Yan Wu

Zhikang Li, Yan Wu, Xin Hu et al.

In multi-modal image registration, the primary challenge lies in shared structural information extraction. Compared to Transformers, Structured State Space Duality (SSD) offers greater global structural feature extraction with higher efficiency during training and inference. Inspired by these advantages, we propose a novel algorithm for multi-modal image registration, named RegNetMamba-2. Our algorithm incorporates SSD into coarse-to-fine matching process to extract local and global structural features effectively. Firstly, SSD is applied in three different scales for multi-modal feature extraction in our network. To strengthen local representation, we pay more attention on foreground edge and structural information by feature scaling function of SSD. Secondly, for shared feature extraction of input images and multi-modal feature fusion in all scales, we propose cross-modality feature fusion model based on SSD, consisting of Cross-Modality feature Interaction (CMI) module and Multi-Scale feature Fusion (MSF) module. CMI module is designed for cross-modality feature extraction of each scale by SSD in cross form. MSF module is designed to employ a progressive upward fusion in feature-level to obtain fine features, consisting of multi-modal features in all scales. Following coarse-to-fine, the features in 1/8 scale from CMI and 1/2 scale from MSF are collected to calculate matching probability scores. Then we respectively establish matching process by correspondences of pixel-wise. Extensive experiments demonstrate that comparing with state-of-the-art deep-learning based algorithms, RegNetMamba-2 has achieved good effects in both performance and efficiency for multi-modal image registration on the following datasets: VIS-SAR (OSDataset), VIS-IR (LGHD/RoadSence) and VIS-NIR (RGB-NIR sense).

Nathan C. Frey, Daniel Berenberg, Karina Zadorozhny et al. · berkeley

We resolve difficulties in training and sampling from a discrete generative model by learning a smoothed energy function, sampling from the smoothed data manifold with Langevin Markov chain Monte Carlo (MCMC), and projecting back to the true data manifold with one-step denoising. Our Discrete Walk-Jump Sampling formalism combines the contrastive divergence training of an energy-based model and improved sample quality of a score-based model, while simplifying training and sampling by requiring only a single noise level. We evaluate the robustness of our approach on generative modeling of antibody proteins and introduce the distributional conformity score to benchmark protein generative models. By optimizing and sampling from our models for the proposed distributional conformity score, 97-100% of generated samples are successfully expressed and purified and 70% of functional designs show equal or improved binding affinity compared to known functional antibodies on the first attempt in a single round of laboratory experiments. We also report the first demonstration of long-run fast-mixing MCMC chains where diverse antibody protein classes are visited in a single MCMC chain.

Nataša Tagasovska, Nathan C. Frey, Andreas Loukas et al. · berkeley

Deep generative models have emerged as a popular machine learning-based approach for inverse design problems in the life sciences. However, these problems often require sampling new designs that satisfy multiple properties of interest in addition to learning the data distribution. This multi-objective optimization becomes more challenging when properties are independent or orthogonal to each other. In this work, we propose a Pareto-compositional energy-based model (pcEBM), a framework that uses multiple gradient descent for sampling new designs that adhere to various constraints in optimizing distinct properties. We demonstrate its ability to learn non-convex Pareto fronts and generate sequences that simultaneously satisfy multiple desired properties across a series of real-world antibody design tasks.

Jiawei Fu, Yunlong Song, Yan Wu et al.

Autonomous drones can operate in remote and unstructured environments, enabling various real-world applications. However, the lack of effective vision-based algorithms has been a stumbling block to achieving this goal. Existing systems often require hand-engineered components for state estimation, planning, and control. Such a sequential design involves laborious tuning, human heuristics, and compounding delays and errors. This paper tackles the vision-based autonomous-drone-racing problem by learning deep sensorimotor policies. We use contrastive learning to extract robust feature representations from the input images and leverage a two-stage learning-by-cheating framework for training a neural network policy. The resulting policy directly infers control commands with feature representations learned from raw images, forgoing the need for globally-consistent state estimation, trajectory planning, and handcrafted control design. Our experimental results indicate that our vision-based policy can achieve the same level of racing performance as the state-based policy while being robust against different visual disturbances and distractors. We believe this work serves as a stepping-stone toward developing intelligent vision-based autonomous systems that control the drone purely from image inputs, like human pilots.

Yan Wu, Esther Wershof, Sebastian M Schmon et al.

We introduce a comprehensive framework for modeling single cell transcriptomic responses to perturbations, aimed at standardizing benchmarking in this rapidly evolving field. Our approach includes a modular and user-friendly model development and evaluation platform, a collection of diverse perturbational datasets, and a set of metrics designed to fairly compare models and dissect their performance. Through extensive evaluation of both published and baseline models across diverse datasets, we highlight the limitations of widely used models, such as mode collapse. We also demonstrate the importance of rank metrics which complement traditional model fit measures, such as RMSE, for validating model effectiveness. Notably, our results show that while no single model architecture clearly outperforms others, simpler architectures are generally competitive and scale well with larger datasets. Overall, this benchmarking exercise sets new standards for model evaluation, supports robust model development, and furthers the use of these models to simulate genetic and chemical screens for therapeutic discovery.

Yan Wu, Yang Yang, Jun Fan et al.

Radiated noise in unmanned underwater vehicles (UUVs) is an important indicator for characterizing acoustic signatures and evaluating platform performance. To address the strong dependence of traditional physics-based modeling and numerical simulation methods on target structural information and environmental boundary conditions, and their inability to achieve continuous spatial spectrum-response modeling in three-dimensional scenes, this paper proposes a neural radiated-noise field (NRNF). An NRNF represents the UUV radiated-noise spectrum as a continuous function of the three-dimensional UUV position, the three-dimensional hydrophone position, the UUV yaw angle, and the frequency, enabling query-based prediction at arbitrary spatial locations. The proposed method employs sinusoidal encoding for position and frequency, and introduces a learnable three-dimensional scene feature grid to explicitly represent environmental structure and propagation effects. A spectrum-prediction dataset is constructed from lake trials, and the proposed model is evaluated under three settings: horizontal extrapolation, depth extrapolation, and cross-run generalization. Results show that the NRNF achieves an average prediction error of 3.5 dB in the 50 to 5000 Hz band. Horizontal extrapolation is easiest, depth extrapolation is the most challenging, and cross-run generalization is of intermediate difficulty. Further ablation results demonstrate that the scene feature grid significantly improves the prediction stability and spatial generalization of the model.

Karolis Martinkus, Jan Ludwiczak, Kyunghyun Cho et al.

