AAAI 2025 论文列表

To diminish the substantial communication costs incurred by federated learning during the training of the global model and enhance the model update efficiency across both clients and server domains, we have integrated knowledge distillation into the federated learning framework. This integration has led to the development of a novel approach termed ClientsToServerKDFL, which streamlines the distillation process by directly transferring model insights from clients to the server for computational learning without the need for extensive computations across numerous clients. This iterative process ensures model accuracy and curtails communication expenses. Experimental data analysis has validated the efficacy of this algorithm.

We design sensitivity oracles for error-prone networks. For a network problem Π, the data structure preprocesses a network G=(V,E) and sensitivity parameter f such that, for any set F of up to f link or node failures, it can report the solution of Π in G-F. We study three network problems Π. - L-Hop Shortest Path: Given s,t in V, is there a shortest s-t-path in G-F with at most L links? - k-Path: Does G-F contain a simple path with k links? - k-Clique: Does G-F contain a clique of k nodes? Our main technical contribution is a new construction of (L,f)-replacement path coverings ((L,f)-RPC) in the parameter realm where f = o(log L). An (L,f)-RPC is a family G' of subnetworks of G which, for every set F of at most f links, has a subfamily G'_F such that (i) no subnetwork in G'_F contains a link of F and (ii) for each s,t in V, if G-F contains a shortest s-t-path with at most L links, then some subnetwork in G'_F retains at least one such path. Our (L,f)-RPC has almost the same size as the one by Weimann and Yuster (2013) but it improves the time to query G'_F from Õ(f^2 L^f) to Õ(f^(5/2) L^o(1)). It also improves over the size and query time of the (L,f)-RPC by Karthik and Parter (2021) by nearly a factor of L. From this construction, we derive oracles for L-Hop Shortest Path, k-Path, and k-Clique. Notably, our solution for k-Path improves the query time of the one by Bilò for f=o(log k).

Event cameras have recently been introduced into image semantic segmentation, owing to their high temporal resolution and other advantageous properties. However, existing event-based semantic segmentation methods often fail to fully exploit the complementary information provided by frames and events, resulting in complex training strategies and increased computational costs. To address these challenges, we propose an efficient hybrid framework for image semantic segmentation, comprising a Spiking Neural Network branch for events and an Artificial Neural Network branch for frames. Specifically, we introduce three specialized modules to facilitate the interaction between these two branches: the Adaptive Temporal Weighting (ATW) Injector, the Event-Driven Sparse (EDS) Injector, and the Channel Selection Fusion (CSF) module. The ATW Injector dynamically integrates temporal features from event data into frame features, enhancing segmentation accuracy by leveraging critical dynamic temporal information. The EDS Injector effectively combines sparse event data with rich frame features, ensuring precise temporal and spatial information alignment. The CSF module selectively merges these features to optimize segmentation performance. Experimental results demonstrate that our framework not only achieves state-of-the-art accuracy across the DDD17-Seg, DSEC-Semantic, and M3ED-Semantic datasets but also significantly reduces energy consumption, achieving a 65% reduction on the DSEC-Semantic dataset.

The domain gap resulting from mismatches in acquisition details like protocol and scanner between training and test data hinders the deployment of the trained model in clinical practice. To address this issue, Continual test-time adaptation (CTTA) has been proposed to adapt the source model to continually changing unlabeled domains without accessing the source data. Existing methods learn an image-level visual prompt for target domains and inject the trainable prompt into the input space. However, they either combine the input with a prompt of equal scale or determine the prompt injection position through complex strategies such as uncertainty estimation or Fourier Transform. These approaches substantially increase the number of trainable parameters and computational burden, especially in high-dimensional medical imaging data. To overcome these challenges, we propose the Efficient Deformable Convolutional Prompt (EDCP), which leverages the inductive bias of convolution to reduce trainable parameters compared to standard prompts. We further enhance convolution by making it deformable, addressing fine-grained domain shifts at the pixel level through an offset branch. To improve training efficiency and balance parameters between the convolution and offset branches, we decompose the offset transformation into two parts, storing one in an offset bank that also serves as a domain indicator. This bank accelerates training by skipping test images similar to those already stored. Prompt updates are guided by layer-wise alignment of source-target statistics without unfreezing batch normalization layers. Extensive experiments demonstrate the superiority of our method in 2D and 3D medical image segmentation tasks.

Reconstructing the intricate local morphology of neurons and their long-range projecting axons can address many connectivity related questions in neuroscience. The main bottleneck in connectomics pipelines is correcting topological errors, as multiple entangled neuronal arbors is a challenging instance segmentation problem. More broadly, segmentation of curvilinear, filamentous structures continues to pose significant challenges. To address this problem, we extend the notion of simple points from digital topology to connected sets of voxels (i.e. supervoxels) and propose a topology-aware neural network segmentation method with minimal computational overhead. We demonstrate its effectiveness on a new public dataset of 3-d light microscopy images of mouse brains, along with the benchmark datasets DRIVE, ISBI12, and CrackTree.

