Clustered Federated Learning (CFL) addresses the challenges posed by non-IID data by training multiple group- or cluster-specific expert models. However, existing methods often overlook the shared information across clusters, which represents the generalizable knowledge valuable to all participants in the Federated Learning (FL) system. To overcome this limitation, this paper introduces a novel FL framework that distills a universal expert model from the knowledge of multiple clusters. This universal expert captures globally shared information across all clients and is subsequently distributed to each client as the initialization for the next round of model training. The proposed FL framework operates in three iterative steps: (1) local model training at each client, (2) cluster-specific model aggregation, and (3) universal expert distillation. This three-step learning paradigm ensures the preservation of fine-grained non-IID characteristics while effectively incorporating shared knowledge across clusters. Compared to traditional gradient-based aggregation methods, the distillation-based model aggregation introduces greater flexibility in handling model heterogeneity and reduces conflicts among cluster-specific experts. Extensive experimental results demonstrate the superior performance of the proposed method across various scenarios, highlighting its potential to advance the state of CFL by balancing personalized and shared knowledge more effectively.
论文检索
输入标题、作者或关键词，从 1,280 篇学术成果中精准定位
CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting
PDF ↗Multivariate long-term time series forecasting is critical for applications such as weather prediction, and traffic analysis. In addition, the implementation of Transformer variants has improved prediction accuracy. Following these variants, different input data process approaches also enhanced the field, such as tokenization techniques including point-wise, channel-wise, and patch-wise tokenization. However, previous studies still have limitations in time complexity, computational resources, and cross-dimensional interactions. To address these limitations, we introduce a novel CNN Autoencoder-based Score Attention mechanism (CASA), which can be introduced in diverse Transformers model-agnosticically by reducing memory and leading to improvement in model performance. Experiments on eight real-world datasets validate that CASA decreases computational resources by up to 77.7%, accelerates inference by 44.0%, and achieves state-of-the-art performance, ranking first in 87.5% of evaluated metrics. Our code is available at https://github.com/lmh9507/CASA.
The principal portfolio approach is an emerging method in signal-based trading. However, these principal portfolios may not be diversified to explore the key features of the prediction matrix or robust to different situations. To address this problem, we propose a novel linear trading position with sparse spectrum that can explore a larger spectral region of the prediction matrix. We also develop a Krasnosel'skii-Mann fixed-point algorithm to optimize this trading position, which possesses the descent property and achieves a linear convergence rate in the objective value. This is a new theoretical result for this type of algorithms. Extensive experiments show that the proposed method achieves good and robust performance in various situations.
Multi-label learning (MLL) has gained attention for its ability to represent real-world data. Label Distribution Learning (LDL), an extension of MLL to learning from label distributions, faces challenges in collecting accurate label distributions. To address the issue of biased annotations, based on the low-rank assumption, existing works recover true distributions from biased observations by exploring the label correlations. However, recent evidence shows that the label distribution tends to be full-rank, and naive apply of low-rank approximation on biased observation leads to inaccurate recovery and performance degradation. In this paper, we address the LDL with biased annotations problem from a novel perspective, where we first degenerate the soft label distribution into a hard multi-hot label and then recover the true label information for each instance. This idea stems from an insight that assigning hard multi-hot labels is often easier than assigning a soft label distribution, and it shows stronger immunity to noise disturbances, leading to smaller label bias. Moreover, assuming that the multi-label space for predicting label distributions is low-rank offers a more reasonable approach to capturing label correlations. Theoretical analysis and experiments confirm the effectiveness and robustness of our method on real-world datasets.