We introduce AbDiffuser, an equivariant and physics-informed diffusion model for the joint generation of antibody 3D structures and sequences. AbDiffuser is built on top of a new representation of protein structure, relies on a novel architecture for aligned proteins, and utilizes strong diffusion priors to improve the denoising process. Our approach improves protein diffusion by taking advantage of domain knowledge and physics-based constraints; handles sequence-length changes; and reduces memory complexity by an order of magnitude, enabling backbone and side chain generation. We validate AbDiffuser in silico and in vitro. Numerical experiments showcase the ability of AbDiffuser to generate antibodies that closely track the sequence and structural properties of a reference set. Laboratory experiments confirm that all 16 HER2 antibodies discovered were expressed at high levels and that 57.1% of the selected designs were tight binders.

Xiang Tan, Run He, Yawen Cui et al.

Class-Incremental Learning (CIL) with pre-trained models (PTMs) aims to sequentially adapt PTMs to new categories without forgetting old knowledge. Built upon PTMs, existing adapter-based methods mainly train models via distinct task-specific adapters, and present a uniform knowledge allocation for each adapter during inference. However, this allocation mechanism ignores the nature of task discrepancy and leads to suboptimal utilization of adapters. Also, under CIL constraint, an allocator is prone to forgetting when tasks evolve. To address these issues, we propose a Non-Forgetting Allocation with Bi-Level Competition (NoFA-BC). NoFA-BC constructs a non-forgetting allocator (NFA) by transforming the allocator training into a recursive least-squares problem and achieves an allocator equivalent to that trained with all data. Based on the NFA, a Bi-Level Competition (BLC) including an intra-task level Winner-Takes-All (WTA) mechanism and inter-task Last-Ones-Fall (LOF) elimination is proposed to provide better allocation of adapter knowledge. WTA extracts the most significant logit within a task to represent the adapter's contribution and LOF suppresses the irrelevant adapters. With BLC, participation ratio of each adapter can be tailored for each input. Moreover, a Stability Enhancement (SE) process is incorporated to further improve the performance of old tasks.

Qianli Xu, Nicolas Gauthier, Wenyu Liang et al.

When deploying a robot to a new task, one often has to train it to detect novel objects, which is time-consuming and labor-intensive. We present TAILOR -- a method and system for object registration with active and incremental learning. When instructed by a human teacher to register an object, TAILOR is able to automatically select viewpoints to capture informative images by actively exploring viewpoints, and employs a fast incremental learning algorithm to learn new objects without potential forgetting of previously learned objects. We demonstrate the effectiveness of our method with a KUKA robot to learn novel objects used in a real-world gearbox assembly task through natural interactions.

Zichao Yan, Yan Wu, Mica Xu Ji et al.

Predicting the effects of perturbations in-silico on cell state can identify drivers of cell behavior at scale and accelerate drug discovery. However, modeling challenges remain due to the inherent heterogeneity of single cell gene expression and the complex, latent gene dependencies. Here, we present PRiMeFlow, an end-to-end flow matching based approach to directly model the effects of genetic and small molecule perturbations in the gene expression space. The distribution-fitting approach taken by PRiMeFlow enables it to accurately approximate the empirical distribution of single-cell gene expression, which we demonstrate through extensive benchmarking inside PerturBench. Through ablation studies, we also validate important model design choices such as operating in gene expression space and parameterizing the velocity field with a U-Net architecture. The PRiMeFlow architecture was used as the basis for the model that won the Generalist Prize in the first ARC Virtual Cell Challenge.

Kaira Samuel, Matthew LaRosa, Kyle McAlpin et al.

Artificial intelligence (AI) has enormous potential to improve Air Force pilot training by providing actionable feedback to pilot trainees on the quality of their maneuvers and enabling instructor-less flying familiarization for early-stage trainees in low-cost simulators. Historically, AI challenges consisting of data, problem descriptions, and example code have been critical to fueling AI breakthroughs. The Department of the Air Force-Massachusetts Institute of Technology AI Accelerator (DAF-MIT AI Accelerator) developed such an AI challenge using real-world Air Force flight simulator data. The Maneuver ID challenge assembled thousands of virtual reality simulator flight recordings collected by actual Air Force student pilots at Pilot Training Next (PTN). This dataset has been publicly released at Maneuver-ID.mit.edu and represents the first of its kind public release of USAF flight training data. Using this dataset, we have applied a variety of AI methods to separate "good" vs "bad" simulator data and categorize and characterize maneuvers. These data, algorithms, and software are being released as baselines of model performance for others to build upon to enable the AI ecosystem for flight simulator training.

Mihaela Rosca, Yan Wu, Chongli Qin et al.

The recipe behind the success of deep learning has been the combination of neural networks and gradient-based optimization. Understanding the behavior of gradient descent however, and particularly its instability, has lagged behind its empirical success. To add to the theoretical tools available to study gradient descent we propose the principal flow (PF), a continuous time flow that approximates gradient descent dynamics. To our knowledge, the PF is the only continuous flow that captures the divergent and oscillatory behaviors of gradient descent, including escaping local minima and saddle points. Through its dependence on the eigendecomposition of the Hessian the PF sheds light on the recently observed edge of stability phenomena in deep learning. Using our new understanding of instability we propose a learning rate adaptation method which enables us to control the trade-off between training stability and test set evaluation performance.

Matthew Kolodner, Mingxuan Ju, Zihao Fan et al.

Improving recommendation systems (RS) can greatly enhance the user experience across many domains, such as social media. Many RS utilize embedding-based retrieval (EBR) approaches to retrieve candidates for recommendation. In an EBR system, the embedding quality is key. According to recent literature, self-supervised multitask learning (SSMTL) has showed strong performance on academic benchmarks in embedding learning and resulted in an overall improvement in multiple downstream tasks, demonstrating a larger resilience to the adverse conditions between each downstream task and thereby increased robustness and task generalization ability through the training objective. However, whether or not the success of SSMTL in academia as a robust training objectives translates to large-scale (i.e., over hundreds of million users and interactions in-between) industrial RS still requires verification. Simply adopting academic setups in industrial RS might entail two issues. Firstly, many self-supervised objectives require data augmentations (e.g., embedding masking/corruption) over a large portion of users and items, which is prohibitively expensive in industrial RS. Furthermore, some self-supervised objectives might not align with the recommendation task, which might lead to redundant computational overheads or negative transfer. In light of these two challenges, we evaluate using a robust training objective, specifically SSMTL, through a large-scale friend recommendation system on a social media platform in the tech sector, identifying whether this increase in robustness can work at scale in enhancing retrieval in the production setting. Through online A/B testing with SSMTL-based EBR, we observe statistically significant increases in key metrics in the friend recommendations, with up to 5.45% improvements in new friends made and 1.91% improvements in new friends made with cold-start users.

