With the popularity of cloud storage, video data, especially videos containing personal information, faces serious security challenges. However, existing video encryption schemes typically adopt full-frame encryption, resulting in high computational overhead, low efficiency, and difficulty balancing privacy protection and lightweight encryption requirements. To this end, a lightweight video encryption scheme based on object detection and adaptive frame multiplexing is designed in this paper. First, the YOLOv8 algorithm is employed to accurately identify multiple human target regions in videos, thereby avoiding the computational resource waste associated with full-frame encryption. Second, combined with SCI-HMC hyperchaotic map, an adaptive frame multiplexing strategy is adopted to realize the dynamic adjustment of chaotic sequences by correlating the encryption process before and after. On this basis, a cube lightweight encryption algorithm based on video frame combination is designed, which is spliced layer by layer in terms of color channels, traversed in terms of 3 layers of data (one video frame), and performs the 3D confusion and 3D mod diffusion sequentially according to the target coordinates, and finally realizes the accurate encryption of the region of interest. The experimental results show that the scheme performs well in terms of practicality and resistance to attacks. The information entropy of the encrypted area is as high as 7.9982, and the encryption speed can be increased to 0.2343 seconds per frame. This deep collaboration framework tightly integrates modern object detection with dynamic encryption processes, providing a balanced security and lightweight solution for cloud video data protection.

In high-voltage devices, the voltage-blocking structure is composed of chip termination and packaging, which is a semiconductor with multilayer dielectrics stacked structure. However, existing TCAD (Technology Computer-Aided Design) software fails to consider the Maxwell-Wagner effect in multilayer dielectrics. To address this issue, a transient state electric field model considering the Maxwell-Wagner effect for dielectric materials is developed and integrated into the standard TCAD workflow with acceptable computational cost. By experimental verification, the proposed model corresponds well with the measurement results and it reveals the influence of the Maxwell-Wagner effect in the multilayer dielectrics on the electric field characteristics of semiconductor. Moreover, the maximum relative error between the simulation based on the proposed model and measurement is 2.3%, while it is 15.38% when using original model in existing TCAD software. In addition, the influence of mesh size and parameters of dielectrics are discussed. Overall, the proposed model yields more accurate simulation results and can be used for robust insulation design under transient state.

To address the high computational costs of full-frame encryption and the risk of exposing sensitive locations in partial encryption, this paper proposes a video selective encryption and steganography scheme based on object detection and image inpainting. First, YOLOv8 is employed to achieve real-time and accurate detection of human targets in video frames. Then, the LIS-HMC hyperchaotic map and a new chaotic-driven interframe chain modulation (CDICM) strategy, combined with a designed row-column interchange and Roller confusion algorithm, are applied to selectively encrypt the target regions. Next, the globally and locally consistent image completion (GLCIC) algorithm is used to restore the background panoramically, eliminating visual discontinuities. Meanwhile, based on the Walsh-Hadamard transform (WHT), a multi-round embedding (MRE) steganography strategy is developed to hide the encrypted information within the restored background. Experimental results show that the encrypted data achieve an information entropy of 7.9925, a steganographic capacity of 0.75 bpp, and a PSNR above 44.91 dB after data embedding, demonstrating that the proposed method provides a new solution for video privacy protection that balances security, real-time performance, and visual naturalness.

Contemporary recommendation systems predominantly rely on ID embedding to capture latent associations among users and items. However, this approach overlooks the wealth of semantic information embedded within textual descriptions of items, leading to suboptimal performance and poor generalizations. Leveraging the capability of large language models to comprehend and reason about textual content presents a promising avenue for advancing recommendation systems. To achieve this, we propose an Llm-driven knowlEdge Adaptive RecommeNdation (LEARN) framework that synergizes open-world knowledge with collaborative knowledge. We address computational complexity concerns by utilizing pretrained LLMs as item encoders and freezing LLM parameters to avoid catastrophic forgetting and preserve open-world knowledge. To bridge the gap between the open-world and collaborative domains, we design a twin-tower structure supervised by the recommendation task and tailored for practical industrial application. Through experiments on the real large-scale industrial dataset and online A/B tests, we demonstrate the efficacy of our approach in industry application. We also achieve state-of-the-art performance on six Amazon Review datasets to verify the superiority of our method.

