FASP: Fast and Accurate Structured Pruning of Large Language Models

The rapid increase in the size of large language models (LLMs) has significantly escalated their computational and memory demands, posing challenges for efficient deployment, especially on resource-constrained devices. Structured pruning has emerged as an effective model compression method that can reduce these demands while preserving performance. In this paper, we introduce FASP (Fast and Accurate Structured Pruning), a novel structured pruning framework for LLMs that emphasizes both speed and accuracy. FASP employs a distinctive pruning structure that interlinks sequential layers, allowing for the removal of columns in one layer while simultaneously eliminating corresponding rows in the preceding layer without incurring additional performance loss. The pruning metric, inspired by Wanda, is computationally efficient and effectively selects components to prune. Additionally, we propose a restoration mechanism that enhances model fidelity by adjusting the remaining weights post-pruning. We evaluate FASP on the OPT and LLaMA model families, demonstrating superior performance in terms of perplexity and accuracy on downstream tasks compared to state-of-the-art methods. Our approach achieves significant speed-ups, pruning models such as OPT-125M in 17 seconds and LLaMA-30B in 15 minutes on a single NVIDIA RTX 4090 GPU, making it a highly practical solution for optimizing LLMs.

Intelligent Reflecting Surfaces Aided Wireless Network: Deployment Architectures and Solutions

Intelligent reflecting surfaces (IRSs) have emerged as a transformative technology for wireless networks by improving coverage, capacity, and energy efficiency through intelligent manipulation of wireless propagation environments. This paper provides a comprehensive study on the deployment and coordination of IRSs for wireless networks. By addressing both single- and multi-reflection IRS architectures, we examine their deployment strategies across diverse scenarios, including point-to-point, point-to-multipoint, and point-to-area setups. For the single-reflection case, we highlight the trade-offs between passive and active IRS architectures in terms of beamforming gain, coverage extension, and spatial multiplexing. For the multi-reflection case, we discuss practical strategies to optimize IRS deployment and element allocation, balancing cooperative beamforming gains and path loss. The paper further discusses practical challenges in IRS implementation, including environmental conditions, system compatibility, and hardware limitations. Numerical results and field tests validate the effectiveness of IRS-aided wireless networks and demonstrate their capacity and coverage improvements. Lastly, promising research directions, including movable IRSs, near-field deployments, and network-level optimization, are outlined to guide future investigations.

Learning from Ambiguous Data with Hard Labels

Real-world data often contains intrinsic ambiguity that the common single-hard-label annotation paradigm ignores. Standard training using ambiguous data with these hard labels may produce overly confident models and thus leading to poor generalization. In this paper, we propose a novel framework called Quantized Label Learning (QLL) to alleviate this issue. First, we formulate QLL as learning from (very) ambiguous data with hard labels: ideally, each ambiguous instance should be associated with a ground-truth soft-label distribution describing its corresponding probabilistic weight in each class, however, this is usually not accessible; in practice, we can only observe a quantized label, i.e., a hard label sampled (quantized) from the corresponding ground-truth soft-label distribution, of each instance, which can be seen as a biased approximation of the ground-truth soft-label. Second, we propose a Class-wise Positive-Unlabeled (CPU) risk estimator that allows us to train accurate classifiers from only ambiguous data with quantized labels. Third, to simulate ambiguous datasets with quantized labels in the real world, we design a mixing-based ambiguous data generation procedure for empirical evaluation. Experiments demonstrate that our CPU method can significantly improve model generalization performance and outperform the baselines.

Building Gradient Bridges: Label Leakage from Restricted Gradient Sharing in Federated Learning

The growing concern over data privacy, the benefits of utilizing data from diverse sources for model training, and the proliferation of networked devices with enhanced computational capabilities have all contributed to the rise of federated learning (FL). The clients in FL collaborate to train a global model by uploading gradients computed on their private datasets without collecting raw data. However, a new attack surface has emerged from gradient sharing, where adversaries can restore the label distribution of a victim's private data by analyzing the obtained gradients. To mitigate this privacy leakage, existing lightweight defenses restrict the sharing of gradients, such as encrypting the final-layer gradients or locally updating the parameters within. In this paper, we introduce a novel attack called Gradient Bridge (GDBR) that recovers the label distribution of training data from the limited gradient information shared in FL. GDBR explores the relationship between the layer-wise gradients, tracks the flow of gradients, and analytically derives the batch training labels. Extensive experiments show that GDBR can accurately recover more than 80% of labels in various FL settings. GDBR highlights the inadequacy of restricted gradient sharing-based defenses and calls for the design of effective defense schemes in FL.