Cihan Acar, Kuluhan Binici, Alp Tekirdağ et al.

The use of multi-camera views simultaneously has been shown to improve the generalization capabilities and performance of visual policies. However, the hardware cost and design constraints in real-world scenarios can potentially make it challenging to use multiple cameras. In this study, we present a novel approach to enhance the generalization performance of vision-based Reinforcement Learning (RL) algorithms for robotic manipulation tasks. Our proposed method involves utilizing a technique known as knowledge distillation, in which a pre-trained ``teacher'' policy trained with multiple camera viewpoints guides a ``student'' policy in learning from a single camera viewpoint. To enhance the student policy's robustness against camera location perturbations, it is trained using data augmentation and extreme viewpoint changes. As a result, the student policy learns robust visual features that allow it to locate the object of interest accurately and consistently, regardless of the camera viewpoint. The efficacy and efficiency of the proposed method were evaluated both in simulation and real-world environments. The results demonstrate that the single-view visual student policy can successfully learn to grasp and lift a challenging object, which was not possible with a single-view policy alone. Furthermore, the student policy demonstrates zero-shot transfer capability, where it can successfully grasp and lift objects in real-world scenarios for unseen visual configurations.

Kevin Yuchen Ma, Heng Zhang, Weisi Lin et al.

Generalizing tool manipulation requires both semantic planning and precise physical control. Modern generalist robot policies, such as Vision-Language-Action (VLA) models, often lack the physical grounding required for contact-rich tool manipulation. Conversely, existing contact-aware policies that leverage tactile or haptic sensing are typically instance-specific and fail to generalize across diverse tool geometries. Bridging this gap requires learning representations that are both semantically transferable and physically grounded, yet a fundamental barrier remains: diverse real-world tactile data are prohibitive to collect at scale, while direct zero-shot sim-to-real transfer is challenging due to the complex nonlinear deformation of soft tactile sensors. To address this, we propose Semantic-Contact Fields (SCFields), a unified 3D representation that fuses visual semantics with dense extrinsic contact estimates, including contact probability and force. SCFields is learned through a two-stage Sim-to-Real Contact Learning Pipeline: we first pre-train on large-scale simulation to learn geometry-aware contact priors, then fine-tune on a small set of real data pseudo-labeled via geometric heuristics and force optimization to align real tactile signals. The resulting force-aware representation serves as the dense observation input to a diffusion policy, enabling physical generalization to unseen tool instances. Experiments on scraping, crayon drawing, and peeling demonstrate robust category-level generalization, significantly outperforming vision-only and raw-tactile baselines. Project page: https://kevinskwk.github.io/SCFields/.

Haoran Zhang, Aparna Balagopalan, Nassim Oufattole et al.

Large repositories of image-caption pairs are essential for the development of vision-language models. However, these datasets are often extracted from noisy data scraped from the web, and contain many mislabeled instances. In order to improve the reliability of downstream models, it is important to identify and filter images with incorrect captions. However, beyond filtering based on image-caption embedding similarity, no prior works have proposed other methods to filter noisy multimodal data, or concretely assessed the impact of noisy captioning data on downstream training. In this work, we propose, theoretically justify, and empirically validate LEMoN, a method to identify label errors in image-caption datasets. Our method leverages the multimodal neighborhood of image-caption pairs in the latent space of contrastively pretrained multimodal models to automatically identify label errors. Through empirical evaluations across eight datasets and twelve baselines, we find that LEMoN outperforms the baselines by over 3% in label error detection, and that training on datasets filtered using our method improves downstream captioning performance by more than 2 BLEU points over noisy training.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Xiaobo Zhu, Yan Wu, Qinhu Zhang et al.

Modelling temporal networks for dynamic link prediction of new nodes has many real-world applications, such as providing relevant item recommendations to new customers in recommender systems and suggesting appropriate posts to new users on social platforms. Unlike old nodes, new nodes have few historical links, which poses a challenge for the dynamic link prediction task. Most existing dynamic models treat all nodes equally and are not specialized for new nodes, resulting in suboptimal performances. In this paper, we consider dynamic link prediction of new nodes as a few-shot problem and propose a novel model based on the meta-learning principle to effectively mitigate this problem. Specifically, we develop a temporal encoder with a node-level span memory to obtain a new node embedding, and then we use a predictor to determine whether the new node generates a link. To overcome the few-shot challenge, we incorporate the encoder-predictor into the meta-learning paradigm, which can learn two types of implicit information during the formation of the temporal network through span adaptation and node adaptation. The acquired implicit information can serve as model initialisation and facilitate rapid adaptation to new nodes through a fine-tuning process on just a few links. Experiments on three publicly available datasets demonstrate the superior performance of our model compared to existing state-of-the-art methods.

Dong Guo, Faming Wu, Feida Zhu et al. · pku

We present Seed1.5-VL, a vision-language foundation model designed to advance general-purpose multimodal understanding and reasoning. Seed1.5-VL is composed with a 532M-parameter vision encoder and a Mixture-of-Experts (MoE) LLM of 20B active parameters. Despite its relatively compact architecture, it delivers strong performance across a wide spectrum of public VLM benchmarks and internal evaluation suites, achieving the state-of-the-art performance on 38 out of 60 public benchmarks. Moreover, in agent-centric tasks such as GUI control and gameplay, Seed1.5-VL outperforms leading multimodal systems, including OpenAI CUA and Claude 3.7. Beyond visual and video understanding, it also demonstrates strong reasoning abilities, making it particularly effective for multimodal reasoning challenges such as visual puzzles. We believe these capabilities will empower broader applications across diverse tasks. In this report, we mainly provide a comprehensive review of our experiences in building Seed1.5-VL across model design, data construction, and training at various stages, hoping that this report can inspire further research. Seed1.5-VL is now accessible at https://www.volcengine.com/ (Volcano Engine Model ID: doubao-1-5-thinking-vision-pro-250428)

Zhengbin Zhang, Yan Wu, Hongkun Zhang

Collaborative perception has the potential to significantly enhance perceptual accuracy through the sharing of complementary information among agents. However, real-world collaborative perception faces persistent challenges, particularly in balancing perception performance and bandwidth limitations, as well as coping with localization errors. To address these challenges, we propose Fast2comm, a prior knowledge-based collaborative perception framework. Specifically, (1)we propose a prior-supervised confidence feature generation method, that effectively distinguishes foreground from background by producing highly discriminative confidence features; (2)we propose GT Bounding Box-based spatial prior feature selection strategy to ensure that only the most informative prior-knowledge features are selected and shared, thereby minimizing background noise and optimizing bandwidth efficiency while enhancing adaptability to localization inaccuracies; (3)we decouple the feature fusion strategies between model training and testing phases, enabling dynamic bandwidth adaptation. To comprehensively validate our framework, we conduct extensive experiments on both real-world and simulated datasets. The results demonstrate the superior performance of our model and highlight the necessity of the proposed methods. Our code is available at https://github.com/Zhangzhengbin-TJ/Fast2comm.

