3D, a pseudo-label based test-time adaptative 3D object detection method. By investigating the factor that limits the detection accuracy, we find that regression is essential in this task. To make better regression, we first design a noise consistency pseudo-label generation process to filter pseudo-labels with instability under noise interference and obtain reliable pseudo-labels. Then, confidence-guided regression refinement is introduced, which uses the box regression results of high-confidence boxes to supervise boxes with relatively low confidence, further making the predicted box size gradually approach the distribution of the target domain. Finally, to better update the regression layer and alleviate the class-imbalance is sue, a class-balance EMA updating strategy is proposed. Experimental results on multiple cross-domain scenarios including cross-beam, cross location, and cross-weather demonstrate that Reg-TTA3D can achieve comparable or even better performance compared to unsupervised do main adaptation works by only updating less than 0.1% parameters within less than 1% time.

Recent advances in Large Multimodal Models (LMM) have made it possible for various applications in human machine interactions. However, developing LMMs that can comprehend, reason, and plan in complex and diverse 3D environments remains a challenging topic, especially considering the demand for understanding permutation invariant point cloud 3D representations of the 3D scene. Existing works seek help from multi-view images, and project 2D features to 3D space as 3D scene representa tions. This, however, leads to huge computational over head and performance degradation. In this paper, we present LL3DA, a Large Language 3D Assistant that takes point cloud as direct input and respond to both textual instructions and visual-prompts. This help LMMs better comprehend human interactions and further help to remove the ambiguities in cluttered 3D scenes. Experiments show that LL3DAachievesremarkableresults, and surpasses var ious 3D vision-language models on both 3D Dense Cap tioning and 3D Question Answering.

Though the advancement of pre-trained large language models unfolds, the exploration of building a unified model for language and other multi-modal data, such as motion, remains challenging and untouched so far. Fortunately, human motion displays a semantic coupling akin to human language, often perceived as a form of body language. By fusing language data with large-scale motion models, motion-language pre-training that can enhance the performance of motion-related tasks becomes feasible. Driven by this insight, we propose MotionGPT, a unified, versatile, and user-friendly motion-language model to handle multiple motion-relevant tasks. Specifically, we employ the discrete vector quantization for human motion and transfer 3D motion into motion tokens, similar to the generation process of word tokens. Building upon this "motion vocabulary", we perform language modeling on both motion and text in a unified manner, treating human motion as a specific language. Moreover, inspired by prompt learning, we pre-train MotionGPT with a mixture of motion-language data and fine-tune it on prompt-based question-and-answer tasks. Extensive experiments demonstrate that MotionGPT achieves state-of-the-art performances on multiple motion tasks including text-driven motion generation, motion captioning, motion prediction, and motion in-between.

Equivariant networks enhance model efficiency and generalization by embedding symmetry priors into their architectures. However, most existing methods, primarily based on group convolutions and steerable convolutions, face significant limitations when dealing with complex transformation groups, particularly the projective group, which plays a crucial role in vision. In this work, we tackle the challenge by constructing projective equivariant networks based on differential invariants. Using the moving frame method with a carefully selected cross section tailored for multi-dimensional functions, we derive a complete and concise set of second-order fundamental differential invariants of the projective group. We provide a rigorous analysis of the properties and transformation relationships of their underlying components, yielding a further simplified and unified set of fundamental differential invariants, which facilitates both theoretical analysis and practical applications. Building on this foundation, we develop the first deep projective equivariant networks, PDINet, which achieve full projective equivariance without discretizing or sampling the group. Empirical results on the projectively transformed STL-10 and Imagenette datasets show that PDINet achieves improvements of 11.39% and 5.66% in accuracy over the respective standard baselines under out-of-distribution settings, demonstrating its strong generalization to complex geometric transformations.

This paper introduces PaZO, a preconditioned accelerated zeroth-order optimization algorithm for fine-tuning large language models (LLMs). First, we theoretically demonstrate the necessity of preconditioning in zeroth-order optimization, proving that zeroth-order stochastic gradient descent (ZO-SGD) alone fails to achieve the ideal convergence rate. Building on this, we propose a Preconditioned Simultaneous Perturbation Stochastic Approximation (PSPSA) and theoretical version of PaZO, and demonstrate that setting the order of preconditioner as -1/2 in PSPSA yields the improved convergence rate for PaZO. Moreover, we design a practical version of PaZO that stabilizes training via diagonal Hessian estimate and moving average technique. Extensive experiments on diverse downstream tasks with models like RoBERTa-large and OPT show PaZO's effectiveness. Compared to other zeroth order baselines, PaZO achieves better performance across models and tasks.

