The success of self-supervised contrastive learning hinges on identifying positive data pairs that, when pushed together in embedding space, encode useful information for subsequent downstream tasks. However, in time-series, this is challenging because creating positive pairs via augmentations may break the original semantic meaning. We hypothesize that if we can retrieve information from one subsequence to successfully reconstruct another subsequence, then they should form a positive pair. Harnessing this intuition, we introduce our novel approach: REtrieval-BAsed Reconstruction (REBAR) contrastive learning. First, we utilize a convolutional cross-attention architecture to calculate the REBAR error between two different time-series. Then, through validation experiments, we show that the REBAR error is a predictor of mutual class membership, justifying its usage as a positive/negative labeler. Finally, once integrated into a contrastive learning framework, our REBAR method can learn an embedding that achieves state-of-the-art performance on downstream tasks across various modalities.
In this work, we investigate the performance of a joint sensing and communication (JSC) network consisting of multiple base stations (BSs) that cooperate through a fusion center (FC) to exchange information about the sensed environment while concurrently establishing communication links with a set of user equipments (UEs). Each BS within the network operates as a monostatic radar system, enabling comprehensive scanning of the monitored area and generating range-angle maps that provide information regarding the position of a group of heterogeneous objects. The acquired maps are subsequently fused in the FC. Then, a convolutional neural network (CNN) is employed to infer the category of the targets, e.g., pedestrians or vehicles, and such information is exploited by an adaptive clustering algorithm to group the detections originating from the same target more effectively. Finally, two multi-target tracking algorithms, the probability hypothesis density (PHD) filter and multi-Bernoulli mixture (MBM) filter, are applied to estimate the state of the targets. Numerical results demonstrated that our framework could provide remarkable sensing performance, achieving an optimal sub-pattern assignment (OSPA) less than 60 cm, while keeping communication services to UEs with a reduction of the communication capacity in the order of 10% to 20%. The impact of the number of BSs engaged in sensing is also examined, and we show that in the specific case study, 3 BSs ensure a localization error below 1 m.
In this paper, we propose an approach to address the problem of 3D reconstruction of scenes from a single image captured by a light-field camera equipped with a rolling shutter sensor. Our method leverages the 3D information cues present in the light-field and the motion information provided by the rolling shutter effect. We present a generic model for the imaging process of this sensor and a two-stage algorithm that minimizes the re-projection error while considering the position and motion of the camera in a motion-shape bundle adjustment estimation strategy. Thereby, we provide an instantaneous 3D shape-and-pose-and-velocity sensing paradigm. To the best of our knowledge, this is the first study to leverage this type of sensor for this purpose. We also present a new benchmark dataset composed of different light-fields showing rolling shutter effects, which can be used as a common base to improve the evaluation and tracking the progress in the field. We demonstrate the effectiveness and advantages of our approach through several experiments conducted for different scenes and types of motions. The source code and dataset are publicly available at: https://github.com/ICB-Vision-AI/RSLF
RGB-based surface anomaly detection methods have advanced significantly. However, certain surface anomalies remain practically invisible in RGB alone, necessitating the incorporation of 3D information. Existing approaches that employ point-cloud backbones suffer from suboptimal representations and reduced applicability due to slow processing. Re-training RGB backbones, designed for faster dense input processing, on industrial depth datasets is hindered by the limited availability of sufficiently large datasets. We make several contributions to address these challenges. (i) We propose a novel Depth-Aware Discrete Autoencoder (DADA) architecture, that enables learning a general discrete latent space that jointly models RGB and 3D data for 3D surface anomaly detection. (ii) We tackle the lack of diverse industrial depth datasets by introducing a simulation process for learning informative depth features in the depth encoder. (iii) We propose a new surface anomaly detection method 3DSR, which outperforms all existing state-of-the-art on the challenging MVTec3D anomaly detection benchmark, both in terms of accuracy and processing speed. The experimental results validate the effectiveness and efficiency of our approach, highlighting the potential of utilizing depth information for improved surface anomaly detection.
Learning generalized representations from limited training samples is crucial for applying deep neural networks in low-resource scenarios. Recently, methods based on Contrastive Language-Image Pre-training (CLIP) have exhibited promising performance in few-shot adaptation tasks. To avoid catastrophic forgetting and overfitting caused by few-shot fine-tuning, existing works usually freeze the parameters of CLIP pre-trained on large-scale datasets, overlooking the possibility that some parameters might not be suitable for downstream tasks. To this end, we revisit CLIP's visual encoder with a specific focus on its distinctive attention pooling layer, which performs a spatial weighted-sum of the dense feature maps. Given that dense feature maps contain meaningful semantic information, and different semantics hold varying importance for diverse downstream tasks (such as prioritizing semantics like ears and eyes in pet classification tasks rather than side mirrors), using the same weighted-sum operation for dense features across different few-shot tasks might not be appropriate. Hence, we propose fine-tuning the parameters of the attention pooling layer during the training process to encourage the model to focus on task-specific semantics. In the inference process, we perform residual blending between the features pooled by the fine-tuned and the original attention pooling layers to incorporate both the few-shot knowledge and the pre-trained CLIP's prior knowledge. We term this method as Semantic-Aware FinE-tuning (SAFE). SAFE is effective in enhancing the conventional few-shot CLIP and is compatible with the existing adapter approach (termed SAFE-A).