A key challenge in multi-agent collaborative tasks is reducing uncertainty about teammates to enhance cooperative performance. Explicit communication methods can reduce uncertainty about teammates, but the associated high communication costs limit their practicality. Alternatively, implicit consensus learning can promote cooperation without incurring communication costs. However, its performance declines significantly when local observations are severely limited. This paper introduces a novel multi-agent learning framework that combines the strengths of these methods. In our framework, agents generate a consensus about the group based on their local observations and then use both the consensus and local observations to produce messages. Since the consensus provides a certain level of global guidance, communication can be disabled when not essential, thereby reducing overhead. Meanwhile, communication can provide supplementary information to the consensus when necessary. Experimental results demonstrate that our algorithm significantly reduces inter-agent communication overhead while ensuring efficient collaboration.

Transformer-based Super-Resolution (SR) methods have demonstrated superior performance compared to convolutional neural network (CNN)-based SR approaches due to their capability to capture long-range dependencies. However, their high computational complexity necessitates the development of lightweight approaches for practical use. To address this challenge, we propose the Attention-Sharing Information Distillation (ASID) network, a lightweight SR network that integrates attention-sharing and an information distillation structure specifically designed for Transformer-based SR methods. We modify the information distillation scheme, originally designed for efficient CNN operations, to reduce the computational load of stacked self-attention layers, effectively addressing the efficiency bottleneck. Additionally, we introduce attention-sharing across blocks to further minimize the computational cost of self-attention operations. By combining these strategies, ASID achieves competitive performance with existing SR methods while requiring only around 300K parameters – significantly fewer than existing CNN-based and Transformer-based SR models. Furthermore, ASID outperforms state-of-the-art SR methods when the number of parameters is matched, demonstrating its efficiency and effectiveness.

Efficient anomaly detection of irregular sequences, especially those characterized by non-uniform sampling from discontinuous operations or unreliable sensors, presents challenges across various fields. In response, this paper introduces irregular-sequence classification in ''Ct-Echo Model Space''. A novel Continuous-time Echo Network (Ct-Echo) is proposed to fit irregular sequences, efficiently capturing their inherent dynamic characteristics. Ct-Echo utilizes the ''Echo'' mechanism, where history information influences the current state and diminishes over time, and employs Ordinary Differential Equation (ODE) to construct continuous-time transition of hidden states. Each sequence is individually fitted via Ct-Echo to derive a readout model. These fitted models, capturing the dynamic characteristics of the original data, serve as representations of the corresponding sequences, thus mapping the original data from the data space to the Ct-Echo model space. Anomaly detection is further performed in this model space, evaluating differences between models rather than directly on the original sequences. Our method enhances real-time processing and lessens reliance on the amount of labeled training data, as demonstrated by experimental studies.

Spiking Neural Networks (SNNs) provide an energy-efficient way to extract 3D spatio-temporal features. Point clouds are sparse 3D spatial data, which suggests that SNNs should be well-suited for processing them. However, when applying SNNs to point clouds, they often exhibit limited performance and fewer application scenarios. We attribute this to inappropriate preprocessing and feature extraction methods. To address this issue, we first introduce the Spike Voxel Coding (SVC) scheme, which encodes the 3D point clouds into a sparse spike train space, reducing the storage requirements and saving time on point cloud preprocessing. Then, we propose a Spike Sparse Convolution (SSC) model for efficiently extracting 3D sparse point cloud features. Combining SVC and SSC, we design an efficient 3D SNN backbone (E-3DSNN), which is friendly with neuromorphic hardware. For instance, SSC can be implemented on neuromorphic chips with only minor modifications to the addressing function of vanilla spike convolution. Experiments on ModelNet40, KITTI, and Semantic KITTI datasets demonstrate that E-3DSNN achieves state-of-the-art (SOTA) results with remarkable efficiency. Notably, our E-3DSNN (1.87M) obtained 91.7% top-1 accuracy on ModelNet40, surpassing the current best SNN baselines (14.3M) by 3.0%. To our best knowledge, it is the first direct training 3D SNN backbone that can simultaneously handle various 3D computer vision tasks (e.g., classification, detection, and segmentation) with an event-driven nature.

