Multi-objective multi-armed bandit (MOMAB) problems are crucial for complex decision-making scenarios where multiple conflicting objectives must be simultaneously optimized. However, most existing works are based on the linear assumption of the feedback rewards, which significantly constrains their applicability and efficacy in capturing the intricate dynamics of real-world environments. This paper explores a multi-objective neural bandit (MONB) framework, which integrates the universal approximators, neural networks, with the classical MOMABs. We adopt random scalarization to accommodate the special needs of a practitioner by setting an appropriate distribution on the regions of interest. Using the trade-off capabilities of upper confidence bound (UCB) and Thompson sampling (TS) strategies, we propose two novel algorithms, MONeural-UCB and MONeural-TS. Theoretical and empirical analysis demonstrate the superiority of our methods in multi-objective or multi-task bandit problems, which makes great improvement over the classical linear MOMABs.
论文检索
输入标题、作者或关键词，从 2,893 篇学术成果中精准定位
Real-life deployment of federated Learning (FL) often faces non-IID data, which leads to poor accuracy and slow convergence. Personalized FL (pFL) tackles these issues by tailoring local models to individual data sources and using weighted aggregation methods for client-specific learning. However, existing pFL methods often fail to provide each local model with global knowledge on demand while maintaining low computational overhead. Additionally, local models tend to over-personalize their data during the training process, potentially dropping previously acquired global information. We propose FLAYER, a novel layer-wise learning method for pFL that optimizes local model personalization performance. FLAYER considers the different roles and learning abilities of neural network layers of individual local models. It incorporates global information for each local model as needed to initialize the local model cost-effectively. It then dynamically adjusts learning rates for each layer during local training, optimizing the personalized learning process for each local model while preserving global knowledge. Additionally, to enhance global representation in pFL, FLAYER selectively uploads parameters for global aggregation in a layer-wise manner. We evaluate FLAYER on four representative datasets in computer vision and natural language processing domains. Compared to eight state-of-the-art pFL methods, FLAYER improves the inference accuracy, on average, by 5.20% (up to 14.29%). Code is available at https://github.com/lancasterJie/FLAYER/.
Adaptive Gradient Learning for Spiking Neural Networks by Exploiting Membrane Potential Dynamics
PDF ↗Recent advancements have focused on directly training high-performance spiking neural networks (SNNs) by estimating the approximate gradients of spiking activity through a continuous function with constant sharpness, known as surrogate gradient (SG) learning. However, as spikes propagate within neurons and among layers, the distribution of membrane potential dynamics (MPD) will deviate from the gradient-available interval of fixed SG, hindering SNNs from searching the optimal solution space. To maintain the stability of gradient flows, SG needs to align with evolving MPD. Here, we propose a novel adaptive gradient learning for SNNs by exploiting MPD, namely MPD-AGL. It fully accounts for the underlying factors contributing to membrane potential shifts and establishes a dynamic association between SG and MPD at different timesteps to relax gradient estimation, which provides a new degree of freedom for SG learning. Experimental results demonstrate that our method achieves excellent performance at low latency. Moreover, it increases the proportion of neurons that fall into the gradient-available interval compared to fixed SG, effectively mitigating the gradient vanishing problem. Code is available at https://github.com/jqjiang1999/MPD-AGL.
With the rapid development of recommendation models and device computing power, device-based recommendation has become an important research area due to its better real-time performance and privacy protection. Previously, Transformer-based sequential recommendation models have been widely applied in this field because they outperform Recurrent Neural Network (RNN)-based recommendation models in terms of performance. However, as the length of interaction sequences increases, Transformer-based models introduce significantly more space and computational overhead compared to RNN-based models, posing challenges for device-based recommendation. To balance real-time performance and high performance on devices, we propose Device-Cloud Collaborative Correction Framework for On-Device Recommendation (CoCorrRec). CoCorrRec uses a self-correction network (SCN) to correct parameters with extremely low time cost. By updating model parameters during testing based on the input token, it achieves performance comparable to current optimal but more complex Transformer-based models. Furthermore, to prevent SCN from overfitting, we design a global correction network (GCN) that processes hidden states uploaded from devices and provides a global correction solution. Extensive experiments on multiple datasets show that CoCorrRec outperforms existing Transformer-based and RNN-based device recommendation models in terms of performance, with fewer parameters and lower FLOPs, thereby achieving a balance between real-time performance and high efficiency. Code is available at https: //github.com/Yuzt-zju/CoCorrRec.
