With the rapid growth in electromagnetic device quantities, various forms of communication interference have emerged, significantly impacting the accuracy of signal classification. Existing classification algorithms mainly focus on unintentional interference, such as co-channel interference and noise, with limited research on the problem of malicious interference in Multiple Input Multiple Output (MIMO) signal classification. This study proposes an intelligent MIMO signal classification algorithm based on fractional graph feature fusion. Initially, a high-order cumulant tensor model is constructed and regularized tensor decomposition is applied to reconstruct the MIMO signals. Subsequently, a feature extraction model using a fractional wavelet scattering network is designed to effectively capture the distinguishing features of signal constellations. Finally, a collaborative representation classifier based on the Grassmann manifold is utilized to amplify the differences between modulation categories, thereby improving classification performance. Simulation results indicate that the proposed algorithm effectively suppresses common communication interference and successfully classifies MIMO signals. Compared to existing methods, the proposed approach demonstrates significant performance improvements without requiring prior knowledge, such as noise power or channel coefficients.

As the electromagnetic environment becomes more complex, a significant number of interferences and malfunctions of authorized equipment can result in anomalies in spectrum usage. Utilizing intelligent spectrum technology to sense and identify anomalies in the electromagnetic space is of great significance for the efficient use of the electromagnetic space. In this paper, a method for intelligent sensing and identification of anomalies in spectrum with alpha‐stable noise is proposed. First, we use a delayed feedback network (DFN) to suppress alpha‐stable noise. Then, we use a long short‐term memory (LSTM) autoencoder‐based attention mechanism to sense anomaly. Finally, we use the deep forest model to identify abnormal spectrum. Simulation results demonstrate that the proposed method effectively suppresses alpha‐stable noise, and it outperforms existing methods in abnormal spectrum sensing and identification.

This article presents a novel distributed robust fuzzy control scheme to address the global consensus problem of unknown nonlinear multiagent systems (MASs). By replacing the nonlinear dynamic model constrained by the global Lipschitz condition with a more general system model, the proposed approach enhances applicability. A robust fuzzy control scheme based on a smooth switching function is introduced, effectively resolving the global consensus problem for unknown nonlinear systems. Furthermore, time-varying σ-modification terms are incorporated into the adaptive parameter design, replacing constant terms to avoid asymptotically uniform ultimate boundedness and ensuring global asymptotic consensus of the closed-loop systems. The efficacy of the proposed scheme is demonstrated through simulation results.

Passive sensing using communication signal waveforms is considered to be a promising technology for target monitoring in Internet-of-Everything. Conventional passive sensing schemes require accurate estimation of the time difference of arrival (TDOA) and frequency difference of arrival (FDOA), which is leading to high complexity but low accuracy. In this paper, a robust passive sensing algorithm using multiple illumination of opportunities is proposed to improve the detection performance while avoiding separate estimation of TDOA and FDOA. The proposed method first combines the linear constrained minimum variance adaptive filter with the wide nulling algorithm to achieve target direction finding while separating the direct wave and suppressing multipath interference. Then, the Linear Canonical Transformation-based Cross Ambiguity Function (LCTCAF) is employed to estimate the distance and radial velocity of the target. Relying on the relationship between distance to time and velocity to Doppler, a Distance-Velocity transformation-based Cross Ambiguity Function (DVCAF) is introduced to characterize the distance and radial velocity of the target. Finally, a spectral peak search scheme is exploited in DVCAF to estimate the time delay and Doppler shift so as to identify the target parameters directly. Its’ Cramer-Rao Low Bound is derived. Simulation results validate that the performance of the proposed algorithm outperforms the conventional estimators based on the cross ambiguity function.