This paper targets on the regularization effect of momentum-based methods in regression settings and analyzes the popular diagonal linear networks to precisely characterize the implicit bias of continuous versions of heavy-ball (HB) and Nesterov's method of accelerated gradients (NAG). We show that, HB and NAG exhibit different implicit bias compared to GD for diagonal linear networks, which is different from the one for classic linear regression problem where momentum-based methods share the same implicit bias with GD. Specifically, the role of momentum in the implicit bias of GD is twofold: (a) HB and NAG induce extra initialization mitigation effects similar to SGD that are beneficial for generalization of sparse regression; (b) the implicit regularization effects of HB and NAG also depend on the initialization of gradients explicitly, which may not be benign for generalization. As a result, whether HB and NAG have better generalization properties than GD jointly depends on the aforementioned twofold effects determined by various parameters such as learning rate, momentum factor, and integral of gradients. Our findings highlight the potential beneficial role of momentum and can help understand its advantages in practice such as when it will lead to better generalization performance.

Flight trajectory data plays a vital role in the traffic management community, especially for downstream tasks such as trajectory prediction, flight recognition, and anomaly detection. Existing works often utilize handcrafted features and design models for different tasks individually, which heavily rely on domain expertise and are hard to extend. We argue that different flight analysis tasks share the same useful features of the trajectory. Jointly learning a unified representation for flight trajectories could be beneficial for improving the performance of various tasks. However, flight trajectory representation learning (TRL) faces two primary challenges, \ie unbalanced behavior density and 3D spatial continuity, which disable recent general TRL methods. In this paper, we propose Flight2Vec, a flight-specific representation learning method to address these challenges. Specifically, a behavior-adaptive patching mechanism is used to inspire the learned representation to pay more attention to behavior-dense segments. Moreover, we introduce a motion trend learning technique that guides the model to memorize not only the precise locations, but also the motion trend to generate better representations. Extensive experimental results demonstrate that Flight2Vec significantly improves performance in downstream tasks such as flight trajectory prediction, flight recognition, and anomaly detection.

Recent advances indicate that diffusion model holds great promise in image super-resolution. While latest methods are primarily based on latent diffusion models with convolutional neural networks, there are few attempts to explore transformers, which have demonstrated remarkable performance in image generation. In this work, we design an effective diffusion transformer for image super resolution (DiT-SR) that achieves the visual quality of prior-based methods, but through a training-from-scratch manner. In practice, DiT-SR leverages an overall U-shaped architecture, and adopts uniform isotropic design for all the transformer blocks across different stages. The former facilitates multi-scale hierarchical feature extraction, while the latter reallocate the computational resources to critical layers to further enhance performance. Moreover, we thoroughly analyze the limitation of the widely used AdaLN, and present a frequency-adaptive time-step conditioning module, enhancing the model's capacity to process distinct frequency information at different time steps. Extensive experiments demonstrate that DiT-SR outperforms the existing training-from-scratch diffusion-based SR methods significantly, and even beats some of the prior-based methods on pretrained Stable Diffusion, proving the superiority of diffusion transformer in image super resolution.

Model editing is a novel research topic in large language models (LLMs), aimed at efficiently handling various knowledge editing tasks. Since irrelevant knowledge is difficult to measure, existing editing methods often lack explicit ways to preserve it, especially for editing methods based on the fine-tuning paradigm. They generally control the locality performance of model editing by constraining the range of changes in model parameters. However, their performance improvements are not always ideal, and may even lead to a decrease in the editing reliability. In this paper, we try to explore effective editing locality control methods based on the relationship between the stored knowledge and the strongly associated model components. Based on the discovery of ``knowledge neurons'' and enough experimental results, we further explore the potential characteristics between knowledge and model components, confirm and point out: (1) only 1% neurons have significant contributions to specific knowledge storage, and (2) these targeted neurons often have a high overlap for knowledge with similar relational descriptions, which means that knowledge with similar relationships may be severely affected when these targeted neurons are modified. Based on these findings, we propose Targeted Neurons Fine-tuning with Data Augmentation (TNF-DA), which performs data augmentation based on the relational representation of edited knowledge to improve editing locality. By freezing most of the model parameters and only fine-tuning the highly contributing neurons corresponding to the edited knowledge, we obtain desirable results in terms of generalization and specificity compared with previous fine-tuning-based methods. Extensive experiments have demonstrated the superior editing performance achieved by our proposed method.

The rapid development of diffusion models has significantly advanced AI-generated content (AIGC), particularly in Text-to-Image (T2I) and Text-to-Video (T2V) generation. Text-based video editing, leveraging these generative capabilities, has emerged as a promising field, enabling precise modifications to videos based on text prompts. Despite the proliferation of innovative video editing models, there is a conspicuous lack of comprehensive evaluation benchmarks that holistically assess these models’ performance across various dimensions. Existing evaluations are limited and inconsistent, typically summarizing overall performance with a single score, which obscures models’ effectiveness on individual editing tasks. To address this gap, we propose EditBoard, the first comprehensive evaluation benchmark for text-based video editing models. EditBoard encompasses nine automatic metrics across four dimensions, evaluating models on four task categories and introducing three new metrics to assess fidelity. This task-oriented benchmark facilitates objective evaluation by detailing model performance and providing insights into each model’s strengths and weaknesses. By open-sourcing EditBoard, we aim to standardize evaluation and advance the development of robust video editing models.