On-device sequential recommendation (SR) systems are designed to make local inferences using real-time features, thereby alleviating the communication burden on server-based recommenders when handling concurrent requests from millions of users. However, the resource constraints of edge devices, including limited memory and computational capacity, pose significant challenges to deploying efficient SR models. Inspired by the energy-efficient and sparse computing properties of deep Spiking Neural Networks (SNNs), we propose a cost-effective on-device SR model named SSR, which encodes dense embedding representations into sparse spike-wise representations and integrates novel spiking filter modules to extract temporal patterns and critical features from item sequences, optimizing computational and memory efficiency without sacrificing recommendation accuracy. Extensive experiments on real-world datasets demonstrate the superiority of SSR. Compared to other SR baselines, SSR achieves comparable recommendation performance while reducing energy consumption by an average of 59.43%. In addition, SSR significantly lowers memory usage, making it particularly well-suited for deployment on resource-constrained edge devices.
Graph Neural Network-based methods face privacy leakage risks due to the introduction of topological structures about the targets, which allows attackers to bypass the target's prior knowledge of the sensitive attributes and realize membership inference attacks (MIA) by observing and analyzing the topology distribution. As privacy concerns grow, the assumption of MIA, which presumes that attackers can obtain an auxiliary dataset with the same distribution, is increasingly deviating from reality. In this paper, we categorize the distribution diversity issue in real-world MIA scenarios as an Out-Of-Distribution (OOD) problem, and propose a novel Graph OOD Membership Inference Attack (GOOD-MIA) to achieve cross-domain graph attacks. Specifically, we construct shadow subgraphs with distributions from different domains to model the diversity of real-world data. We then explore the stable node representations that remain unchanged under external influences and consider eliminating redundant information from confounding environments and extracting task-relevant key information to more clearly distinguish between the characteristics of training data and unseen data. This OOD-based design makes cross-domain graph attacks possible. Finally, we perform risk extrapolation to optimize the attack's domain adaptability during attack inference to generalize the attack to other domains. Experimental results demonstrate that GOOD-MIA achieves superior attack performance in datasets designed for multiple domains.
SEP: A General Lossless Compression Framework with Semantics Enhancement and Multi-Stream Pipelines
PDF ↗Deep-learning-based lossless compression is of immense importance in real-world applications, such as cold data persistence, sensor data collection, and astronomical data transmission. However, existing compressors typically model data using single-byte symbols as tokens, which makes it hard to capture the inherent correlations and cannot effectively utilize the parallel capabilities of GPU and multi-core CPU. This paper proposes SEP, a novel lossless compression framework for most time-series backbone neural networks. We first introduce a semantic enhancement module to capture the complex intra-patch relationships of binary byte streams. To improve the compression speed, we design multi-stream pipelines that dynamically assign parallel tasks to GPU streams and multi-cores. We further propose a novel GPU memory optimization strategy, which reuses GPU memory by a shared pool across streams. We conduct experiments on seven real-world datasets and the results demonstrate that our SEP framework outperforms state-of-the-art compressors with an average speed improvement of 30.0% and an average compression ratio gain of 5.1%, which is further elevated to 7.6% with the use of pre-training models. The GPU memory footprint is reduced by as high as 63.1% and by an average of 36.2%. The source code is available at: https://github.com/damonwan1/SEP.
Detecting Out-of-Distribution (OOD) inputs is crucial for improving the reliability of deep neural networks in the real-world deployment. In this paper, inspired by the inherent distribution shift between in-distribution (ID) and OOD data, we propose a novel method that leverages optimal transport to measure the distribution discrepancy between test inputs and ID prototypes. The resulting transport costs are used to quantify the individual contribution of each test input to the overall discrepancy, serving as a desirable measure for OOD detection. To address the issue that solely relying on the transport costs to ID prototypes is inadequate for identifying OOD inputs closer to ID data, we generate virtual outliers to approximate the OOD region via linear extrapolation. By combining the transport costs to ID prototypes with the costs to virtual outliers, the detection of OOD data near ID data is emphasized, thereby enhancing the distinction between ID and OOD inputs. Extensive evaluations demonstrate the superiority of our method over state-of-the-art methods.