Deep learning is being increasingly applied to image and video compression in a new paradigm known as neural video compression. While achieving impressive rate–distortion (RD) performance, neural video codecs (NVC) require heavy neural networks, which in turn have large memory and computational costs and often lack important functionalities such as variable rate. These are significant limitations to their practical application. Addressing these problems, recent slimmable image codecs can dynamically adjust their model capacity to elegantly reduce the memory and computation requirements, without harming RD performance. However, the extension to video is not straightforward due to the non-trivial interplay with complex motion estimation and compensation modules in most NVC architectures. In this paper we propose the slimmable video codec framework (SlimVC) that integrates an slimmable autoencoder and a motion-free conditional entropy model. We show that the slimming mechanism is also applicable to the more complex case of video architectures, providing SlimVC with simultaneous control of the computational cost, memory and rate, which are all important requirements in practice. We further provide detailed experimental analysis, and describe application scenarios that can benefit from slimmable video codecs.

This work developed a kernel-based residual learning framework for quadrupedal robotic locomotion. Ini-tially, a kernel neural network is trained with data collected from an MPC controller. Alongside a frozen kernel network, a residual controller network is trained using reinforcement learning to acquire generalized locomotion skills and robust-ness against external perturbations. The proposed framework successfully learns a robust quadrupedal locomotion controller with high sample efficiency and controllability, which can provide omnidirectional locomotion at continuous velocities. We validated its versatility and robustness on unseen terrains that the expert MPC controller failed to traverse. Furthermore, the learned kernel can produce a range of functional locomotion behaviors and can generalize to unseen gaits.

We present a modular framework designed to enable a robot hand-arm system to learn how to catch flying objects, a task that requires fast, reactive, and accurately-timed robot motions. Our framework consists of five core modules: (i) an object state estimator that learns object trajectory prediction, (ii) a catching pose quality network that learns to score and rank object poses for catching, (iii) a reaching control policy trained to move the robot hand to pre-catch poses, (iv) a grasping control policy trained to perform soft catching motions for safe and robust grasping, and (v) a gating network trained to synthesize the actions given by the reaching and grasping policy. The former two modules are trained via supervised learning and the latter three use deep reinforcement learning in a simulated environment. We conduct extensive evaluations of our framework in simulation for each module and the integrated system, to demonstrate high success rates of in-flight catching and robustness to perturbations and sensory noise. Whilst only simple cylindrical and spherical objects are used for training, the integrated system shows successful generalization to a variety of household objects that are not used in training.

Deep learning models have achieved promising disease prediction performance of the Electronic Health Records (EHR) of patients. However, most models developed under the I.I.D. hypothesis fail to consider the agnostic distribution shifts, diminishing the generalization ability of deep learning models to Out-Of-Distribution (OOD) data. In this setting, spurious statistical correlations between procedures and diagnoses that may change in different environments will be exploited, which can cause sub-optimal performances of deep learning models and spurious correlation between historical EHR and future diagnosis. To address this problem, we propose to use a causal representation learning method called Causal Healthcare Embedding (CHE). CHE aims at eliminating the spurious statistical relationship by removing the dependencies between diagnoses and procedures. We introduce the Hilbert-Schmidt Independence Criterion (HSIC) to measure the degree of independence between the embedded diagnosis and procedure features. Based on causal view analyses, we perform the sample weighting technique to get rid of such spurious relationship for the stable learning of EHR across different environments. Moreover, our proposed CHE method can be used as a flexible plug-and-play module to enhance existing deep learning models on EHR. Extensive experiments on two public datasets and five state-of-the-art baselines unequivocally show that CHE can improve the prediction accuracy of deep learning models on out-of-distribution data by a large margin. In addition, the interpretability study shows that CHE could successfully leverage causal structures to reflect a more reasonable contribution of historical records for predictions.

Recently, Deep Neural Networks (DNNs) have made remarkable progress for text classification, which, however, still require a large number of labeled data. To train high-performing models with the minimal annotation cost, active learning is proposed to select and label the most informative samples, yet it is still challenging to measure informativeness of samples used in DNNs. In this paper, inspired by piece-wise linear interpretability of DNNs, we propose a novel Active Learning with DivErse iNterpretations (ALDEN) approach. With local interpretations in DNNs, ALDEN identifies linearly separable regions of samples. Then, it selects samples according to their diversity of local interpretations and queries their labels. To tackle the text classification problem, we choose the word with the most diverse interpretations to represent the whole sentence. Extensive experiments demonstrate that ALDEN consistently outperforms several state-of-the-art deep active learning methods.

For reliability, machine learning models in some areas, e.g., finance and healthcare, require to be both accurate and globally interpretable. Among them, credit risk assessment is a major application of machine learning for financial institutions to evaluate credit of users and detect default or fraud. Simple white-box models, such as Logistic Regression (LR), are usually used for credit risk assessment, but not powerful enough to model complex nonlinear interactions among features. In contrast, complex black-box models are powerful at modeling, but lack of interpretability, especially global interpretability. Fortunately, automatic feature crossing is a promising way to find cross features to make simple classifiers to be more accurate without heavy handcrafted feature engineering. However, existing automatic feature crossing methods have problems in efficiency on credit risk assessment, for corresponding data usually contains hundreds of feature fields.