Learning robust representations from data often requires scale, which has led to the success of recent zero-shot models such as CLIP. However, the obtained robustness can easily be deteriorated when these models are fine-tuned on other downstream tasks (e.g., of smaller scales). Previous works often interpret this phenomenon in the context of domain shift, developing fine-tuning methods that aim to preserve the original domain as much as possible. However, in a different context, fine-tuned models with limited data are also prone to learning features that are spurious to humans, such as background or texture. In this paper, we propose StarFT (Spurious Textual Alignment Regularization), a novel framework for fine-tuning zero-shot models to enhance robustness by preventing them from learning spuriosity. We introduce a regularization that aligns the output distribution for spuriosity-injected labels with the original zero-shot model, ensuring that the model is not induced to extract irrelevant features further from these descriptions. We leverage recent language models to get such spuriosity-injected labels by generating alternative textual descriptions that highlight potentially confounding features. Extensive experiments validate the robust generalization of StarFT and its emerging properties: zero-shot group robustness and improved zero-shot classification. Notably, StarFT boosts both worst-group and average accuracy by 14.30% and 3.02%, respectively, in the Waterbirds group shift scenario, where other robust fine-tuning baselines show even degraded performance.
Dynamic edge networks revolutionize mobile edge computing by enabling real-time applications in intelligent transportation, augmented reality, and industrial Internet of Things (IoT). Efficient workload offloading in dynamic edge networks is crucial for addressing the increasing demands of time-varying workloads while contending with limited computational and communication resources. Existing deep reinforcement learning (DRL)-based offloading decision-making schemes are inadequate for managing scenarios involving multiple workloads and edge servers, particularly when faced with time-varying workload arrivals and fluctuating channel states. To this end, we propose a flexible module weighted fusion DRL framework (DRL-MWF) for scalable and robust multi-workload offloading in edge environments. Unlike traditional monolithic networks, DRL-MWF employs a weighted fusion modular architecture that adapts flexibly to diverse workload distributions. Specifically, DRL-MWF introduces a state representation and normalization strategy to model state and workload characteristics, enabling precise and adaptive decision-making. Furthermore, we design two key mechanisms: a weighted policy correction method to stabilize learning and a prioritized experience replay with weighted importance sampling to accelerate convergence by emphasizing critical transitions. Extensive evaluations on real-world datasets demonstrate that DRL-MWF consistently outperforms state-of-the-art baselines. These results reveal DRL-MWF's potential to transform workload offloading in next-generation edge computing systems, ensuring high performance in dynamic scenarios.
Mixed precision quantization (MPQ) is an effective quantization approach to achieve accuracy-complexity trade-off of neural network, through assigning different bit-widths to network activations and weights in each layer. The typical way of existing MPQ methods is to optimize quantization policies (i.e., bit-width allocation) in a gradient descent manner, termed as Differentiable MPQ (DMPQ). At the end of the search, the bit-width associated to the quantization parameters which has the largest value will be selected to form the final mixed precision quantization policy, with the implicit assumption that the values of quantization parameters reflect the operation contribution to the accuracy improvement. While much has been discussed about the MPQ’s improvement, the bit-width selection process has received little attention. We study this problem and argue that the magnitude of quantization parameters does not necessarily reflect the actual contribution of the bit-width to the task performance. Then, we propose a Shapley-based MPQ (SMPQ) method, which measures the bit-width operation’s direct contribution on the MPQ task. To reduce computation cost, a Monte Carlo sampling-based approximation strategy is proposed for Shapley computation. Extensive experiments on mainstream benchmarks demonstrate that our SMPQ consistently achieves state-of-the-art performance than gradient-based competitors.
A Multi-view Fusion Approach for Enhancing Speech Signals via Short-time Fractional Fourier Transform
PDF ↗Deep learning-based speech enhancement (SE) methods focus on reconstructing speech from the time or frequency domain. However, these domains cannot provide enough information to capture the dynamics of non-stationary signals accurately. To enrich information, this work proposes a multi-view fusion SE method (MFSE). Specifically, MFSE extends the representation space of speech to the dynamic domain (also called fractional domain) between the time and frequency domains by using the short-time fractional Fourier transform (STFrFT). Subsequently, we construct inputs as modes of the primary short-time Fourier transform (STFT) spectrum and the auxiliary STFrFT spectrum views and adaptively identify the optimal fractional STFrFT spectrum from the infinitely continuous fractional domain by leveraging the average spectral centroids. The framework extracts potential features through multiple designed convolutional modules and captures the correlation between different speech frequencies through multi-granularity attention. Experimental results show that the proposed method significantly improves performance in several metrics compared to existing single-channel SE methods based on time and frequency domains. Furthermore, the results of its generalizability evaluation show that the multi-view method outperforms the single-view method under a wide range of SNR conditions.