Memory-based temporal graph neural networks (MTGNN) use node memory to store historical information, enabling efficient processing of large dynamic graphs through batch parallel training, with larger batch sizes leading to increased training efficiency. However, this approach overlooks the interdependency among edges within the same batch, leading to outdated memory states and reduced training accuracy. Previous studies have attempted to mitigate this issue through methods such as measuring memory loss, overlap training, and additional compensation modules. Despite these efforts, challenges persist, including imprecise coarse-grained memory loss measurement and ineffective compensation modules. To address these challenges, we propose the Refined Batch parallel Training (RBT) framework, which accurately evaluates intra-batch information loss and optimizes batch partitioning to minimize loss, enhancing the training process's effectiveness and efficiency. RBT also includes a precise and efficient memory compensation algorithm. Experimental results demonstrate RBT's superior performance compared to existing MTGNN frameworks like TGL, ETC, and PRES in terms of training efficiency and accuracy across various dynamic graph datasets. Our code is made publicly available at https://github.com/fengwudi/RBT.
Multimodal hashing projects multimodal data into compact binary codes, enabling rapid and storage-efficient retrieval of large-scale multimedia content. In practical scenarios, the issue of missing modality frequently arises when dealing with multimodal data. Existing incomplete multimodal hashing techniques directly recover missing modalities by neural networks, resulting in a disjointed representation space between the recovered and true data. In this paper, we present a novel recovery paradigm, namely Prototype-based Modality Completion Hashing (PMCH). Instead of directly synthesizing it from available modalities, PMCH adaptively aggregates associated within-modality prototypes to recover missing modality data. Specifically, PMCH introduces an within-modality prototype learning module to optimize representative prototypes for each modality. These prototypes act as recovery anchors and reside within the same representation space as their corresponding modality data. Subsequently, PMCH adaptively aggregates the associated within-modality prototypes with coefficients derived from the modality-specific Weight-Net. By utilizing prototypes from the same modality, the semantic disparity between the reconstructed and authentic data can be substantially diminished. Extensive experiments on three widely used benchmark datasets demonstrate that PMCH can effectively recover the missing modality, and attain state-of-the-art performance in both complete and incomplete multimodal retrieval scenarios. Code is available at https://github.com/Sasa77777779/PMCH.git.
Deep Opinion-Unaware Blind Image Quality Assessment by Learning and Adapting from Multiple Annotators
PDF ↗Existing deep neural network (DNN)-based blind image quality assessment (BIQA) methods primarily rely on human-rated datasets for training. However, collecting human labels is extremely time-consuming and labor-intensive, posing a significant bottleneck for practical applications. To address this challenge, we propose a Deep opinion-Unaware BIQA model by learning and adapting from Multiple Annotators, termed DUBMA, thereby eliminating the need for human annotations. Specifically, we first generate a large-scale set of distorted image pairs and then assign relative quality rankings using existing full-reference IQA models. The resulting dataset is subsequently employed for training our DUBMA. Due to the inherent discrepancies between synthetic and real-world distortions, a domain shift may occur. To address this, we propose an outlier-robust unsupervised domain adaptation approach leveraging optimal transport. This strategy effectively reduces the gap between synthetic and real-world distortion domains, thereby boosting the model’s adaptability and overall performance. Extensive experiments show that DUBMA outperforms existing opinion-unaware BIQA methods in terms of prediction accuracy across multiple datasets.
Conventional image set methods typically learn from image sets stored in a single location. However, in real-world applications, image sets are often distributed across different locations. Learning from such distributed sets using deep neural networks poses challenges for efficient image set classification and retrieval. To address this, we propose Distributed Cascade Manifold Hashing Network (DCMHN) for compact image set representation. DCMHN represents each image set using an SPD manifold and utilizes a manifold hashing network to generate hash codes, enabling efficient classification and retrieval. The network is trained in a cascaded manner, where the bilinear mapping in the BiMap layer is learned first, followed by joint learning of the hash function and classifier in the hash layer. DCMHN enforces local consistency on global variables across neighboring nodes, allowing parallel optimization. Extensive experiments on three benchmark image set datasets demonstrate that the proposed DCMHN achieves competitive accuracies in distributed settings, and outperforms state-of-the-arts in terms of computation and storage efficiency.