Xiuchao Sui, Shaohua Li, Xue Geng et al.

Optical flow estimation aims to find the 2D motion field by identifying corresponding pixels between two images. Despite the tremendous progress of deep learning-based optical flow methods, it remains a challenge to accurately estimate large displacements with motion blur. This is mainly because the correlation volume, the basis of pixel matching, is computed as the dot product of the convolutional features of the two images. The locality of convolutional features makes the computed correlations susceptible to various noises. On large displacements with motion blur, noisy correlations could cause severe errors in the estimated flow. To overcome this challenge, we propose a new architecture "CRoss-Attentional Flow Transformer" (CRAFT), aiming to revitalize the correlation volume computation. In CRAFT, a Semantic Smoothing Transformer layer transforms the features of one frame, making them more global and semantically stable. In addition, the dot-product correlations are replaced with transformer Cross-Frame Attention. This layer filters out feature noises through the Query and Key projections, and computes more accurate correlations. On Sintel (Final) and KITTI (foreground) benchmarks, CRAFT has achieved new state-of-the-art performance. Moreover, to test the robustness of different models on large motions, we designed an image shifting attack that shifts input images to generate large artificial motions. Under this attack, CRAFT performs much more robustly than two representative methods, RAFT and GMA. The code of CRAFT is is available at https://github.com/askerlee/craft.

Chin-Hung Chen, Ivana Nikoloska, Wim van Houtum et al.

This paper presents a fully blind phase-aware expectation-maximization (EM) algorithm for OFDM systems with the phase-shift keying (PSK) modulation. We address the well-known local maximum problem of the EM algorithm for blind channel estimation. This is primarily caused by the unknown phase ambiguity in the channel estimates, which conventional blind EM estimators cannot resolve. To overcome this limitation, we propose to exploit the extrinsic information from the decoder as model evidence metrics. A finite set of candidate models is generated based on the inherent symmetries of PSK modulation, and the decoder selects the most likely candidate model. Simulation results demonstrate that, when combined with a simple convolutional code, the phase-aware EM algorithm reliably resolves phase ambiguity during the initialization stage and reduces the local convergence rate from 80% to nearly 0% in frequency-selective channels with a constant phase ambiguity. The algorithm is invoked only once after the EM initialization stage, resulting in negligible additional complexity during subsequent turbo iterations.

Yuwei Zhang, Yan Wu, Yanming Liu et al.

Object detection methods under known single degradations have been extensively investigated. However, existing approaches require prior knowledge of the degradation type and train a separate model for each, limiting their practical applications in unpredictable environments. To address this challenge, we propose a chain-of-thought (CoT) prompted adaptive enhancer, CPA-Enhancer, for object detection under unknown degradations. Specifically, CPA-Enhancer progressively adapts its enhancement strategy under the step-by-step guidance of CoT prompts, that encode degradation-related information. To the best of our knowledge, it's the first work that exploits CoT prompting for object detection tasks. Overall, CPA-Enhancer is a plug-and-play enhancement model that can be integrated into any generic detectors to achieve substantial gains on degraded images, without knowing the degradation type priorly. Experimental results demonstrate that CPA-Enhancer not only sets the new state of the art for object detection but also boosts the performance of other downstream vision tasks under unknown degradations.

Ting Xu, Xiaoxiao Deng, Xiandong Meng et al.

This paper addresses the challenges posed by the unstructured nature and high-dimensional semantic complexity of electronic health record texts. A deep learning method based on attention mechanisms is proposed to achieve unified modeling for information extraction and multi-label disease prediction. The study is conducted on the MIMIC-IV dataset. A Transformer-based architecture is used to perform representation learning over clinical text. Multi-layer self-attention mechanisms are employed to capture key medical entities and their contextual relationships. A Sigmoid-based multi-label classifier is then applied to predict multiple disease labels. The model incorporates a context-aware semantic alignment mechanism, enhancing its representational capacity in typical medical scenarios such as label co-occurrence and sparse information. To comprehensively evaluate model performance, a series of experiments were conducted, including baseline comparisons, hyperparameter sensitivity analysis, data perturbation studies, and noise injection tests. Results demonstrate that the proposed method consistently outperforms representative existing approaches across multiple performance metrics. The model maintains strong generalization under varying data scales, interference levels, and model depth configurations. The framework developed in this study offers an efficient algorithmic foundation for processing real-world clinical texts and presents practical significance for multi-label medical text modeling tasks.

Xiandong Meng, Yan Wu, Yexin Tian et al.

This paper addresses the challenges of high computational cost and slow inference in deploying large language models. It proposes a distillation strategy guided by multiple teacher models. The method constructs several teacher models and integrates their output probability distributions and intermediate semantic features. This guides the student model to learn from multiple sources of knowledge. As a result, the student model gains stronger language understanding and generation ability while maintaining a small parameter size. To achieve this, the paper introduces a weighted output fusion mechanism, a feature alignment loss function, and an entropy-driven dynamic teacher weighting strategy. These components improve the quality and stability of knowledge transfer during distillation. Under multi-teacher guidance, the student model captures semantic information more effectively and demonstrates strong performance across multiple evaluation metrics. In particular, the method shows high consistency in expression, generalization ability, and task adaptability in tasks such as language modeling, text generation, and multi-task learning. The experiments compare the proposed method with several widely adopted distillation approaches. The results further confirm its overall advantages in perplexity, distillation loss, and generation quality. This study provides a feasible technical path for the efficient compression of large-scale language models. It also demonstrates the effectiveness of multi-teacher collaborative mechanisms in complex language modeling tasks.

Yan Wu, Korrawe Karunratanakul, Zhengyi Luo et al.