MONOPOLY: Learning to Price Public Facilities for Revaluing Private Properties with Large-Scale Urban Data

The value assessment of private properties is an attractive but challenging task which is widely concerned by a majority of people around the world. A prolonged topic among us is ``\textit{how much is my house worth?}''. To answer this question, most experienced agencies would like to price a property given the factors of its attributes as well as the demographics and the public facilities around it. However, no one knows the exact prices of these factors, especially the values of public facilities which may help assess private properties. In this paper, we introduce our newly launched project ``Monopoly'' (named after a classic board game) in which we propose a distributed approach for revaluing private properties by learning to price public facilities (such as hospitals etc.) with the large-scale urban data we have accumulated via Baidu Maps. To be specific, our method organizes many points of interest (POIs) into an undirected weighted graph and formulates multiple factors including the virtual prices of surrounding public facilities as adaptive variables to parallelly estimate the housing prices we know. Then the prices of both public facilities and private properties can be iteratively updated according to the loss of prediction until convergence. We have conducted extensive experiments with the large-scale urban data of several metropolises in China. Results show that our approach outperforms several mainstream methods with significant margins. Further insights from more in-depth discussions demonstrate that the ``Monopoly'' is an innovative application in the interdisciplinary field of business intelligence and urban computing, and it will be beneficial to tens of millions of our users for investments and to the governments for urban planning as well as taxation.

Effects of Muscle Synergy during Overhead Work with a Passive Shoulder Exoskeleton: A Case Study

Objective: Shoulder exoskeletons can effectively assist with overhead work. However, their impacts on muscle synergy remain unclear. The objective is to systematically investigate the effects of the shoulder exoskeleton on muscle synergies during overhead work.Methods: Eight male participants were recruited to perform a screwing task both with (Intervention) and without (Normal) the exoskeleton. Eight muscles were monitored and muscle synergies were extracted using non-negative matrix factorization and electromyographic topographic maps. Results: The number of synergies extracted was the same (n = 2) in both conditions. Specifically, the first synergies in both conditions were identical, with the highest weight of AD and MD; while the second synergies were different between conditions, with highest weight of PM and MD, respectively. As for the first synergy in the Intervention condition, the activation profile significantly decreased, and the average recruitment level and activation duration were significantly lower (p<0.05). The regression analysis for the muscle synergies across conditions shows the changes of muscle synergies did not influence the sparseness of muscle synergies (p=0.7341). In the topographic maps, the mean value exhibited a significant decrease (p<0.001) and the entropy significantly increased (p<0.01). Conclusion: The exoskeleton does not alter the number of synergies and existing major synergies but may induce new synergies. It can also significantly decrease neural activation and may influence the heterogeneity of the distribution of monitored muscle activations. Significance: This study provides insights into the potential mechanisms of exoskeleton-assisted overhead work and guidance on improving the performance of exoskeletons.

Pseudo-labeling with Keyword Refining for Few-Supervised Video Captioning

Video captioning generate a sentence that describes the video content. Existing methods always require a number of captions (\eg, 10 or 20) per video to train the model, which is quite costly. In this work, we explore the possibility of using only one or very few ground-truth sentences, and introduce a new task named few-supervised video captioning. Specifically, we propose a few-supervised video captioning framework that consists of lexically constrained pseudo-labeling module and keyword-refined captioning module. Unlike the random sampling in natural language processing that may cause invalid modifications (\ie, edit words), the former module guides the model to edit words using some actions (\eg, copy, replace, insert, and delete) by a pretrained token-level classifier, and then fine-tunes candidate sentences by a pretrained language model. Meanwhile, the former employs the repetition penalized sampling to encourage the model to yield concise pseudo-labeled sentences with less repetition, and selects the most relevant sentences upon a pretrained video-text model. Moreover, to keep semantic consistency between pseudo-labeled sentences and video content, we develop the transformer-based keyword refiner with the video-keyword gated fusion strategy to emphasize more on relevant words. Extensive experiments on several benchmarks demonstrate the advantages of the proposed approach in both few-supervised and fully-supervised scenarios. The code implementation is available at https://github.com/mlvccn/PKG_VidCap