Pansharpening combines panchromatic and low-resolution multispectral images to generate high-resolution multispectral images. Previous studies have explored the connection between pansharpening and the frequency domain, but mostly in the real-valued domain, leaving the complex domain relatively unexplored. To redefine the pansharpening task, we propose a complex-valued spatial-frequency dual-domain framework, PanComplex. To achieve this, we first establish complex representations and introduce basic complex operators tailored to pansharpening, enabling the transformation of multispectral real-valued signals into the complex domain for learning. We then model both spatial and frequency branches to capture global frequency features and local spatial features comprehensively. Finally, we employ a complex-based interaction module to fuse the spatial and frequency features, achieving complementary information across both domains. By using the representation power of the complex domain, PanComplex effectively extracts complementary features from PAN and MS images, thereby enhancing pansharpening performance. Experiments on multiple datasets demonstrate that our method achieves optimal performance with the fewest parameters and exhibits strong generalization ability to other tasks. The source code for this work is publicly available at https://github.com/lch-ustc/PanComplex.
Tackling Long-Tailed Data Challenges in Spiking Neural Networks via Heterogeneous Knowledge Distillation
PDF ↗Spiking Neural Networks (SNNs), inspired by the behavior of biological neurons, have gained significant research interest for resource-constrained edge devices and neuromorphic hardware due to their use of binary spike signals for inter-unit communication with low power consumption. However, the absence of research on spiking neural networks on long-tailed data has severely limited the deployment and application of this emerging network in practical scenarios. To fill this gap, this paper proposes a long-tail learning framework based on spiking neural networks, named LT-SpikingFormer, to alleviate the distribution bias between head and tail classes. LT-SpikingFormer adopts a widely trained Convolutional Neural Network to construct a heterogeneous knowledge distillation paradigm, offering balanced and reliable prior knowledge. Moreover, a multi-granularity hierarchical feature distillation objective is proposed to leverage cross-layer local features and network global predictions to facilitate refined information distillation to optimize the network, specifically for the performance of the tailed classes. Extensive experimental results demonstrate that our method performs well on several benchmark datasets.
In order to boost the performance of a convolutional neural network (CNN), several approaches have shown the benefit of enhancing the spatial encoding of feature maps. However, few works paid attention to the positional properties of convolutional kernels. In this paper, we demonstrate that different kernel positions are of different importance, which depends on the task, dataset and architecture, and adaptively emphasizing the informative parts in convolutional kernels can lead to considerable improvement. Therefore, we propose a novel structural re-parameterization Position Boosting Convolution (PBConv) to exploit and enhance the position information in the convolutional kernel. PBConv consists of several concurrent small convolutional kernels, which can be equivalently converted to the original kernel and bring no extra inference cost. Different from existing structural re-parameterization methods, PBconv searches for the optimal re-parameterized structure by a fast heuristic algorithm based on the dispersion of kernel weights. Such heuristic search is efficient yet effective, well adapting the varying kernel weight distribution. As a result, PBConv can significantly improve the representational power of a model, especially its ability to extract fine-grained low-level features. Importantly, PBConv is orthogonal to procedural re-parameterization methods and can further boost performance based on them.
Oriented object detection has advanced with the development of convolutional neural networks (CNNs) and transformers. However, modern detectors still rely on predefined object candidates, such as anchors in CNN-based methods or queries in transformer-based methods, which struggle to capture spatial information effectively. To address the limitations, we propose GSDet, a novel framework that formulates oriented object detection as Gaussian splatting. Specifically, our approach performs detection within a 3D feature space constructed from image features, where 3D Gaussians are employed to represent oriented objects. These 3D Gaussians are projected onto the image plane to form 2D Gaussians, which are then transformed into oriented boxes. Furthermore, we optimize the mean, anisotropic covariance, and confidence scores of these randomly initialized 3D Gaussians, using a decoder that incorporates 3D Gaussian sampling. Moreover, our method exhibits flexibility, enabling adaptive control and a dynamic number of Gaussians during inference. Experiments on 3 datasets indicate that GSDet achieves AP50 gains of 0.7% on DIOR-R, 0.3% on DOTA-v1.0, and 0.55% on DOTA-v1.5 when evaluated with adaptive control and outperforms mainstream detectors.