FedCPD:Personalized Federated Learning with Prototype-Enhanced Representation and Memory Distillation
PDF ↗Federated learning, as a distributed learning framework, aims to develop a global model while preserving client privacy. However, heterogeneity of client data leads to fairness issues and reduced performance. Techniques like parameter decoupling and prototype learning appear promising, yet challenges such as forgetting historical data and limited generalization persist. These methods also lack local insights, with locally trained features prone to overfitting, which affects generalization in global parameter aggregation. To address these challenges, we propose FedCPD, a personalized federated learning framework. FedCPD maintains historical information, reduces information loss, and increases personalization through hierarchical feature distillation and cross-layer feature fusion. Moreover, we utilize representation techniques like prototype contrastive learning and prototype alignment to capture diverse client data features, thus improving model generalization and fairness. Experiments show FedCPD outperforms state-of-the-art models, enhancing generalization by up to 10.40% and personalization by up to 4.90%, highlighting its effectiveness and superiority.
Due to the notorious modality imbalance phenomenon, multimodal learning (MML) struggles to achieve satisfactory performance. Recently, multimodal learning with alternating unimodal adaptation (MLA) has been proven effective in mitigating the interference between modalities by capturing interaction through orthogonal projection, thus relieving modality imbalance phenomenon to some extent. However, the projection strategy orthogonal to the original space can lead to poor plasticity as the alternating learning proceeds, thus affecting model performance. To address this issue, in this paper, we propose a novel multimodal learning method called interactiveMML via flat gradient modification (IGM) by employing a flat gradient modification strategy to enhance interactive MML. Specifically, we first employ a flat projection-based gradient modification strategy that is independent to the original space, aiming to avoid the poor plasticity issue. Then we introduce the sharpness-aware minimization (SAM)-based optimization strategy to fully exploit the flatness of the learning objective and further enhance interaction during learning. To this end, the plasticity problem can be avoided and the overall performance is improved. Extensive experiments on widely used datasets demonstrate that IGM outperforms various state-of-the-art (SOTA) baselines, achieving superior performance. The source code is available at https://anonymous.4open.science/r/method-CC45.
Classical Federated Learning (FL) encounters significant challenges when deploying large models on power-constrained clients. To tackle this, we propose an asymmetric FL mechanism that enables the aggregation of compact client models into a comprehensive server model. We design the server model as a Mixture-of-Experts (MoE), where each expert has the same architecture as each client model. This uniformity allows for efficient fusion of the most pertinent client models to update each server expert, based on the measured relevance between each client and server expert. To address the Non-IID data issue, we further optimize the server-side MoE architecture by incorporating a main expert that always activates alongside a set of selectively activated routed experts. This configuration ensures a balance between learning general knowledge and specific data distribution. Our Fed-MoE framework is model-agnostic and has demonstrated notable improvements on vision FL tasks with million-scale ResNet backbones, and language tasks with billion-scale BERT and GPT-2 backbones.
Learning to Explain: Towards Human-Aligned Explainability in Deep Reinforcement Learning via Attention Guidance
PDF ↗Recent advances in explainable deep reinforcement learning (DRL) have provided insights into the reasoning behind decisions made by DRL agents. However, existing methods often overlook the subjective nature of explanations and fail to consider human cognitive styles and preferences. Such ignorance tends to reduce the interpretability and relevance of the generated explanations from a human evaluator's perspective. To address this issue, we introduce human cognition into the explaining procedure by integrating DRL with attention guidance in a novel manner. The proposed concept proximal policy optimization (Concept-PPO) learns to generate human-aligned explanations by jointly optimizing the DRL performance and the discrepancy between generated explanations and human annotations. Its key component is a specially designed spatial concept transformer that can enhance explaining efficiency by premasking decision-irrelevant information. Experiments on the ATARI benchmark demonstrate that Concept-PPO achieves better policies than its black-box counterparts, and user studies confirm its superiority in generating human-aligned explanations compared to existing explainable DRL methods.