The expanding computational costs and limited resources underscore the critical need for budgeted-iteration training, which aims to achieve optimal learning within predetermined iteration budgets. While learning rate schedules fundamentally govern the performance of different networks and tasks, particularly in budgeted-iteration scenarios, their design remains largely heuristic, lacking theoretical foundations. In addition, the optimal learning rate schedule requires extensive trial-and-error selection, making the training process inefficient. In this work, we propose the Unified Budget-Aware (UBA) schedule, a theoretically grounded learning rate schedule that consistently outperforms commonly-used schedules among diverse architectures and tasks under different constrained training budgets. First, we bridge the gap by constructing a novel training budget-aware optimization framework, which explicitly accounts for the robustness to landscape curvature variations. From this framework, we derive the UBA schedule, controlled by a single hyper-parameter $\varphi$ that provides a trade-off between flexibility and simplicity, eliminating the need for per-network numerical optimization. Moreover, we establish a theoretical connection between $\varphi$ and the condition number, adding interpretation and justification to our approach. Besides, we prove the convergence for different values of $\varphi$. We offer practical guidelines for its selection via theoretical analysis and empirical results. Extensive experimental results show that UBA \textit{consistently surpasses} the commonly-used schedules across diverse vision and language tasks, spanning network architectures (e.g., ResNet, OLMo) and scales, under different training-iteration budgets.

Convolutional neural networks benefit from translation equivariance, achieving tremendous success. Equivariant networks further extend this property to other transformation groups. However, most existing methods require discretization or sampling of groups, leading to increased model sizes for larger groups, such as the affine group. In this paper, we build affine equivariant networks based on differential invariants from the viewpoint of symmetric PDEs, without discretizing or sampling the group. To address the division-by-zero issue arising from fractional differential invariants of the affine group, we construct a new kind of affine invariants by normalizing polynomial relative differential invariants to replace classical differential invariants. For further flexibility, we design an equivariant layer, which can be directly integrated into convolutional networks of various architectures. Moreover, our frame work for the affine group is also applicable to its continuous subgroups. We implement equivariant networks for the scale group, the rotation-scale group, and the affine group. Numerical experiments demonstrate the outstanding performance of our framework across classification tasks involving transformations of these groups. Remarkably, under the out-of-distribution setting, our model achieves a 3.37% improvement in accuracy over the main counterpart affConv on the affNIST dataset.

Recent research has shown that incorporating equivariance into neural network architectures is very helpful, and there have been some works investigating the equivariance of networks under group actions. However, as digital images and feature maps are on the discrete meshgrid, corresponding equivariance-preserving transformation groups are very limited. 
In this work, we deal with this issue from the connection between convolutions and partial differential operators (PDOs). In theory, assuming inputs to be smooth, we transform PDOs and propose a system which is equivariant to a much more general continuous group, the n-dimension Euclidean group. In implementation, we discretize the system using the numerical schemes of PDOs, deriving approximately equivariant convolutions (PDO-eConvs). Theoretically, the approximation error of PDO-eConvs is of the quadratic order. It is the first time that the error analysis is provided when the equivariance is approximate. Extensive experiments on rotated MNIST and natural image classification show that PDO-eConvs perform competitively yet use parameters much more efficiently. Particularly, compared with Wide ResNets, our methods result in better results using only 12.6% parameters.

Event cameras offer unique advantages in scenarios involving high speed, low light, and high dynamic range, yet their asynchronous and sparse nature poses significant challenges to efficient spatiotemporal representation learning. Specifically, despite notable progress in the field, effectively modeling the full spatiotemporal context, selectively attending to salient dynamic regions, and robustly adapting to the variable density and dynamic nature of event data remain key challenges. Motivated by these challenges, this paper proposes EventMG, a lightweight, efficient, multilevel Mamba-Graph architecture designed for learning high-quality spatiotemporal event representations. EventMG employs a multilevel approach, jointly modeling information at the micro (single event) and macro (event cluster) levels to comprehensively capture the multi-scale characteristics of event data. At the micro-level, it focuses on spatiotemporal details, employing State Space Model (SSM) based Mamba, to precisely capture long-range dependencies among numerous event nodes. Concurrently, at the macro-level, Component Graphs are introduced to efficiently encode the local semantics and global topology of dense event regions. Furthermore, to better accommodate the dynamic and sparse characteristics of data, we propose the Spatiotemporal-aware Event Scanning Technology (SEST), integrating the Adaptive Perturbation Network (APN) and Multidirectional Scanning Module (MSM), which substantially enhances the model's ability to perceive and focus on key spatiotemporal patterns. By employing this novel collaborative paradigm, EventMG demonstrates the ability to effectively capture multi-level spatiotemporal characteristics of event data while maintaining a low parameter count and linear computational complexity, suggesting a promising direction for event representation learning.