Cognitive radio-enabled Internet of Things (CR-IoT) is considered as a promising technology to handle spectrum scarcity for IoT applications. Spectrum sensing enables unlicensed secondary users to exploit spectrum holes under the condition of avoiding interference with primary users in CR-IoT networks. Previous studies often assume that the noise is Gaussian while ignoring the influence of non-Gaussian noise. Moreover, multi-antenna-based spectrum sensing algorithms only consider the partial information of covariance matrix. This paper develops two multi-antenna-based spectrum sensing schemes, using fractional low-order covariance matrices to address the issue of performance degradation in impulsive noise. Specifically, the first scheme, namely, diagonal element weighting detection, exploits the diagonal element weighting of the fractional low-order covariance matrix. The latter scheme is called off-diagonal element weighting detection, which adopts the diagonal matrix weighting strategy that exploits the off-diagonal elements of fractional low-order covariance matrices. The approximate analytical expressions of the false alarm probability and detection probability are derived. These developed schemes do not employ any priori knowledge of the primary user signal. Simulation results indicate that two proposed schemes achieve acceptable performance and are robust to the characteristic exponent of the alpha-stable noise, e.g., these proposed methods could achieve a detection probability of 90% with a false alarm probability of 0.1 at GSNR = -16dB, respectively.

Signal identification, a vital task of intelligent communication radios, finds its applications in various military and civil communication systems. Previous works on identification for space-time block codes (STBC) of multiple-input multiple-output (MIMO) system employing orthogonal frequency division multiplexing (OFDM) are limited to additive white Gaussian noise. In this paper, we develop a novel automatic identification algorithm to exploit the generalized cross-correntropy function of the received signals to classify STBC-OFDM signals in the presence of Gaussian noise and impulsive interference. This algorithm first introduces the generalized cross-correntropy function to fully utilize the space-time redundancy of STBC-OFDM signals. The strongly-distinguishable discriminating matrix is then constructed by using the generalized cross-correntropy for multiple receive antennas. Finally, a decision tree identification algorithm is employed to identify the STBC-OFDM signals which is extended by the binary hypothesis test. The proposed algorithm avoids the traditionally required pre-processing tasks, such as channel coefficient estimation, noise and interference statistics prediction and modulation type recognition. Numerical results are presented to show that the proposed scheme provides good identification performance by exploiting the generalized cross-correntropy function of STBC-OFDM signals under impulsive interference circumstances.

Several prior studies have suggested that word frequency biases can cause the Bert model to learn indistinguishable sentence embeddings. Contrastive learning schemes such as SimCSE and ConSERT have already been adopted successfully in unsupervised sentence embedding to improve the quality of embeddings by reducing this bias. However, these methods still introduce new biases such as sentence length bias and false negative sample bias, that hinders model's ability to learn more fine-grained semantics. In this paper, we reexamine the challenges of contrastive sentence embedding learning from a debiasing perspective and argue that effectively eliminating the influence of various biases is crucial for learning high-quality sentence embeddings. We think all those biases are introduced by simple rules for constructing training data in contrastive learning and the key for contrastive learning sentence embedding is to "mimic" the distribution of training data in supervised machine learning in unsupervised way. We propose a novel contrastive framework for sentence embedding, termed DebCSE, which can eliminate the impact of these biases by an inverse propensity weighted sampling method to select high-quality positive and negative pairs according to both the surface and semantic similarity between sentences. Extensive experiments on semantic textual similarity (STS) benchmarks reveal that DebCSE significantly outperforms the latest state-of-the-art models with an average Spearman's correlation coefficient of 80.33% on BERTbase.

New findings in natural language processing (NLP) demonstrate that the strong memorization capability contributes a lot to the success of Large Language Models (LLM). This inspires us to explicitly bring an independent memory mechanism into CTR ranking model to learn and memorize cross features' representations. In this paper, we propose multi-Hash Codebook NETwork (HCNet) as the memory mechanism for efficiently learning and memorizing representations of cross features in CTR tasks. HCNet uses a multi-hash codebook as the main memory place and the whole memory procedure consists of three phases: multi-hash addressing, memory restoring, and feature shrinking. We also propose a new CTR model named MemoNet which combines HCNet with a DNN backbone. Extensive experimental results on three public datasets and online test show that MemoNet reaches superior performance over state-of-the-art approaches. Besides, MemoNet shows scaling law of large language model in NLP, which means we can enlarge the size of the codebook in HCNet to sustainably obtain performance gains. Our work demonstrates the importance and feasibility of learning and memorizing representations of cross features, which sheds light on a new promising research direction. The source code is in https://github.com/ptzhangAlg/RecAlg.