Explainable Artificial Intelligence (XAI) is critical for attaining trust in the operation of AI systems. A key question of an AI system is ``why was this decision made this way''. Formal approaches to XAI use a formal model of the AI system to identify abductive explanations. While abductive explanations may be applicable to a large number of inputs sharing the same concrete values, more general explanations may be preferred for numeric inputs. So-called inflated abductive explanations give intervals for each feature ensuring that any input whose values fall withing these intervals is still guaranteed to make the same prediction. Inflated explanations cover a larger portion of the input space, and hence are deemed more general explanations. But there can be many (inflated) abductive explanations for an instance. Which is the best? In this paper, we show how to find a most general abductive explanation for an AI decision. This explanation covers as much of the input space as possible, while still being a correct formal explanation of the model's behaviour. Given that we only want to give a human one explanation for a decision, the most general explanation gives us the explanation with the broadest applicability, and hence the one most likely to seem sensible.
Explainability and uncertainty quantification are key to trustable artificial intelligence. However, the reasoning behind uncertainty estimates is generally left unexplained. Identifying the drivers of uncertainty complements explanations of point predictions in recognizing model limitations and enhancing transparent decision-making. So far, explanations of uncertainties have been rarely studied. The few exceptions rely on Bayesian neural networks or technically intricate approaches, such as auxiliary generative models, thereby hindering their broad adoption. We propose a straightforward approach to explain predictive aleatoric uncertainties. We estimate uncertainty in regression as predictive variance by adapting a neural network with a Gaussian output distribution. Subsequently, we apply out-of-the-box explainers to the model's variance output. This approach can explain uncertainty influences more reliably than complex published approaches, which we demonstrate in a synthetic setting with a known data-generating process. We substantiate our findings with a nuanced, quantitative benchmark including synthetic and real, tabular and image datasets. For this, we adapt metrics from conventional XAI research to uncertainty explanations. Overall, the proposed method explains uncertainty estimates with little modifications to the model architecture and outperforms more intricate methods in most settings.
Prompt learning is an effective way to exploit the potential of large-scale pre-trained foundational models. Continuous prompts parameterize context tokens in prompts by turning them into differentiable vectors. Deep continuous prompts insert prompts not only in the input but also in the intermediate hidden representations. Manually designed deep continuous prompts exhibit a remarkable improvement compared to the zero-shot pre-trained model on downstream tasks. How to automate the continuous prompt design is an underexplored area, and a fundamental question arises, is manually designed deep prompt strategy optimal? To answer this question, we propose a method dubbed differentiable prompt learning (DPL). The DPL method is formulated as an optimization problem to automatically determine the optimal context length of the prompt to be added to each layer, where the objective is to maximize the performance. We test the DPL method on the pre-trained CLIP. We empirically find that by using only limited data, our DPL method can find deep continuous prompt configuration with high confidence. The performance on the downstream tasks exhibits the superiority of the automatic design: our method boosts the average test accuracy by 2.60% on 11 datasets compared to baseline methods. Besides, our method focuses only on the prompt configuration (i.e. context length for each layer), which means that our method is compatible with the baseline methods that have sophisticated designs to boost the performance. We release our code in https://github.com/Zhenhan-Huang/Differentiable-Prompt-Learn.
Visual Perturbation and Adaptive Hard Negative Contrastive Learning for Compositional Reasoning in Vision-Language Models
PDF ↗Vision-Language Models (VLMs) are essential for multimodal tasks, especially compositional reasoning (CR) tasks, which require distinguishing fine-grained semantic differences between visual and textual embeddings. However, existing methods primarily fine-tune the model by generating text-based hard negative samples, neglecting the importance of image-based negative samples, which results in insufficient training of the visual encoder and ultimately impacts the overall performance of the model. Moreover, negative samples are typically treated uniformly, without considering their difficulty levels, and the alignment of positive samples is insufficient, which leads to challenges in aligning difficult sample pairs. To address these issues, we propose Adaptive Hard Negative Perturbation Learning (AHNPL). AHNPL translates text-based hard negatives into the visual domain to generate semantically disturbed image-based negatives for training the model, thereby enhancing its overall performance. AHNPL also introduces a contrastive learning approach using a multimodal hard negative loss to improve the model's discrimination of hard negatives within each modality and a dynamic margin loss that adjusts the contrastive margin according to sample difficulty to enhance the distinction of challenging sample pairs. Experiments on three public datasets demonstrate that our method effectively boosts VLMs' performance on complex CR tasks. The source code is available at https://github.com/nynu-BDAI/AHNPL.