Generating natural and physically plausible character motion remains challenging, particularly for long-horizon control with diverse guidance signals. While prior work combines high-level diffusion-based motion planners with low-level physics controllers, these systems suffer from domain gaps that degrade motion quality and require task-specific fine-tuning. To tackle this problem, we introduce UniPhys, a diffusion-based behavior cloning framework that unifies motion planning and control into a single model. UniPhys enables flexible, expressive character motion conditioned on multi-modal inputs such as text, trajectories, and goals. To address accumulated prediction errors over long sequences, UniPhys is trained with the Diffusion Forcing paradigm, learning to denoise noisy motion histories and handle discrepancies introduced by the physics simulator. This design allows UniPhys to robustly generate physically plausible, long-horizon motions. Through guided sampling, UniPhys generalizes to a wide range of control signals, including unseen ones, without requiring task-specific fine-tuning. Experiments show that UniPhys outperforms prior methods in motion naturalness, generalization, and robustness across diverse control tasks.

Chia-Wen Chen, Yan Wu, Korrawe Karunratanakul et al.

Achieving precise, versatile whole-body character control in physics-based animation remains challenging. Recent diffusion-based policies generate rich and expressive motions but typically rely on gradient-based test-time guidance to satisfy task objectives, which is slow and can reduce robustness. We introduce NaP-Control (Navigating Diffusion Prior for Versatile and Fast Character Control), abbreviated as NaP. Our method uses reinforcement learning to manipulate the latent noise of a task-agnostic diffusion policy prior, steering it toward task-specific behaviors for fast, robust control with high motion fidelity. In contrast to methods that rely solely on offline training, NaP interacts with the environment during training to correct motions and optimize task rewards, improving success rates and enabling adaptation to challenging scenarios. By directly predicting task-optimized diffusion noise, NaP eliminates iterative guidance during denoising and enables efficient inference. Experiments show that NaP attains higher success rates and faster inference while preserving natural motion across diverse tasks.

Chin-Hung Chen, Wen-Hung Huang, Boris Karanov et al.

Recently, new types of interference in electric vehicles (EVs), such as converters switching and/or battery chargers, have been found to degrade the performance of wireless digital transmission systems. Measurements show that such an interference is characterized by impulsive behavior and is widely varying in time. This paper uses recorded data from our EV testbed to analyze the impulsive interference in the digital audio broadcasting band. Moreover, we use our analysis to obtain a corresponding interference model. In particular, we studied the temporal characteristics of the interference and confirmed that its amplitude indeed exhibits an impulsive behavior. Our results show that impulsive events span successive received signal samples and thus indicate a bursty nature. To this end, we performed a data-driven modification of a well-established model for bursty impulsive interference, the Markov-Middleton model, to produce synthetic noise realization. We investigate the optimal symbol detector design based on the proposed model and show significant performance gains compared to the conventional detector based on the additive white Gaussian noise assumption.

Boris Karanov, Chin-Hung Chen, Yan Wu et al.

We developed machine learning approaches for data-driven trellis-based soft symbol detection in coded transmission over intersymbol interference (ISI) channels in presence of bursty impulsive noise (IN), for example encountered in wireless digital broadcasting systems and vehicular communications. This enabled us to obtain optimized detectors based on the Bahl-Cocke-Jelinek-Raviv (BCJR) algorithm while circumventing the use of full channel state information (CSI) for computing likelihoods and trellis state transition probabilities. First, we extended the application of the neural network (NN)-aided BCJR, recently proposed for ISI channels with additive white Gaussian noise (AWGN). Although suitable for estimating likelihoods via labeling of transmission sequences, the BCJR-NN method does not provide a framework for learning the trellis state transitions. In addition to detection over the joint ISI and IN states we also focused on another scenario where trellis transitions are not trivial: detection for the ISI channel with AWGN with inaccurate knowledge of the channel memory at the receiver. Without access to the accurate state transition matrix, the BCJR- NN performance significantly degrades in both settings. To this end, we devised an alternative approach for data-driven BCJR detection based on the unsupervised learning of a hidden Markov model (HMM). The BCJR-HMM allowed us to optimize both the likelihood function and the state transition matrix without labeling. Moreover, we demonstrated the viability of a hybrid NN and HMM BCJR detection where NN is used for learning the likelihoods, while the state transitions are optimized via HMM. While reducing the required prior channel knowledge, the examined data-driven detectors with learned trellis state transitions achieve bit error rates close to the optimal full CSI-based BCJR, significantly outperforming detection with inaccurate CSI.

Heng Zhang, Kevin Yuchen Ma, Mike Zheng Shou et al.

Dexterous grasping with multi-fingered hands remains challenging due to high-dimensional articulations and the cost of optimization-based pipelines. Existing end-to-end methods require training on large-scale datasets for specific hands, limiting their ability to generalize across different embodiments. We propose an eigengrasp-based, end-to-end framework for cross-embodiment grasp generation. From a hand's morphology description, we derive a morphology embedding and an eigengrasp set. Conditioned on these, together with the object point cloud and wrist pose, an amplitude predictor regresses articulation coefficients in a low-dimensional space, which are decoded into full joint articulations. Articulation learning is supervised with a Kinematic-Aware Articulation Loss (KAL) that emphasizes fingertip-relevant motions and injects morphology-specific structure. In simulation on unseen objects across three dexterous hands, our model attains a 91.9% average grasp success rate with less than 0.4 seconds inference per grasp. With few-shot adaptation to an unseen hand, it achieves 85.6% success on unseen objects in simulation, and real-world experiments on this few-shot generalized hand achieve an 87% success rate. The code and additional materials will be made available upon publication on our project website https://connor-zh.github.io/cross_embodiment_dexterous_grasping.

Jijun Wang, Yan Wu, Yujian Mo et al.

Existing LiDAR-Camera fusion methods have achieved strong results in 3D object detection. To address the sparsity of point clouds, previous approaches typically construct spatial pseudo point clouds via depth completion as auxiliary input and adopts a proposal-refinement framework to generate detection results. However, introducing pseudo points inevitably brings noise, potentially resulting in inaccurate predictions. Considering the differing roles and reliability levels of each modality, we propose LDRFusion, a novel Lidar-dominant two-stage refinement framework for multi-sensor fusion. The first stage soley relies on LiDAR to produce accurately localized proposals, followed by a second stage where pseudo point clouds are incorporated to detect challenging instances. The instance-level results from both stages are subsequently merged. To further enhance the representation of local structures in pseudo point clouds, we present a hierarchical pseudo point residual encoding module, which encodes neighborhood sets using both feature and positional residuals. Experiments on the KITTI dataset demonstrate that our framework consistently achieves strong performance across multiple categories and difficulty levels.

Yujian Mo, Yan Wu, Junqiao Zhao et al.