Click-Through Rate (CTR) estimation has become one of the most fundamental tasks in many real-world applications and various deep models have been proposed. Some research has proved that FiBiNet is one of the best performance models and outperforms all other models on Avazu dataset. However, the large model size of FiBiNet hinders its wider application. In this paper, we propose a novel FiBiNet++ model to redesign FiBiNet's model structure, which greatly reduces model size while further improves its performance. One of the primary techniques involves our proposed "Low Rank Layer" focused on feature interaction, which serves as a crucial driver of achieving a superior compression ratio for models. Extensive experiments on three public datasets show that FiBiNet++ effectively reduces non-embedding model parameters of FiBiNet by 12x to 16x on three datasets. On the other hand, FiBiNet++ leads to significant performance improvements compared to state-of-the-art CTR methods, including FiBiNet. The source code is in https://github.com/recommendation-algorithm/FiBiNet.

Signal classification plays a pivotal role in cognitive radio networks. This problem becomes more challenging for multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) systems employing linear space-time block code (STBC). This paper introduces a novel classification scheme to exploit the generalized cross-correntropy function to classify STBC-OFDM signals in the presence of impulsive interference. This scheme relies on the generalized cross-correntropy statistics of the STBC-OFDM signals to construct the strongly-distinguishable discriminating matrix. The proposed scheme avoids the estimation of channel coefficients, noise and interference statistics, and modulation types. Numerical results are presented to show that the proposed scheme provides acceptable classification performance in the presence of Gaussian noise and impulsive interference.

We propose a novel multimodal video benchmark - the Perception Test - to evaluate the perception and reasoning skills of pre-trained multimodal models (e.g. Flamingo, BEiT-3, or GPT-4). Compared to existing benchmarks that focus on computational tasks (e.g. classification, detection or tracking), the Perception Test focuses on skills (Memory, Abstraction, Physics, Semantics) and types of reasoning (descriptive, explanatory, predictive, counterfactual) across video, audio, and text modalities, to provide a comprehensive and efficient evaluation tool. The benchmark probes pre-trained models for their transfer capabilities, in a zero-shot / few-shot or limited finetuning regime. For these purposes, the Perception Test introduces 11.6k real-world videos, 23s average length, designed to show perceptually interesting situations, filmed by around 100 participants worldwide. The videos are densely annotated with six types of labels (multiple-choice and grounded video question-answers, object and point tracks, temporal action and sound segments), enabling both language and non-language evaluations. The fine-tuning and validation splits of the benchmark are publicly available (CC-BY license), in addition to a challenge server with a held-out test split. Human baseline results compared to state-of-the-art video QA models show a significant gap in performance (91.4% vs 45.8%), suggesting that there is significant room for improvement in multimodal video understanding.Dataset, baselines code, and challenge server are available at https://github.com/deepmind/perception_test

Unmanned aerial vehicle (UAV) communication is a promising technology that provides swift and flexible on-demand wireless connectivity for devices without infrastructure support. The proliferation of UAV communication equipment is causing the limited spectrum to become crowded. To deal with this issue, spectrum sharing policy (SSP) is introduced to support UAV communication. Spectrum sensing in SSP must be carefully formulated to control interference to the primary users and ground communications. In this paper, we propose spectrum sensing for opportunistic spectrum access in UAV communication to improve the spectrum utilization efficiency. Different from most existing works, we focus on the problem of spectrum sensing with randomly arriving primary signals in the presence of non-Gaussian noise/interference. We propose a novel spectrum sensing scheme to improve the spectrum utilization efficiency in UAV communication. We construct the p-norm decision statistic based on the assumption that the random arrivals of signals follow a Poisson process. Simulation results illustrate the validity and superiority of the proposed scheme when the primary signals are corrupted by additive non-Gaussian noise and are arriving randomly during spectrum sensing in the UAV communication.

Radio frequency fingerprint identification (RFFI) technology identifies the emitter by extracting one or more unintentional features of the signal from the emitter. To solve the problem that the traditional deep learning network is not highly adaptable for the contour features extracted from the signal, this paper proposes a novel RFFI method based on a deformable convolutional network. This network makes the convolution operation more biased towards the useful information content in the feature map with higher energy, and ignores part of the background noise information. The proposed blind identification method requires less information and no training sequences and pilots, Thus, it achieves energy and spectrum efficient radio communications. Simulation verifies that the proposed method can achieve better recognition performance and is beneficial for green radios.