Enhancing Mixture of Experts with Independent and Collaborative Learning for Long-Tail Visual Recognition
PDF ↗Deep neural networks (DNNs) face substantial challenges in Long-Tail Visual Recognition (LTVR) due to the inherent class imbalances in real-world data distributions. The Mixture of Experts (MoE) framework has emerged as a promising approach to addressing these issues. However, in MoE systems, experts are typically trained to optimize a collective objective, often neglecting the individual optimality of each expert. This individual optimality usually contributes to the overall performance, as the goals of different experts are not mutually exclusive. We propose the Independent and Collaborative Learning (ICL) framework to optimize each expert independently while ensuring global optimality. First, Diverse Optimization Learning (DOL) is introduced to enhance expert diversity and individual performance. Then, we conceptualize experts as parallel circuit branches and introduce Competition and Collaboration Learning (CoL). Competition Learning amplifies the gradients of better-performing experts to preserve individual optimality, and Collaboration Learning encourages collaboration through mutual distillation to enhance optimal knowledge sharing. ICL achieves state-of-the-art accuracy in experiments on CIFAR-100/10-LT, ImageNet-LT, and iNaturalist 2018, respectively. Our code is available at https://github.com/PolarisLight/ICL.
Efficient Counterexample-Guided Fairness Verification and Repair of Neural Networks Using Satisfiability Modulo Convex Programming
PDF ↗Ensuring fairness is essential for ethical decision-making in various domains. Informally, a neural network is considered fair if and only if it treats similar individuals similarly in a given task. We introduce FaVeR (Fairness Verification and Repair), a framework for efficiently verifying and repairing pre-trained neural networks with respect to individual fairness properties. FaVeR ensures fairness via iterative search of high-sensitivity neurons and backward adjustment of their weights, guided by counterexamples generated from fairness verification using satisfiability modulo convex programming. By addressing fairness at the neuron level, FaVeR minimizes the impact of neural network repair on the overall performance. Experimental evaluations on common fairness datasets show that FaVeR achieves a 100% fairness repair rate across all models, with accuracy reduction of less than 2.27%. Moreover, its significantly lower average runtime makes it suitable for practical applications.
SuBiTO: Synopsis-based Training Optimization for Continuous Real-Time Neural Learning over Big Streaming Data
PDF ↗In machine learning applications over Big streaming Data, Neural Networks (NNs) are continuously and rapidly trained over voluminous data arriving at high speeds. As soon as a new version of the NN becomes available, it gets deployed for prediction purposes (e.g. classification). The real-time character of such applications greatly depends on the volume and velocity of the data streams, as well as the NN complexity. Training on large volume of ingested streams or using complex NNs, potentially increases accuracy, but may compromise the real-time character of those applications. In this work, we present SuBiTO, a framework that automatically and continuously learns the training time vs accuracy trade-offs as new data stream in and fine tunes: (i) the number, size and type of NN layers; (ii) the size of the ingested data via stream synopses specific parameters; and (iii) the number of training epochs. Finally, SuBiTO suggests optimal sets of such parameters and detects concept drifts, enabling the human operator adapt these parameters on-the-fly, at runtime.
Artificial neural networks (ANNs) struggle with continual learning, sacrificing performance on previously learned tasks to acquire new task knowledge. Here we propose a new approach allowing to mitigate catastrophic forgetting during continuous task learning. Typically a new task is trained until it reaches maximal performance, causing complete catastrophic forgetting of the previous tasks. In our new approach, termed Optimal Stopping (OS), network training on each new task continues only while the mean validation accuracy across all the tasks (current and previous) increases. The stopping criterion creates an explicit balance: lower performance on new tasks is accepted in exchange for preserving knowledge of previous tasks, resulting in higher overall network performance. The overall performance is further improved when OS is combined with Sleep Replay Consolidation (SRC), wherein the network converts to a Spiking Neural Network (SNN) and undergoes unsupervised learning modulated by Hebbian plasticity. During the SRC, the network spontaneously replays activation patterns from previous tasks, helping to maintain and restore prior task performance. This combined approach offers a promising avenue for enhancing the robustness and longevity of learned representations in continual learning models, achieving over twice the mean accuracy of baseline continuous learning while maintaining stable performance across tasks.