Recent advances in foundation models have opened up new possibilities for enhancing 3D perception. In particular, DepthAnything offers dense and reliable geometric priors from monocular RGB images, which can complement sparse LiDAR data in autonomous driving scenarios. However, such priors remain underutilized in LiDAR-based 3D object detection. In this paper, we address the limited expressiveness of raw LiDAR point features, especially the weak discriminative capability of the reflectance attribute, by introducing depth priors predicted by DepthAnything. These priors are fused with the original LiDAR attributes to enrich each point's representation. To leverage the enhanced point features, we propose a point-wise feature extraction module. Then, a Dual-Path RoI feature extraction framework is employed, comprising a voxel-based branch for global semantic context and a point-based branch for fine-grained structural details. To effectively integrate the complementary RoI features, we introduce a bidirectional gated RoI feature fusion module that balances global and local cues. Extensive experiments on the KITTI benchmark show that our method consistently improves detection accuracy, demonstrating the value of incorporating visual foundation model priors into LiDAR-based 3D object detection.

Chin-Hung Chen, Boris Karanov, Wim van Houtom et al.

It has been recognized that the impulsive noise (IN) generated by power devices poses significant challenges to wireless receivers. In this paper, we comprehensively assess the achievable information rate (AIR) for the well-established Markov-Middleton IN model with a phase-shift keying (PSK) input sequence across various channel conditions, including matched and mismatched decoding scenarios. Upon determining information-theoretic bounds, we propose an optimal turbo-differentially encoded (DE)-PSK-IN receiver design based on a commonly used commercial transmission setup consisting of a convolutional encoder, bit-level interleaver, and a DE-PSK symbol mapper. We show that by incorporating the differential decoder into the maximum a-posteriori-based (MAP) IN detector, we can significantly enhance the receiver performance with a 4.5 dB gain compared to the conventional MAP-based turbo-PSK-IN receiver and a gap of around 1 dB to the theoretical bounds. We also propose a suboptimal separate receiver design that can be implemented with half the complexity of the joint design and near-optimal performance. We have evaluated the performance of the proposed receiver designs through extensive simulations, demonstrating their effectiveness in real-world scenarios with limited interleaver depth and mismatched state implementation.

Yan Wu, Lihong Pei, Yukai Han et al.

Long-term traffic emission forecasting is crucial for the comprehensive management of urban air pollution. Traditional forecasting methods typically construct spatiotemporal graph models by mining spatiotemporal dependencies to predict emissions. However, due to the multi-scale entanglement of traffic emissions across time and space, these spatiotemporal graph modeling method tend to suffer from cascading error amplification during long-term inference. To address this issue, we propose a Scale-Disentangled Spatio-Temporal Modeling (SDSTM) framework for long-term traffic emission forecasting. It leverages the predictability differences across multiple scales to decompose and fuse features at different scales, while constraining them to remain independent yet complementary. Specifically, the model first introduces a dual-stream feature decomposition strategy based on the Koopman lifting operator. It lifts the scale-coupled spatiotemporal dynamical system into an infinite-dimensional linear space via Koopman operator, and delineates the predictability boundary using gated wavelet decomposition. Then a novel fusion mechanism is constructed, incorporating a dual-stream independence constraint based on cross-term loss to dynamically refine the dual-stream prediction results, suppress mutual interference, and enhance the accuracy of long-term traffic emission prediction. Extensive experiments conducted on a road-level traffic emission dataset within Xi'an's Second Ring Road demonstrate that the proposed model achieves state-of-the-art performance.

Xinyue Xin, Ming Li, Yan Wu et al.

The collaborative classification of dual-frequency PolSAR images is a meaningful but also challenging research. The effect of regional consistency on classification information learning and the rational use of dual-frequency data are two main difficulties for dual-frequency collaborative classification. To tackle these problems, a selected knowledge distillation network with statistical-based sample rectification (SKDNet-SSR) is proposed in this article. First, in addition to applying CNN and ViT as local and global feature extractors, a statistical-based dynamic sample rectification (SDSR) module is designed to avoid the impact of poor regional consistency on spatial information learning process. Specifically, based on the fact that the PolSAR covariance matrix conforms to the complex Wishart distribution, SDSR first dynamically evaluates the sample purity, and then performs pixel selection and pixel generation to remove noisy pixels, thereby avoiding the feature interaction between informative pixels and noisy pixels and improving the classification feature extraction process. Next, a dual-frequency gate-selected distillation (DGSD) module is constructed to emphasize the advantages of different frequency bands and perform complementary learning on dual-frequency data. It uses the dominant single-frequency branch on each sample as teacher model to train the dual-frequency student model, enabling the student model to learn the optimal results and realizing complementary utilization of dual-frequency data on different terrain objects. Comprehensive experiments on four measured dual-frequency PolSAR data demonstrate that the proposed SKDNet-SSR outperforms other related methods.

Chin-Hung Chen, Ivana Nikoloska, Wim van Houtum et al.

Impulsive noise (IN) commonly generated by power devices can severely degrade the performance of high sensitivity wireless receivers. Accurate channel state information (CSI) knowledge is essential for designing optimal maximum a posteriori detectors. This paper examines blind channel estimation methods based on the expectation-maximization (EM) algorithm tailored for scenarios impacted by bursty IN, which can be described by the Markov-Middleton model. We propose a constrained EM algorithm that exploits the trellis structure of the IN model and the transmitted binary phase shift keying (BPSK) symbols. By enforcing shared variance among specific trellis states and symmetry in the transition matrix, the proposed constrained EM algorithm adapted for the bursty IN channel has an almost two times faster convergence rate and better estimation performance than the standard EM approach. We comprehensively evaluate the robustness of both standard and constrained EM estimators under different types of CSI uncertainties. The results indicate that the final estimations of both EM estimators are robust enough to mismatch Markov-Middleton model parameters. However, as the level of CSI uncertainty increases, the convergence rate decreases.

Xiaobo Zhu, Yan Wu, Zhipeng Li et al.

Recently, representation learning over graph networks has gained popularity, with various models showing promising results. Despite this, several challenges persist: 1) most methods are designed for static or discrete-time dynamic graphs; 2) existing continuous-time dynamic graph algorithms focus on a single evolving perspective; and 3) many continuous-time dynamic graph approaches necessitate numerous temporal neighbors to capture long-term dependencies. In response, this paper introduces the Multi-Perspective Feedback-Attention Coupling (MPFA) model. MPFA incorporates information from both evolving and raw perspectives, efficiently learning the interleaved dynamics of observed processes. The evolving perspective employs temporal self-attention to distinguish continuously evolving temporal neighbors for information aggregation. Through dynamic updates, this perspective can capture long-term dependencies using a small number of temporal neighbors. Meanwhile, the raw perspective utilizes a feedback attention module with growth characteristic coefficients to aggregate raw neighborhood information. Experimental results on a self-organizing dataset and seven public datasets validate the efficacy and competitiveness of our proposed model.

