Accurate classification of healthcare time series is critical for clinical decision-making. However, existing models often struggle under real-world data shifts and lack interpretability- two key requirements for reliable medical deployment. To address these challenges, we propose SHINE, a novel endto-end framework that learns disentangled and shift-invariant representations by modeling the generative process of multivariate healthcare signals. Specifically, SHINE first introduces a genuine data representation learning that disentangles healthcare signals into trend, seasonality, and noise components, reflecting distinct temporal dynamics of healthcare series. Then, we inject several inductive biases into each component to encourage latent representations to be invariant to data shifts and aligned with their corresponding semantic units. Extensive experiments on six healthcare benchmarks spanning ECG, EEG, and continuous glucose monitoring (CGM) domains-under a variety of simulated real-world shift scenarios-demonstrate that SHINE consistently outperforms state-of-the-art baselines, providing robust performance and clinically meaningful interpretations grounded in the estimated components.

Summary Convolutional neural networks (CNNs) have shown remarkable advantages in a wide range of domains at the expense of huge parameters and computations. Modern CNNs still tend to be more complex and larger to achieve better inference accuracy. However, the complex and large structures of CNNs could slow down the inference speed. Recently, Compressing the convolutional weights to be sparse by pruning the unimportant parameters has been demonstrated as an efficient way to reduce the computations of CNNs. On the other hand, field‐programmable gate arrays (FPGAs) have been a popular hardware platform to accelerate CNN inference. In this paper, we propose an algorithm/hardware co‐optimized method for accelerating CNN inference on FPGAs. For the algorithm, we take advantage of unstructured and structured parameter sparsifying methods to achieve high sparsity and keep the regularity of convolutional weights. Correspondingly, hardware‐friendly index representations of sparse convolutional weights are proposed. For the hardware architecture, we propose row‐wise input‐stationary dataflow, which is tightly coupled with the algorithm. A row‐wise computing engine (RConv Engine) is proposed, which is based on the dataflow. Inside the RConv Engine, the scalar‐vector structure is applied to implement the basic processing elements (PEs). To flexibly calculate the feature map with various sizes, the PEs are organized in a 2D structure with two work modes. The experimental results demonstrate that our co‐optimized method implements high sparsity of convolutional weights, and the computing engine achieves high computation efficiency. Compared with other accelerators, our co‐optimized method implements a 10.9 speedup on FPS at most with the highest sparsity of convolutional weights and negligible accuracy loss.

Streaming Recommender Systems (SRSs) have emerged to deliver recommendations based on pervasive data streams, which are a sequence of user-item interactions with multiple behavior types (e.g., purchase, add-to-cart, and view). However, existing SRSs all rely on a single behavior type (e.g., purchase) to make streaming recommendations, and commonly suffer from the data sparsity problem. To address this issue, the relatively more abundant multi-behavior interactions (i.e., interactions with multiple behavior types) could be well leveraged for more accurate streaming recommendations. However, it remains a challenge on how to effectively leverage the commonly-existing and complex multi-behavior interactions for improving the accuracy of streaming recommendations. Targeting at this challenge, we propose the first Multi-behavior Streaming Recommender System in the literature, called MbSRS, to elaborately exploit multi-behavior interactions for delivering accurate recommendations in streaming scenarios. In MbSRS, we first learn instant user preferences and unified item characteristics collaboratively from multi-behavior interactions. Then, we attentively learn long-term user preferences from the historical items interacted by the corresponding users. After that, we wisely fuse the learned instant and long-term user preferences via a gate mechanism. Finally, a novel multi-behavior-specific training process is devised for more effectively learning user preferences towards items from multi-behavior interactions. Extensive experiments on three real-world datasets demonstrate that the proposed MbSRS significantly outperforms the state-of-the-art baselines.

Using fake audio to spoof the audio devices in the Internet of Things has become an important problem in modern network security. Aiming at the problem of lack of robust features in fake audio detection, an audio streams’ hidden feature extraction method based on a heuristic mask for empirical mode decomposition (HM-EMD) is proposed in this paper. First, using HM-EMD, each signal is decomposed into several monotonic intrinsic mode functions (IMFs). Then, on the basis of IMFs, basic features and hidden information features HCFs of audio streams are constructed, respectively. Finally, a machine learning method is used to classify audio streams based on these features. The experimental results show that hidden information features of audio streams based on HM-EMD can effectively supplement the nonlinear and nonstationary information that traditional features such as mel cepstrum features cannot express and can better realize the representation of hidden acoustic events, which provide a new research idea for fake audio detection.

Due to the increasing scale of cloud data center, the issue of energy consumption is becoming pretty significant. To tackle this problem, an extremely effective approach is increasing the utilization of resource in data center. Researchers have found that accurate load prediction can help allocator distribute resource reasonably, so as to increase the utilization. There are a lot of traditional prediction algorithms which have been applied to cloud data center, such as linear regression. However, with the development of technologies, a number of novel prediction algorithms are brought out, for example, neural network. This paper assesses and analyzes the performance of several different prediction algorithms applying on data sets from real world. We get some meaningful and interesting conclusions from comparison among these algorithms, which may offer references for system designers of cloud data center.

The fault correction process is very important in software testing, and it has been considered into some software reliability growth models (SRGMs). In these models, the time-delay functions are often used to describe the dependency of the fault detection and correction processes. In this paper, a more direct variable "correction lag", which is defined as the difference between the detected and corrected fault numbers, is addressed to characterize the dependency of the two processes. We investigate the correction lag and find that it appears Bell-shaped. Therefore, we adopt the Gamma function to describe the correction lag. Based on this function, a new SRGM which includes the fault correction process is proposed. And the experimental results show that the new model gives better fit and prediction than other models.

The testing and operation environments may be essentially different, thus the fault detection rate (FDR) of testing phase is different from that of the operation phase. In this paper, based on the representative model, G-O model, of nonhomogeneous Poisson process (NHPP), a transformation is performed between the FDR of the testing phase to that of the operation considering the profile differences of the two phases, and then a software reliability growth model (SRGM) called TO-SRGM describing the differences of the FDR between the testing phase and the operation phase is proposed. Finally, the parameters of the model are estimated using the least squares estimate (LSE) based on normalized failure data. Experiment results show that the goodness-of-fit of the TO-SRGM is better than that of the G-0 model and the PZ-SRGM on the normalized failure data set.

This paper presents a SRGM (software reliability growth model) from testing to operation based on NHPP (nonhomogeneous Poisson process). Although a few research projects have been devoted to the differences between testing environment and operational environment, consideration of the variation of environmental influence factors along testing time in the existing models is limited. The model in this paper is the first scheme of a few NHPP models which take environmental factors experimented from actual failure data as a function of testing time. FDR (fault detection rate) is usually used to measure the effectiveness of fault detection by test techniques and test cases. A bell-shaped FDR function is proposed which integrate both environmental factors and inherent FDR per fault. A NHPP SRGM from testing to operation that incorporates environmental factors called TOE-SRGM is built which integrates FDR of testing phase and the proposed FDR function of operational phase. TOE-SRGM is evaluated using a set of software failure data. The results show that TOE-SRGM fits the failure data better than G-O model and PZ-SRGM.