Identification of transmit-antenna number is of importance in cognitive Internet of Things (IoT) with multiple-input–multiple-output (MIMO). Previous studies on transmit-antenna number detection only consider Gaussian noise and ignore impulsive interference. In the practical wireless communication, impulsive interference may exist due to low-frequency atmospheric noise, multiple access, and electromagnetic disturbance. Such interference can usually be modeled as symmetric alpha stable ($S\alpha S$), which cause the performance degradation of conventional algorithms based on the Gaussian model. In this article, we present a novel scheme to detect the transmit-antenna number for MIMO systems in cognitive IoT, assuming that signals are corrupted by both$S\alpha S$interference and Gaussian noise. We first introduce a new approach to characterize the generalized correlation matrix (GCM), and provide its bound with$S\alpha S$interference. Then, the discriminating feature vector is constructed by utilizing the higher order moments (HOMs) of eigenvalues of the GCM. Finally, an advanced clustering algorithm is employed to detect the transmit-antenna number, using the cluster where the minimum eigenvalue is located. The proposed algorithm avoids the need fora prioriinformation about the transmitted signals, such as coding mode, modulation type, and pilot patterns. Simulation experiments demonstrate the feasibility of the proposed transmit-antenna number detection scheme in MIMO systems with Gaussian noise and$S\alpha S$interference.

Estimation of communication parameters, a major task of intelligent receivers, has important applications in adaptive wireless systems. Multiple antennas make the identification problem more challenging. In this paper, we focus on the problem of estimating the number of transmit antennas in multiple-input-multiple-output (MIMO) communication systems. A novel estimation algorithm is proposed to determine the number of transmit antennas for MIMO with alpha-stable noise. We first introduce the correlation matrix based on the fractional lower order statistics (FLOS) and provide a particular structure of FLOS-based correlation matrix. Then, the eigenvalues of the FLOS-based correlation matrix are employed to construct a test statistic and the central limit theorem is exploited to obtain the decision threshold. Finally, the transmit-antenna number is estimated using a serial binary hypothesis test. Simulation results are demonstrated to evaluate the effectiveness of the proposed transmit-antenna number estimation algorithm for MIMO with alpha-stable noise.

In this paper, we propose a novel detection algorithm of the number of transmit antennas in multiple-input multiple-output (MIMO) systems, assuming that signals corrupted by non-Gaussian interference and Gaussian noise. We first introduce generalized correlation matrix. Then, the discriminating feature vector is constructed by exploiting the higher-order moment of the eigenvalues. Finally, an advanced clustering algorithm is employed for decision the number of transmit antennas, which is determined by the dimension of the cluster where the minimum eigenvalue is located. The proposed algorithm does not require a priori information about the transmitted signals, such as coding scheme, modulation type, and pilot patterns. Simulation results are demonstrated to evaluate the effectiveness of the proposed transmit-antenna number detection algorithm in MIMO systems with Gaussian Noise and non-Gaussian interference.

The emerging cognitive radio-based Internet-of-Things (CR-IoT) network provides a novel paradigm solution for IoT devices to efficiently utilize spectrum resources. Spectrum sensing is a critical problem in the CR-IoT network which has been investigated extensively under the Gaussian noise/interference. Since most of the interference in an IoT network is non-Gaussian, in this article, we introduce a novel spectrum sensing method for CR-IoT with additive Gaussian mixture noise/interference. The introduced method maps the observation signal matrix from the original input space to a high-dimensional feature space by a nonlinear Gaussian kernel function and then constructs a kernelized test statistic in the feature space. The approximate analytical expressions of the false alarm and detection probability of the proposed scheme are derived under Gaussian mixture noise, and the decision threshold can be determined according to false alarm probability. The simulation results show that the introduced multiple-input-multiple-output (MIMO) spectrum sensing method achieves good performance under Gaussian mixture noise/interference and significantly outperforms existing detectors.