Graph contrastive learning (GCL) aims to learn representations from unlabeled graph data in a self-supervised manner and has developed rapidly in recent years. However, edge-level contrasts are not well explored by most existing GCL methods. Most studies in GCL only regard edges as auxiliary information while updating node features. One of the primary obstacles of edge-based GCL is the heavy computation burden. To tackle this issue, we propose a model that can efficiently learn edge features for GCL, namely Augmentation-Free Edge Contrastive Learning (AFECL) to achieve edge-edge contrast. AFECL depends on no augmentation consisting of two parts. Firstly, we design a novel edge feature generation method, where edge features are computed by embedding concatenation of their connected nodes. Secondly, an edge contrastive learning scheme is developed, where edges connecting the same nodes are defined as positive pairs, and other edges are defined as negative pairs. Experimental results show that compared with recent state-of-the-art GCL methods or even some supervised GNNs, AFECL achieves SOTA performance on link prediction and semi-supervised node classification of extremely scarce labels.

Many approaches to program synthesis perform a combinatorial search within a large space of programs to find one that satisfies a given specification. To tame the search space blowup, previous works introduced probabilistic and neural approaches to guide this combinatorial search by inducing heuristic cost functions. Best-first search algorithms ensure to search in the exact order induced by the cost function, significantly reducing the portion of the program space to be explored. We present a new best-first search algorithm called Eco Search, which is the first no-delay algorithm for pre-generation cost function: the amount of compute required between outputting two programs is constant, and in particular does not increase over time. This key property yields important speedups: we observe that Eco Search outperforms its predecessors on two classical domains.

The area of portrait image animation, propelled by audio input, has witnessed notable progress in the generation of lifelike and dynamic portraits. Conventional methods are limited to utilizing either audios or facial key points to drive images into videos, while they can yield satisfactory results, certain issues exist. For instance, methods driven solely by audios can be unstable at times due to the relatively weaker audio signal, while methods driven exclusively by facial key points, although more stable in driving, can result in unnatural outcomes due to the excessive control of key point information. In addressing the previously mentioned challenges, in this paper, we introduce a novel approach which we named EchoMimic. EchoMimic is concurrently trained using both audios and facial landmarks. Through the implementation of a novel training strategy, EchoMimic is capable of generating portrait videos not only by audios and facial landmarks individually, but also by a combination of both audios and selected facial landmarks. EchoMimic has been comprehensively compared with alternative algorithms across various public datasets and our collected dataset, showcasing superior performance in both quantitative and qualitative evaluations. The code and models are available on the project page.

To extract spatial information, depth estimation using conventional echo-based methods typically employs models with encoder-decoder architectures, such as UNet. However, these methods may face challenges in extracting fine details from echo waveforms and handling multi-scale feature extraction with high precision. To address these challenges, we introduce EchoDiffusion, a framework that incorporates diffusion models conditioned on waveform embeddings for echo-based depth estimation. This framework employs the Multi-Scale Adaptive Latent Feature Network (MALF-Net) to extract multi-scale spatial features and perform adaptive fusion, encoding the echo spectrograms into the latent space. Additionally, we propose the Echo Waveform Detail Embedder (EWDE), which leverages a pre-trained Wav2Vec model to extract detailed spatial information from echo waveforms, using these details as conditional inputs to guide the reverse diffusion process in the latent space. By embedding the echo waveforms into the reverse diffusion process, we can more accurately guide the generation of depth maps. Our extensive evaluations on the Replica and Matterport3D datasets demonstrate that EchoDiffusion establishes new benchmarks for state-of-the-art performance in echo-based depth estimation.

Concept drift, characterized by unpredictable changes in data distribution over time, poses significant challenges to machine learning models in streaming data scenarios. Although error rate-based concept drift detectors are widely used, they often fail to identify drift in the early stages when the data distribution changes but error rates remain constant. This paper introduces the Prediction Uncertainty Index (PU-index), derived from the prediction uncertainty of the classifier, as a superior alternative to the error rate for drift detection. Our theoretical analysis demonstrates that: (1) The PU-index can detect drift even when error rates remain stable. (2) Any change in the error rate will lead to a corresponding change in the PU-index. These properties make the PU-index a more sensitive and robust indicator for drift detection compared to existing methods. We also propose a PU-index-based Drift Detector (PUDD) that employs a novel Adaptive PU-index Bucketing algorithm for detecting drift. Empirical evaluations on both synthetic and real-world datasets demonstrate PUDD’s efficacy in detecting drift in structured and image data.