The next location recommendation is at the core of various location-based applications. Current state-of-the-art models have attempted to solve spatial sparsity with hierarchical gridding and model temporal relation with explicit time intervals, while some vital questions remain unsolved. Non-adjacent locations and non-consecutive visits provide non-trivial correlations for understanding a user’s behavior but were rarely considered. To aggregate all relevant visits from user trajectory and recall the most plausible candidates from weighted representations, here we propose a Spatio-Temporal Attention Network (STAN) for location recommendation. STAN explicitly exploits relative spatiotemporal information of all the check-ins with self-attention layers along the trajectory. This improvement allows a point-to-point interaction between non-adjacent locations and non-consecutive check-ins with explicit spatio-temporal effect. STAN uses a bi-layer attention architecture that firstly aggregates spatiotemporal correlation within user trajectory and then recalls the target with consideration of personalized item frequency (PIF). By visualization, we show that STAN is in line with the above intuition. Experimental results unequivocally show that our model outperforms the existing state-of-the-art methods by 9-17%.

This article describes a comprehensive framework for analysis of time-dependent performance-reliability degradation of an SRAM cache, considering cache configurations, process parameters and their variations, supply voltage, and aging. The framework consists of three parts: microprocessor emulation, activity extraction, and evaluation of performance-reliability metrics. Evaluation of performance-reliability metrics is implemented with a prediction engine involving regression models for the metrics, which evaluates degradation due to various wearout mechanisms, including bias temperature instability (BTI), hot carrier injection (HCI), and random telegraph noise (RTN). The regression models not only enable more than 100× faster computation compared with SPICE simulations but also protect intellectual property. This framework has been applied to study how SRAM instruction cache (I-Cache) configurations, cell structure, inclusion of RTN and gate length variation, voltage scaling, and stress time affect the performance and reliability parameters, such as access time, leakage power, critical charge ( Qcrit), and static noise margin (SNM). We have also studied the impact of configuration parameters on the soft error rate (SER) and the hit rate of the I-Cache, and the impact of single error correction and double error detection (SECDED) error correcting codes (ECCs).

Click-Through Rate (CTR) prediction is a crucial task for various online applications, such as recommendation and online advertising. The task of CTR prediction is to predict the probability of users' clicking behaviors, with high-dimensional input features. To avoid heavy handcrafted feature engineering, the core topic of CTR prediction is the automatic interactions of the input features. Factorization Machine (FM) is an effective approach for modeling second-order feature interactions. Recently, FM has been extended for modeling higher-order feature interactions, such as xDeepFM and Higher-Order Factorization Machine (HOFM). However, these approaches are with either high complexity or iterative computation consuming much time and space. To overcome above problems, we express arbitrary-order FM in the form of power sums according to Newton's identities. Accordingly, we propose a novel Interaction Machine (IM) model. IM is an efficient and exact implementation of high-order FM, whose time complexity linearly grows with the order of interactions and the number of feature fields. Via IM, we can conduct arbitrary-order feature interactions in a very simple way. Moreover, we perform IM together with deep neural networks, and the resulted DeepIM model is more efficient than xDeepFM with comparable or even better performance. We conduct experiments on two real-world datasets, in which effectiveness and efficiency of both IM and DeepIM are strongly verified.

With the influence of information explosion, there are more and more choices exposed to public view. Next item recommendation is being a significant and challenging task. Recently, attention mechanism, Convolutional Neural Networks (CNN) and other kinds of deep components are used to model user behaviors. However, the proposed models often fail to extract the feature of user behaviors in different time periods and the CNN-based models before are hard to make the used CNN interpretable. In this paper, we propose a CNN & Attention-based Sequential Feature Extractor (CASFE) module to capture the possible features of user behaviors at different time intervals. Specifically, we import CNN to extract multi-level features of user behaviors with different time periods. After each CNN layer, we use attention module to emphasize the different effect of behaviors on the prediction result. Besides, the features we try to extract here have the similar concept and meaning with the hand-crafted features in Feature Engineering, which proves the validity of CASFE. Accordingly, CASFE becomes a general sequential feature extractor that can be used in various sequential prediction tasks. With Multi-Layer Perceptron (MLP), CASFE would be a state-of-the-art next item recommendation model. The model obtains good performance on Last.fm_1K dataset and MovieLens_1M dataset. Besides, as a compatible extractor module, it can also promote CTR prediction models as well as other sequential prediction tasks.