The rapid development of Unmanned aerial vehicles (UAVs) technology has spawned a wide variety of applications, such as emergency communications, regional surveillance, and disaster relief. Due to their limited battery capacity and processing power, multiple UAVs are often required for complex tasks. In such cases, a control center is crucial for coordinating their activities, which fits well with the federated learning (FL) framework. However, conventional FL approaches often focus on a single task, ignoring the potential of training multiple related tasks simultaneously. In this paper, we propose a UAVassisted multi-task federated learning scheme, in which data collected by multiple UAVs can be used to train multiple related tasks concurrently. The scheme facilitates the training process by sharing feature extractors across related tasks and introduces a task attention mechanism to balance task performance and encourage knowledge sharing. To provide an analytical description of training performance, the convergence analysis of the proposed scheme is performed. Additionally, the optimal bandwidth allocation for UAVs under limited bandwidth conditions is derived to minimize communication time. Meanwhile, a UAV-EV association strategy based on coalition formation game is proposed. Simulation results validate the effectiveness of the proposed scheme in enhancing multi-task performance and training speed.

Event cameras provide exceptionally high temporal resolution in dynamic vision systems due to their unique event-driven mechanism. However, the sparse and asynchronous nature of event data makes frame-based visual processing methods inappropriate. This study proposes a novel framework, Event-based Graph Spatiotemporal Sensitive Transformer (EGSST), for the exploitation of spatial and temporal properties of event data. Firstly, a well-designed graph structure is employed to model event data, which not only preserves the original temporal data but also captures spatial details. Furthermore, inspired by the phenomenon that human eyes pay more attention to objects that produce significant dynamic changes, we design a Spatiotemporal Sensitivity Module (SSM) and an adaptive Temporal Activation Controller (TAC). Through these two modules, our framework can mimic the response of the human eyes in dynamic environments by selectively activating the temporal attention mechanism based on the relative dynamics of event data, thereby effectively conserving computational resources. In addition, the integration of a lightweight, multi-scale Linear Vision Transformer (LViT) markedly enhances processing efficiency. Our research proposes a fully event-driven approach, effectively exploiting the temporal precision of event data and optimising the allocation of computational resources by intelligently distinguishing the dynamics within the event data. The framework provides a lightweight, fast, accurate, and fully event-based solution for object detection tasks in complex dynamic environments, demonstrating significant practicality and potential for application. 

The beamforming technique has been widely used to improve the link budget in wireless communications. Compared with the digital beamformer, the analog beamformer has much lower hardware complexity and is more suitable for low-cost mobile applications. In this paper, we consider element-level phase and gain control of the analog beamformer using two phase shifters only. By setting the phase shifts properly, simultaneous 360◦ phase and 6-dB gain control (SPGC) can be achieved to form the beam pattern. We first propose a low-complexity SPGC method tailored for massive multiple-input multipleoutput (MIMO). Based on the conventional and proposed SPGC methods, we then design the full-featured accelerator (FFA) and the hardware-efficient accelerator (HEA) to accelerate the computing process. These two accelerators are implemented in 28 nm technology. FFA integrates 137 kilogate equivalents (kGE) in a core area of 0.0693 mm2 and dissipates 104.4 mw at 2.0 GHz with 16 degrees of parallelism, while HEA can reduce the core area by 44.6% and power consumption by 42.7% without significant performance loss.

Sampling and interpolation of continuous graph signals have been extensively studied, in order to reconstruct or estimate the entire graph signal from the signal values on a subset of vertices. Whereas in a lot of real-world scenarios, only the signs of signals are available. For example, a rating system may only provide simple options such as “like” or “dislike”. We are interested in whether it is possible to recover the original signal from such coarse information. In this paper, the reconstruction of bandlimited graph signals based on sign measurements is discussed and a greedy sampling strategy is proposed. The simulation experiments are presented, and the greedy sampling algorithm is compared with the random sampling algorithm, which verifies the feasibility of the proposed approach.