Tianying Wang, En Yen Puang, Marcus Lee et al.

We present an end-to-end Reinforcement Learning(RL) framework for robotic manipulation tasks, using a robust and efficient keypoints representation. The proposed method learns keypoints from camera images as the state representation, through a self-supervised autoencoder architecture. The keypoints encode the geometric information, as well as the relationship of the tool and target in a compact representation to ensure efficient and robust learning. After keypoints learning, the RL step then learns the robot motion from the extracted keypoints state representation. The keypoints and RL learning processes are entirely done in the simulated environment. We demonstrate the effectiveness of the proposed method on robotic manipulation tasks including grasping and pushing, in different scenarios. We also investigate the generalization capability of the trained model. In addition to the robust keypoints representation, we further apply domain randomization and adversarial training examples to achieve zero-shot sim-to-real transfer in real-world robotic manipulation tasks.

Yan Wu, Jiahao Wang, Yan Zhang et al.

The synthesis of human grasping has numerous applications including AR/VR, video games and robotics. While methods have been proposed to generate realistic hand-object interaction for object grasping and manipulation, these typically only consider interacting hand alone. Our goal is to synthesize whole-body grasping motions. Starting from an arbitrary initial pose, we aim to generate diverse and natural whole-body human motions to approach and grasp a target object in 3D space. This task is challenging as it requires modeling both whole-body dynamics and dexterous finger movements. To this end, we propose SAGA (StochAstic whole-body Grasping with contAct), a framework which consists of two key components: (a) Static whole-body grasping pose generation. Specifically, we propose a multi-task generative model, to jointly learn static whole-body grasping poses and human-object contacts. (b) Grasping motion infilling. Given an initial pose and the generated whole-body grasping pose as the start and end of the motion respectively, we design a novel contact-aware generative motion infilling module to generate a diverse set of grasp-oriented motions. We demonstrate the effectiveness of our method, which is a novel generative framework to synthesize realistic and expressive whole-body motions that approach and grasp randomly placed unseen objects. Code and models are available at https://jiahaoplus.github.io/SAGA/saga.html.

Luyu Wang, Pauline Luc, Yan Wu et al.

The ability to learn universal audio representations that can solve diverse speech, music, and environment tasks can spur many applications that require general sound content understanding. In this work, we introduce a holistic audio representation evaluation suite (HARES) spanning 12 downstream tasks across audio domains and provide a thorough empirical study of recent sound representation learning systems on that benchmark. We discover that previous sound event classification or speech models do not generalize outside of their domains. We observe that more robust audio representations can be learned with the SimCLR objective; however, the model's transferability depends heavily on the model architecture. We find the Slowfast architecture is good at learning rich representations required by different domains, but its performance is affected by the normalization scheme. Based on these findings, we propose a novel normalizer-free Slowfast NFNet and achieve state-of-the-art performance across all domains.

Jiayan Gu, Yan Wu, Ashiq Anjum et al.

With the development of Edge Computing and Artificial Intelligence (AI) technologies, edge devices are witnessed to generate data at unprecedented volume. The Edge Intelligence (EI) has led to the emergence of edge devices in various application domains. The EI can provide efficient services to delay-sensitive applications, where the edge devices are deployed as edge nodes to host the majority of execution, which can effectively manage services and improve service discovery efficiency. The multilevel index model is a well-known model used for indexing service, such a model is being introduced and optimized in the edge environments to efficiently services discovery whilst managing large volumes of data. However, effectively updating the multilevel index model by adding new services timely and precisely in the dynamic Edge Computing environments is still a challenge. Addressing this issue, this paper proposes a designated key selection method to improve the efficiency of adding services in the multilevel index models. Our experimental results show that in the partial index and the full index of multilevel index model, our method reduces the service addition time by around 84% and 76%, respectively when compared with the original key selection method and by around 78% and 66%, respectively when compared with the random selection method. Our proposed method significantly improves the service addition efficiency in the multilevel index model, when compared with existing state-of-the-art key selection methods, without compromising the service retrieval stability to any notable level.

Mihaela Rosca, Yan Wu, Benoit Dherin et al.

Gradient-based methods for two-player games produce rich dynamics that can solve challenging problems, yet can be difficult to stabilize and understand. Part of this complexity originates from the discrete update steps given by simultaneous or alternating gradient descent, which causes each player to drift away from the continuous gradient flow -- a phenomenon we call discretization drift. Using backward error analysis, we derive modified continuous dynamical systems that closely follow the discrete dynamics. These modified dynamics provide an insight into the notorious challenges associated with zero-sum games, including Generative Adversarial Networks. In particular, we identify distinct components of the discretization drift that can alter performance and in some cases destabilize the game. Finally, quantifying discretization drift allows us to identify regularizers that explicitly cancel harmful forms of drift or strengthen beneficial forms of drift, and thus improve performance of GAN training.

Wenkai Liang, Yan Wu, Ming Li et al.

Convolutional neural networks (CNNs) have been applied to learn spatial features for high-resolution (HR) synthetic aperture radar (SAR) image classification. However, there has been little work on integrating the unique statistical distributions of SAR images which can reveal physical properties of terrain objects, into CNNs in a supervised feature learning framework. To address this problem, a novel end-to-end supervised classification method is proposed for HR SAR images by considering both spatial context and statistical features. First, to extract more effective spatial features from SAR images, a new deep spatial context encoder network (DSCEN) is proposed, which is a lightweight structure and can be effectively trained with a small number of samples. Meanwhile, to enhance the diversity of statistics, the nonstationary joint statistical model (NS-JSM) is adopted to form the global statistical features. Specifically, SAR images are transformed into the Gabor wavelet domain and the produced multi-subbands magnitudes and phases are modeled by the log-normal and uniform distribution. The covariance matrix is further utilized to capture the inter-scale and intra-scale nonstationary correlation between the statistical subbands and make the joint statistical features more compact and distinguishable. Considering complementary advantages, a feature fusion network (Fusion-Net) base on group compression and smooth normalization is constructed to embed the statistical features into the spatial features and optimize the fusion feature representation. As a result, our model can learn the discriminative features and improve the final classification performance. Experiments on four HR SAR images validate the superiority of the proposed method over other related algorithms.

