Locating boundary is very important for Temporal Action Detection (TAD) and is a key factor affecting the performance of TAD. However, two factors lead to inaccurate boundary localization: the movement feature submergence and the existence of ...

• Devised Movement Feature Extractor enhances movement feature in short-term temporal.
• Proposed Multi-Relation Module highlights the movement feature in long-term temporal.
• Proposed Scale Feature Pyramid Network learns specific ...

Existing knowledge distillation methods directly transfer knowledge from different intermediate layers of the teacher model without differentiating their correctness. However, the student model can obtain explicit and concise decision boundaries ...

• MLKD uses decision boundaries for multi-source feature fusion.
• MLKD is employed to facilitate the student model’s self-reflective distillation.
• MLKD achieves state-of-the-art results in five benchmark datasets.

Transformer-based deep learning methods have significantly facilitated multivariate time series classification (MTSC) tasks. However, due to the inherent operation of self-attention mechanism, most existing methods tend to overlook the internal ...

• A Gaussian-prior Transformer encoder is raised to better capture latent dependencies.
• A data-driven mask strategy is designed to determine crucial timestamps.
• We leverage context-aware positional encoding to improve model’s ...

Precise detection and effective intervention of pathogenic microorganisms are pivotal to ensure crop yield. Traditional open set recognition methods distinguish unknown classes, i.e., those not involved in training, as outliers or separate ...

• SMAD is the first dataset in the field of agricultural microorganisms.
• We propose an OSR framework based on feature representation and feature retrieval.
• The proposed framework can categorize known and unknown pathogenic fungal ...

Although the current two-stage prototypical network approach based on pretrained language models has achieved success in few-shot named entity recognition (NER) tasks, the span boundary at the entity span detection stage is still difficult to ...

Invariance against rotations of 3D objects is an important property in analyzing 3D point set data. Conventional 3D point set DNNs having rotation invariance typically obtain accurate 3D shape features via supervised learning by using labeled 3D ...

• Novel self-supervised learning (SSL) for rotation-invariant 3D point set analysis.
• Our method outperforms competitors both in accuracy and efficiency.
• Demonstrating incompatibility between existing rotation-invariant DNNs and SSL.

Deep learning models have been shown to be vulnerable to critical attacks under adversarial conditions. Attackers are able to generate powerful adversarial examples by searching for adversarial perturbations, without interfering with model ...

Self-supervised and Semi-supervised learning (SSL) on tabular data is an understudied topic. Despite some attempts, there are two major challenges: 1. Imbalanced nature in the tabular dataset; 2. The one-hot encoding used in these methods becomes ...

• The proposed FairCare framework uses an ethnicity-heterogeneous graph neural network to ensure demographic parity in clinical prognosis predictions through reduced bias in electronic health records (EHRs).
• Adversarial training was ...

Electronic health record (EHR) datasets have increasingly been harnessed by artificial intelligence (AI) for predictive modeling, yet the ethnicity fairness of these models remains underexplored. To address this issue, we propose FairCare, a ...

Person search has become one of the most critical and challenging applications in today's video surveillance systems. It helps in locating a person in surveillance videos, which is plausible only with advanced deep learning models, large scale ...

• Examines deep learning methods applied to person search tasks.
• Explores feature representation, loss functions, datasets, and metrics.
• Evaluates image-based person search and person re-identification techniques.
• Summarizes the ...

Knowledge distillation (KD) aims to build a lightweight deep neural network model under the guidance of a large-scale teacher model for model simplicity. Despite improved model efficiency through the KD technique, the performance gap between a ...

The out-of-distribution (OOD) detection task assumes samples that follow the distribution of training data as in-distribution (ID), while samples from other data distributions are considered OOD. In recent years, the OOD detection tasks have made ...

• We decompose the “unclear” definition of the ID into texture and semantics.
• Motivated by real-world problems, we create a new OOD detection benchmark.
• No auxiliary information is needed in our method, unlike previous models.
• ...

Traditional knowledge distillation approaches are typically designed for specific tasks, as they primarily distilling deep features from intermediate layers of a neural network, generally with ingeniously designed knowledge representations, which ...

Autoencoders (AEs) are a type of unsupervised neural networks, which can be used to solve various tasks, e.g., dimensionality reduction, image compression, and image denoising. An AE has two goals: (i) compress the original input to a low-...

• We propose a scheme to avoid redundant features in the bottleneck representation of autoencoders.
• We explicitly penalize the pair-wise correlations between the features and learn a diverse compressed embedding.
• The proposed penalty ...

In recent years, various self-knowledge distillation approaches have been proposed to reduce the cost of training teacher networks. However, these methods often overlook the significance of deep features. To address this limitation and strengthen ...

• The method recognizes the variation contribution of different feature maps in self-distillation.
• A new hierarchical refinement distillation technique via pyramid architecture.
• The approach improves the deep feature representative ...

Metric-based methods aim to predict class labels by computing the similarity between samples using distance functions, which is the mainstream approach to few-shot learning. However, the limited representational space of feature vectors and ...

• An unsupervised way activates discriminative semantic details of feature vectors.
• Mimicking human cognition to learn comprehensive semantic information of the object.
• Redefining the sampling strategy for Triplet loss to obtain high ...

• We propose a more robust feature representation method named HLDR to enhance the ability of uniting intra-class features and separating inter-class features ...

With the development of deep learning technique, many hyperspectral image classification (HIC) methods achieve great progress based on the application of convolution neural network (CNN). However, most of them still face small data ...

The feature representation of wind power sequences is crucial in the modeling of short-tern wind power forecast, but the existing feature representation methods mostly depend on the end-to-end model based on supervised learning, ignoring the ...

With the development of Internet, cybersecurity attracts people's attention. In order to better protect cybersecurity, we can comprehensively analyze the security events based on cyber threat intelligence. We aim to identify correlations between security ...

• A type of Lie Group feature is applied to encode the color, gradient, and shape features and the association information between different features into a ...

Medical image classification is a basic step in medical image analysis and has been an essential task in computer-aided diagnosis. Existing classification methods are proved to be effective in conventional image classification tasks, ...