Gradient Descent Finds Over-Parameterized Neural Networks with Sharp Generalization for Nonparametric Regression: A Distribution-Free Analysis

We study nonparametric regression by an over-parameterized two-layer neural network trained by gradient descent (GD) in this paper. We show that, if the neural network is trained by GD with early stopping, then the trained network renders a sharp rate of the nonparametric regression risk of $\cO(\eps_n^2)$, which is the same rate as that for the classical kernel regression trained by GD with early stopping, where $\eps_n$ is the critical population rate of the Neural Tangent Kernel (NTK) associated with the network and $n$ is the size of the training data. It is remarked that our result does not require distributional assumptions on the training data, in a strong contrast with many existing results which rely on specific distributions such as the spherical uniform data distribution or distributions satisfying certain restrictive conditions. The rate $\cO(\eps_n^2)$ is known to be minimax optimal for specific cases, such as the case that the NTK has a polynomial eigenvalue decay rate which happens under certain distributional assumptions. Our result formally fills the gap between training a classical kernel regression model and training an over-parameterized but finite-width neural network by GD for nonparametric regression without distributional assumptions. We also provide confirmative answers to certain open questions or address particular concerns in the literature of training over-parameterized neural networks by GD with early stopping for nonparametric regression, including the characterization of the stopping time, the lower bound for the network width, and the constant learning rate used in GD.

Beware of Calibration Data for Pruning Large Language Models

As large language models (LLMs) are widely applied across various fields, model compression has become increasingly crucial for reducing costs and improving inference efficiency. Post-training pruning is a promising method that does not require resource-intensive iterative training and only needs a small amount of calibration data to assess the importance of parameters. Previous research has primarily focused on designing advanced pruning methods, while different calibration data's impact on pruning performance still lacks systematical exploration. We fill this blank and surprisingly observe that the effects of calibration data even value more than designing advanced pruning strategies, especially for high sparsity. Our preliminary exploration also discloses that using calibration data similar to the training data can yield better performance. As pre-training data is usually inaccessible for advanced LLMs, we further provide a self-generating calibration data synthesis strategy to construct feasible calibration data. We conduct experiments on the recent strong open-source LLMs (e.g., DCLM, and LLaMA-3), and the results show that the proposed method outperforms commonly used calibration data and can effectively enhance strong pruning methods (e.g., Wanda, OWL).

Toward Generalizing Visual Brain Decoding to Unseen Subjects

Visual brain decoding aims to decode visual information from human brain activities. Despite the great progress, one critical limitation of current brain decoding research lies in the lack of generalization capability to unseen subjects. Prior works typically focus on decoding brain activity of individuals based on the observation that different subjects exhibit different brain activities, while it remains unclear whether brain decoding can be generalized to unseen subjects. This study aims to answer this question. We first consolidate an image-fMRI dataset consisting of stimulus-image and fMRI-response pairs, involving 177 subjects in the movie-viewing task of the Human Connectome Project (HCP). This dataset allows us to investigate the brain decoding performance with the increase of participants. We then present a learning paradigm that applies uniform processing across all subjects, instead of employing different network heads or tokenizers for individuals as in previous methods, which can accommodate a large number of subjects to explore the generalization capability across different subjects. A series of experiments are conducted and we have the following findings. First, the network exhibits clear generalization capabilities with the increase of training subjects. Second, the generalization capability is common to popular network architectures (MLP, CNN and Transformer). Third, the generalization performance is affected by the similarity between subjects. Our findings reveal the inherent similarities in brain activities across individuals. With the emerging of larger and more comprehensive datasets, it is possible to train a brain decoding foundation model in the future. Codes and models can be found at https://github.com/Xiangtaokong/TGBD.