Blind estimation of parameters for M-ary frequency-shift-keying (M-FSK) signals is great of importance in intelligent receivers. Many existing algorithms have assumed white Gaussian noise. However, their performance severely degrades when grossly corrupted data, i.e., outliers, exist. This article solves this issue by developing a novel approach for parameter estimation of M-FSK signals in the presence of alpha-stable noise. Specifically, the proposed method exploits the generalized first- and second-order cyclostationarity of M-FSK signals with alpha-stable noise, which results in closed-form solutions for unknown parameters in both time and frequency domains. As a merit, it is computationally efficient and thus can be used for signal preprocessing, symbol timing estimation, signal and noise power estimation. Furthermore, substantial theoretical analysis on the performance of the proposed approach is provided. Simulations demonstrate that the proposed method is robust to alpha-stable noise and that it outperforms the state-of-the-art algorithms in many challenging scenarios.

Advertising and feed ranking are essential to many Internet companies such as Facebook and Sina Weibo. Among many real-world advertising and feed ranking systems, click through rate (CTR) prediction plays a central role. There are many proposed models in this field such as logistic regression, tree based models, factorization machine based models and deep learning based CTR models. However, many current works calculate the feature interactions in a simple way such as Hadamard product and inner product and they care less about the importance of features. In this paper, a new model named FiBiNET as an abbreviation for Feature Importance and Bilinear feature Interaction NETwork is proposed to dynamically learn the feature importance and fine-grained feature interactions. On the one hand, the FiBiNET can dynamically learn the importance of features via the Squeeze-Excitation network (SENET) mechanism; on the other hand, it is able to effectively learn the feature interactions via bilinear function. We conduct extensive experiments on two real-world datasets and show that our shallow model outperforms other shallow models such as factorization machine(FM) and field-aware factorization machine(FFM). In order to improve performance further, we combine a classical deep neural network(DNN) component with the shallow model to be a deep model. The deep FiBiNET consistently outperforms the other state-of-the-art deep models such as DeepFM and extreme deep factorization machine(XdeepFM).

Social voting is an emerging new feature in online social networks. It poses unique challenges and opportunities for recommendation. In this paper, we develop a set of matrix-factorization (MF) and nearest-neighbor (NN)-based recommender systems (RSs) that explore user social network and group affiliation information for social voting recommendation. Through experiments with real social voting traces, we demonstrate that social network and group affiliation information can significantly improve the accuracy of popularity-based voting recommendation, and social network information dominates group affiliation information in NN-based approaches. We also observe that social and group information is much more valuable to cold users than to heavy users. In our experiments, simple metapath-based NN models outperform computation-intensive MF models in hot-voting recommendation, while users' interests for nonhot votings can be better mined by MF models. We further propose a hybrid RS, bagging different single approaches to achieve the best top-k hit rate.

General information retrieval systems are designed to serve all users without considering individual needs. In this paper, we propose a novel approach to personalized search. It can, in a unified way, exploit and utilize implicit feedback information, such as query logs and immediately viewed documents. Moreover, our approach can implement result re-ranking and query expansion simultaneously and collaboratively. Based on this approach, we develop a client-side personalized web search agent PAIR (Personalized Assistant for Information Retrieval), which supports both English and Chinese. Our experiments on TREC and HTRDP collections clearly show that the new approach is both effective and efficient.

Domain specific research engine has many advantages over its generic counterparts such as high precision and short update period. However it is a time consuming job to construct these applications. We introduce an object-oriented framework for domain specific search engine applications, which contain some design patterns that contribute to this object-oriented architecture, revealing the framework's structure and the forces that shaped it. Using this framework we fix a basic architecture and thus increase ability to construct domain specific search engine application.

With the widespread use of computers in translation work and daily life, there are more and more bilingual corpora becoming available. In this paper, the PECAT (Pilot English-Chinese Computer-Aided Translation) system, based on bilingual corpora, is described. There are mainly three modules in our system: a corpus-processing module, a sentence-matching module and a post-editing module. In order to increase the coverage of input source sentences, the text alignment in the corpus-processing module is based on the chunk level. The matching algorithm of input sentences and source sentences is a three-layer edit-distance algorithm guided by the user, which includes information about word morphology and part-of-speech (POS). Preliminary experiments show promising results.