Jason Ramapuram, Yan Wu, Alexandros Kalousis

Episodic and semantic memory are critical components of the human memory model. The theory of complementary learning systems (McClelland et al., 1995) suggests that the compressed representation produced by a serial event (episodic memory) is later restructured to build a more generalized form of reusable knowledge (semantic memory). In this work we develop a new principled Bayesian memory allocation scheme that bridges the gap between episodic and semantic memory via a hierarchical latent variable model. We take inspiration from traditional heap allocation and extend the idea of locally contiguous memory to the Kanerva Machine, enabling a novel differentiable block allocated latent memory. In contrast to the Kanerva Machine, we simplify the process of memory writing by treating it as a fully feed forward deterministic process, relying on the stochasticity of the read key distribution to disperse information within the memory. We demonstrate that this allocation scheme improves performance in memory conditional image generation, resulting in new state-of-the-art conditional likelihood values on binarized MNIST (<=41.58 nats/image) , binarized Omniglot (<=66.24 nats/image), as well as presenting competitive performance on CIFAR10, DMLab Mazes, Celeb-A and ImageNet32x32.

Yan Wu, Zhiwu Huang, Suryansh Kumar et al.

Modern solutions to the single image super-resolution (SISR) problem using deep neural networks aim not only at better performance accuracy but also at a lighter and computationally efficient model. To that end, recently, neural architecture search (NAS) approaches have shown some tremendous potential. Following the same underlying, in this paper, we suggest a novel trilevel NAS method that provides a better balance between different efficiency metrics and performance to solve SISR. Unlike available NAS, our search is more complete, and therefore it leads to an efficient, optimized, and compressed architecture. We innovatively introduce a trilevel search space modeling, i.e., hierarchical modeling on network-, cell-, and kernel-level structures. To make the search on trilevel spaces differentiable and efficient, we exploit a new sparsestmax technique that is excellent at generating sparse distributions of individual neural architecture candidates so that they can be better disentangled for the final selection from the enlarged search space. We further introduce the sorting technique to the sparsestmax relaxation for better network-level compression. The proposed NAS optimization additionally facilitates simultaneous search and training in a single phase, reducing search time and train time. Comprehensive evaluations on the benchmark datasets show our method's clear superiority over the state-of-the-art NAS in terms of a good trade-off between model size, performance, and efficiency.

Chongli Qin, Yan Wu, Jost Tobias Springenberg et al.

The instability of Generative Adversarial Network (GAN) training has frequently been attributed to gradient descent. Consequently, recent methods have aimed to tailor the models and training procedures to stabilise the discrete updates. In contrast, we study the continuous-time dynamics induced by GAN training. Both theory and toy experiments suggest that these dynamics are in fact surprisingly stable. From this perspective, we hypothesise that instabilities in training GANs arise from the integration error in discretising the continuous dynamics. We experimentally verify that well-known ODE solvers (such as Runge-Kutta) can stabilise training - when combined with a regulariser that controls the integration error. Our approach represents a radical departure from previous methods which typically use adaptive optimisation and stabilisation techniques that constrain the functional space (e.g. Spectral Normalisation). Evaluation on CIFAR-10 and ImageNet shows that our method outperforms several strong baselines, demonstrating its efficacy.

Rhea Sanjay Sukthanker, Zhiwu Huang, Suryansh Kumar et al.

In this paper, we propose a new neural architecture search (NAS) problem of Symmetric Positive Definite (SPD) manifold networks, aiming to automate the design of SPD neural architectures. To address this problem, we first introduce a geometrically rich and diverse SPD neural architecture search space for an efficient SPD cell design. Further, we model our new NAS problem with a one-shot training process of a single supernet. Based on the supernet modeling, we exploit a differentiable NAS algorithm on our relaxed continuous search space for SPD neural architecture search. Statistical evaluation of our method on drone, action, and emotion recognition tasks mostly provides better results than the state-of-the-art SPD networks and traditional NAS algorithms. Empirical results show that our algorithm excels in discovering better performing SPD network design and provides models that are more than three times lighter than searched by the state-of-the-art NAS algorithms.

Xinneng Yang, Yan Wu, Junqiao Zhao et al.

Semantic segmentation has achieved remarkable results with high computational cost and a large number of parameters. However, real-world applications require efficient inference speed on embedded devices. Most previous works address the challenge by reducing depth, width and layer capacity of network, which leads to poor performance. In this paper, we introduce a novel Dense Dual-Path Network (DDPNet) for real-time semantic segmentation under resource constraints. We design a light-weight and powerful backbone with dense connectivity to facilitate feature reuse throughout the whole network and the proposed Dual-Path module (DPM) to sufficiently aggregate multi-scale contexts. Meanwhile, a simple and effective framework is built with a skip architecture utilizing the high-resolution feature maps to refine the segmentation output and an upsampling module leveraging context information from the feature maps to refine the heatmaps. The proposed DDPNet shows an obvious advantage in balancing accuracy and speed. Specifically, on Cityscapes test dataset, DDPNet achieves 75.3% mIoU with 52.6 FPS for an input of 1024 X 2048 resolution on a single GTX 1080Ti card. Compared with other state-of-the-art methods, DDPNet achieves a significant better accuracy with a comparable speed and fewer parameters.

Yan Wu, Aoming Liu, Zhiwu Huang et al.

This paper aims at enlarging the problem of Neural Architecture Search (NAS) from Single-Path and Multi-Path Search to automated Mixed-Path Search. In particular, we model the NAS problem as a sparse supernet using a new continuous architecture representation with a mixture of sparsity constraints. The sparse supernet enables us to automatically achieve sparsely-mixed paths upon a compact set of nodes. To optimize the proposed sparse supernet, we exploit a hierarchical accelerated proximal gradient algorithm within a bi-level optimization framework. Extensive experiments on Convolutional Neural Network and Recurrent Neural Network search demonstrate that the proposed method is capable of searching for compact, general and powerful neural architectures.

Wei Jing, Di Deng, Yan Wu et al.

We present a multi-UAV Coverage Path Planning (CPP) framework for the inspection of large-scale, complex 3D structures. In the proposed sampling-based coverage path planning method, we formulate the multi-UAV inspection applications as a multi-agent coverage path planning problem. By combining two NP-hard problems: Set Covering Problem (SCP) and Vehicle Routing Problem (VRP), a Set-Covering Vehicle Routing Problem (SC-VRP) is formulated and subsequently solved by a modified Biased Random Key Genetic Algorithm (BRKGA) with novel, efficient encoding strategies and local improvement heuristics. We test our proposed method for several complex 3D structures with the 3D model extracted from OpenStreetMap. The proposed method outperforms previous methods, by reducing the length of the planned inspection path by up to 48%