We consider a cognitive radio based Internet of Things (CR-IoT) system where the secondary IoT device (SD) accesses the licensed channel during the transmission vacancies of the primary IoT device (PD). We focus on the impact of the IoT devices’ heterogeneous traffic pattern on the energy efficiency and on the age of information (AoI) performance of the SD. We first derive closed-form expressions of the energy efficiency and the average AoI, and subsequently explore their convexity and monotonicity to the transmit power. Following these characterizations, an optimal transmit power optimization algorithm (TPOA) is proposed for the SD to maximize the energy efficiency while maintaining the average AoI under a predefined threshold. Numerical results verify the different preferences of the SD toward different PD traffic patterns, and provides insights into the tradeoff between the energy efficiency and the average AoI.

Capturing complex high-order interactions among data is an important task in many scenarios. A common way to model high-order interactions is to use hypergraphs whose topology can be mathematically represented by tensors. Existing methods use a fixed-order tensor to describe the topology of the whole hypergraph, which ignores the divergence of differentorder interactions. In this work, we take this divergence into consideration, and propose a multi-order hypergraph Laplacian and the corresponding total variation. Taking this total variation as a regularization term, we can utilize the topology information contained by it to smooth the hypergraph signal. This can help distinguish different-order interactions and represent high-order interactions accurately.

本书概述了机器学习中ADMM 的新进展. 书中全面介绍了各种情形下的ADMM, 包括确定性和随机性的算法、集中式和分布式的算法, 以及求解凸问题和非凸问题的算法, 深入介绍了各个算法的核心思想, 并为算法的收敛性和收敛速度提供了详细的证明。本书面向机器学习和优化领域的研究人员, 也包括人工智能、信号处理、自动控制、网络通信、应用数学等专业的高年级本科生和研究生, 以及从事相关领域产品研发的工程师。

本书概述了机器学习中加速一阶优化算法的新进展。书中全面介绍了各种情形下的加速一阶优化算法，包括确定性和随机性的算法、同步和异步的算法，以求解带约束的问题和无约束的问题、凸问题和非凸问题，对算法思想进行了深入的解读，并对其收敛速度提供了详细的证明。本书面向机器学习和优化领域的研究人员，包括人工智能、信号处理及应用数学特别是计算数学专业高年级本科生、研究生，以及从事人工智能、信号处理领域产品研发的工程师。

Written by experts in machine learning and optimization, this is the first book offering a state-of-the-art review on ADMM under various scenarios, including deterministic and convex optimization, nonconvex optimization, stochastic optimization, and distributed optimization. Offering a rich blend of ideas, theories and proofs, the book is up-to-date and self-contained. It is an excellent reference book for users who are seeking a relatively universal algorithm for constrained problems. Graduate students or researchers can read it to grasp the frontiers of ADMM in machine learning in a short period of time.

Written by experts of machine learning and optimization, this is the first book offering the state-of-the-art review on accelerated first-order optimization algorithms for machine Learning. The book provides a comprehensive introduction to the addressed topic. It covers various methods, including both deterministic and stochastic algorithms, where the algorithms can be synchronous or asynchronous, for both unconstrained and constrained problems, where the problems can be convex or non-convex. Offering a rich blend of ideas, theories and proofs, the book is up-to-date and self-contained. It is an excellent reference book for users who are seeking faster optimization algorithms. Graduate students or researchers can read it to grasp the frontiers of optimization in machine learning in a short period of time.

（该专著获2019年度国家科学技术出版基金资助）

本书参照产业界自动驾驶技术研发的基本流程，充分借鉴了产业界在自动驾驶技术领域中的实际研发经验，以高性能的智能小车和高度仿真的车道沙盘为实验教具和运行环境，深入浅出地讲解自动驾驶技术的原理与实际应用，为初学者打开一扇通往人工智能世界的大门。

本书以帮助初学者如何从无到有地打造出具备自动驾驶功能的智能小车为主线，内容分为看车（了解自动驾驶）、造车（设计智能小车）、开车（收集训练数据）、写车（编写自动驾驶模型）、算车（训练和优化自动驾驶模型）、玩车（部署并验证自动驾驶模型）6章。初学者可以通过边学习理论知识边动手实践的方式，系统学习人工智能的算法理论和应用实例。

本书没有堆砌艰深晦涩的公式推导，力求将枯燥难解的算法原理及模型进行直观的讲解，希望读者在学习的过程中，了解现实中自动驾驶技术的发展并获得运用人工智能解决自动驾驶难题的乐趣。本书适合作为高等院校智能科学与技术、人工智能相关专业的教材，也适合作为人工智能研究人员、开发人员的参考书。