Model quantization is an effective way to compress deep neural networks and accelerate the inference time on edge devices. Existing quantization methods usually require original data for calibration during the compressing process, which may be inaccessible due to privacy issues. A common way is to generate calibration data to mimic the origin data. However, the generators in these methods have the mode collapse problem, making them unable to synthesize diverse data. To solve this problem, we leverage the information from the full-precision model and enhance both inter-class and intra-class diversity for generating better calibration data, by devising a multi-layer features mixer and normalization flow based attention. Besides, novel regulation losses are proposed to make the generator produce diverse data with more patterns from the perspective of activated feature values and for the quantized model to learn better clip ranges adaptive to our diverse calibration data. Extensive experiments show that our method achieves state-of-the-art quantization results for both Transformer and CNN architectures. In addition, we visualize the generated data to verify that our strategies can effectively handle the mode collapse issue. Our codes are available at repo.

We study the problem of time series anomaly prediction, which is relevant to a range of real-world applications. Existing anomaly prediction methods rely on labeled training data for achieving acceptable accuracy. However, such data may be difficult to obtain; and in real-time deployments, anomalies can occur that were not seen in labeled data, thus making them difficult to predict. We provide a theoretical analysis and propose an Importance-based Generative Contrastive Learning method (IGCL) for unsupervised anomaly prediction. IGCL employs a controlled diffusion module to produce anomaly precursor patterns. Next, ICGL learns contextual representations to extract temporal dependencies from pairs of normal time series and anomaly precursors. IGCL is then able to predict anomalies by identifying anomaly precursors that will evolve into future anomalies. To address challenges caused by potentially complex precursor combinations involving multiple variables, we propose a memory bank with importance scores that stores representative samples adaptively and generates more complex anomaly precursors. Extensive experiments on nine benchmark datasets offer evidence that the proposed method is able to outperform state-of-the-art baselines.

The spatial resolution is one of the key issues to be addressed for wireless sensing, which traditionally relies on the signal bandwidth. However, directly enhancing the signal bandwidth is less effective, due to the challenges in (i) hardware implementations for large bandwidth signal generation and reception, and (ii) strict wireless spectrum management. Therefore, solutions those try to combine observations from multi-band signals to improve the spatial resolution have become attractive so far. In this paper, we propose a scattering centers cutting algorithm (SCC) based on the well-known Geometrical Theory of Diffraction (GTD) model, for frequency stitching sensing. A method based on the generalized likelihood ratio test (GLRT) method and SCC are combined, to optimize the target model order and related coefficients, which can greatly improve the sensing performance. Simulation results and real-world experiments are both provided, to show the effectiveness of the proposed SCC solution.

Some special environments, such as closed or harsh spaces, require contactless and real-time humidity monitoring. A flexible 920 MHz antenna sensor with vertical graphene (VG) as a sensitive film is proposed, which can measure humidity and transmit data wirelessly in real time. When the humidity changes, the number of H2O molecules absorbed by the graphene film will vary, which affects the permittivity of the antenna patch, and results in shifts in the resonant frequency. Remote humidity monitoring can be realized by analyzing the frequency response at the receiver. By depositing vertical graphene film on the antenna patch, the sensitivity can be improved from 16.3 to 32.6 kHz/% RH (linear range: 0. 57% RH) and from 145 to 290.5 kHz/% RH (57.90% RH). During humidity change, the radiation direction of the VG-based antenna sensor remains stable, and the maximum wireless transmission distance reaches 100 m. For the first time, the model of the antenna sensor is established. Furthermore, the flexible sensor can conform to various installation shapes. The experimental results show that the wireless VG-based antenna sensor is a good candidate for remote humidity monitoring and the Internet of Things.

Flink clusters often suffer from hotspot issues where the monitored job delay and CPU usage keep rising and remain high. This necessitates the detection of anomalous time series to pinpoint the hotspot machines. However, the state-of-the-art unsupervised time series anomaly detection (UTAD) methods are ineffective in this scenario. We identify two main reasons for this. First, the hotspot scenario requires us to pay particular attention to Flink-specific anomalies, e.g., slow-rising and high-level anomalies, which the existing methods struggle to address. Second, the state-of-the-art anomaly detection methods often assume that training datasets do not contain anomalies, but the data collected from the running Flink clusters contains noise, which causes these methods to learn anomalous patterns as normal patterns. In this paper, we first conduct experiments to analyze why existing methods fail in the Flink scenario. To tackle these challenges, we propose a cross-contrastive approach to learn the context information for each timestamp to enable Flink-specific anomaly detection. Then, to address noisy anomalies, we incorporate prior knowledge to set an anomaly boundary to prevent the model from learning anomalous patterns. Extensive experiments show that our method not only outperforms existing methods in the Flink scenario but also achieves state-of-the-art results on public benchmark datasets.

Accurate human motion classification is required in different fields, which currently mainly relies on WiFi and millimeter waves. Meanwhile, the Ultra-Wideband (UWB) signal offers high spatiotemporal resolution and robust interference resistance, making it suitable for distinguishing limb motions from other body movements, such as body tremors, heartbeat, and respiration. This paper proposes a simple but effective method for arm motion classification and recognition using an Impulse Radio UWB radar. We simplify the motion feature data by extracting the Peak Amplitude Curve (PAC) from the obtained time-frequency spectrum, resulting in reduced dimensionality and sample size. By employing traditional machine learning models instead of complex deep learning models, we achieve a recognition accuracy of up to 94.7%. To demonstrate the reliability of the model, we conducted ten-fold cross-validation, which yielded an average recognition accuracy of 93%.