Multi‐view learning has emerged as a pivotal research area driven by the growing heterogeneity of real‐world data, and graph neural network-based models, modeling multi-view data as multi-view graphs, have achieved remarkable performance by revealing its deep semantics. However, by assuming cross‐view consistency, most approaches collect not only task-relevant (determinative) semantics but also symbiotic yet task-irrelevant (incidental) factors are collected to obscure model inference. Furthermore, these approaches often lack rigorous theoretical analysis that bridges training data to test data. To address these issues, we propose Target-oriented Graph Neural Network (TGNN), a novel framework that goes beyond traditional consistency by prioritizing task-relevant information, ensuring alignment with the target. Specifically, TGNN employs a class-level dual-objective loss to minimize the classification similarity between determinative and incidental factors, accentuating the former while suppressing the latter during model inference. Meanwhile, to ensure consistency between the learned semantics and predictions in representation learning, we introduce a penalty term that aims to amplify the divergence between these two types of factors. Furthermore, we derive an upper bound on the loss discrepancy between training and test data, providing formal guarantees for generalization to test domains. Extensive experiments conducted on three types of multi-view datasets validate the superiority of TGNN.
Deep Operator Network (DeepONet) effectively learns complex operator mappings, especially for systems governed by differential equations. Physics-informed DeepONet (PI-DeepONet) extends these capabilities by integrating physical constraints, enabling robust performance with limited or no labeled data. However, combining operator learning with these constraints increases computational complexity, which makes training more difficult and convergence slower, particularly for nonlinear or high-dimensional problems. In this work, we present an enhanced PI-DeepONet framework, that applies importance sampling to both of DeepONet inputs (i.e., the functions and the collocation points) to alleviate these training challenges. By focusing on critical data regions in both input domains, our approach showcases accelerated convergence and improved accuracy across various complex applications.
FedBG: Proactively Mitigating Bias in Cross-Domain Graph Federated Learning Using Background Data
PDF ↗Federated graph learning is focused on aggregating knowledge from multi-source graph data and training graph neural networks. Unlike the data that traditional federated learning needs to deal with, federated graph learning also needs to face additional topological information. Further, there are also biases in features and topologies among clients, increasing the difficulty of training models. Previous methods usually seek global calibration information, however, this approach may suffer from information bias caused by data skews, and it is also difficult to naturally combine feature and topology information. Therefore, adjusting the bias before it occurs will hopefully address the learning difficulties caused by the skew. In view of this, we employ background graph data, which works as reference information for local training, to proactively correct bias before it occurs. As a kind of graph data, background graphs are naturally capable of combining feature and topology information to accomplish bias correction among clients in a comprehensive way. Mixing strategy is employed on the background graph to additionally provide privacy-preserving capabilities. Graph generation methods are employed to restore the diversity of background graphs that are blurred by the mixing strategy. Extensive experiments on two real-world datasets demonstrate the sufficient motivation and effectiveness of the proposed method.
The combination of Spiking Neural Networks (SNNs) with Vision Transformer architectures has attracted significant attention due to the great potential for energy-efficient and high-performance computing paradigms. However, a substantial performance gap still exists between SNN-based and ANN-based transformer architectures. While existing methods propose spiking self-attention mechanisms that are successfully combined with SNNs, the overall architectures proposed by these methods suffer from a bottleneck in effectively extracting features from different image scales. In this paper, we address this issue and propose MSVIT, a novel spike-driven Transformer architecture, which firstly uses multi-scale spiking attention (MSSA) to enrich the capability of spiking attention blocks. We validate our approach across various main data sets. The experimental results indicate that our MSVIT outperforms existing SNN-based models, positioning itself as a state-of-the-art solution among NN-transformer architectures. The codes are available at https://github.com/Nanhu-AI-Lab/MSViT.