In a data-centric era, concerns regarding privacy and ethical data handling grow as machine learning relies more on personal information. This empirical study investigates the privacy, generalization, and stability of deep learning models in the presence of additive noise in federated learning frameworks. Our main objective is to provide strategies to measure the generalization, stability, and privacy-preserving capabilities of these models and further improve them. To this end, five noise infusion mechanisms at varying noise levels within centralized and federated learning settings are explored. As model complexity is a key component of the generalization and stability of deep learning models during training and evaluation, a comparative analysis of three Convolutional Neural Network (CNN) architectures is provided. The paper introduces Signal-to-Noise Ratio (SNR) as a quantitative measure of the trade-off between privacy and training accuracy of noise-infused models, aiming to find the noise level that yields optimal privacy and accuracy. Moreover, the Price of Stability and Price of Anarchy are defined in the context of privacy-preserving deep learning, contributing to the systematic investigation of the noise infusion strategies to enhance privacy without compromising performance. Our research sheds light on the delicate balance between these critical factors, fostering a deeper understanding of the implications of noise-based regularization in machine learning. By leveraging noise as a tool for regularization and privacy enhancement, we aim to contribute to the development of robust, privacy-aware algorithms, ensuring that AI-driven solutions prioritize both utility and privacy.
The transformer model has gained widespread adoption in computer vision tasks in recent times. However, due to the quadratic time and memory complexity of self-attention, which is proportional to the number of input tokens, most existing Vision Transformers (ViTs) encounter challenges in achieving efficient performance in practical industrial deployment scenarios, such as TensorRT and CoreML, where traditional CNNs excel. Although some recent attempts have been made to design CNN-Transformer hybrid architectures to tackle this problem, their overall performance has not met expectations. To tackle these challenges, we propose an efficient hybrid ViT architecture named FMViT. This approach enhances the model's expressive power by blending high-frequency features and low-frequency features with varying frequencies, enabling it to capture both local and global information effectively. Additionally, we introduce deploy-friendly mechanisms such as Convolutional Multigroup Reparameterization (gMLP), Lightweight Multi-head Self-Attention (RLMHSA), and Convolutional Fusion Block (CFB) to further improve the model's performance and reduce computational overhead. Our experiments demonstrate that FMViT surpasses existing CNNs, ViTs, and CNNTransformer hybrid architectures in terms of latency/accuracy trade-offs for various vision tasks. On the TensorRT platform, FMViT outperforms Resnet101 by 2.5% (83.3% vs. 80.8%) in top-1 accuracy on the ImageNet dataset while maintaining similar inference latency. Moreover, FMViT achieves comparable performance with EfficientNet-B5, but with a 43% improvement in inference speed. On CoreML, FMViT outperforms MobileOne by 2.6% in top-1 accuracy on the ImageNet dataset, with inference latency comparable to MobileOne (78.5% vs. 75.9%). Our code can be found at https://github.com/tany0699/FMViT.
The problem of data sparsity has long been a challenge in recommendation systems, and previous studies have attempted to address this issue by incorporating side information. However, this approach often introduces side effects such as noise, availability issues, and low data quality, which in turn hinder the accurate modeling of user preferences and adversely impact recommendation performance. In light of the recent advancements in large language models (LLMs), which possess extensive knowledge bases and strong reasoning capabilities, we propose a novel framework called LLMRec that enhances recommender systems by employing three simple yet effective LLM-based graph augmentation strategies. Our approach leverages the rich content available within online platforms (e.g., Netflix, MovieLens) to augment the interaction graph in three ways: (i) reinforcing user-item interaction egde, (ii) enhancing the understanding of item node attributes, and (iii) conducting user node profiling, intuitively from the natural language perspective. By employing these strategies, we address the challenges posed by sparse implicit feedback and low-quality side information in recommenders. Besides, to ensure the quality of the augmentation, we develop a denoised data robustification mechanism that includes techniques of noisy implicit feedback pruning and MAE-based feature enhancement that help refine the augmented data and improve its reliability. Furthermore, we provide theoretical analysis to support the effectiveness of LLMRec and clarify the benefits of our method in facilitating model optimization. Experimental results on benchmark datasets demonstrate the superiority of our LLM-based augmentation approach over state-of-the-art techniques. To ensure reproducibility, we have made our code and augmented data publicly available at: https://github.com/HKUDS/LLMRec.git
Clustering of time series is a well-studied problem, with applications ranging from quantitative, personalized models of metabolism obtained from metabolite concentrations to state discrimination in quantum information theory. We consider a variant, where given a set of trajectories and a number of parts, we jointly partition the set of trajectories and learn linear dynamical system (LDS) models for each part, so as to minimize the maximum error across all the models. We present globally convergent methods and EM heuristics, accompanied by promising computational results.
We analyze how symmetries can be used to compress structures (also known as interpretations) onto a smaller domain without loss of information. This analysis suggests the possibility to solve satisfiability problems in the compressed domain for better performance. Thus, we propose a 2-step novel method: (i) the sentence to be satisfied is automatically translated into an equisatisfiable sentence over a ``lifted'' vocabulary that allows domain compression; (ii) satisfiability of the lifted sentence is checked by growing the (initially unknown) compressed domain until a satisfying structure is found. The key issue is to ensure that this satisfying structure can always be expanded into an uncompressed structure that satisfies the original sentence to be satisfied. We present an adequate translation for sentences in typed first-order logic extended with aggregates. Our experimental evaluation shows large speedups for generative configuration problems. The method also has applications in the verification of software operating on complex data structures. Further refinements of the translation are left for future work.