Latent user representations are widely adopted in the tech industry for powering personalized recommender systems. Most prior work infers a single high dimensional embedding to represent user, which is good starting point but falls short delivering full understanding of user's interests. In this work, we introduce PinnerSage, an end-to-end system that represents each via multi-modal embeddings and leverages rich representation users provides quality recommendations. PinnerSage achieves by...

A deception jamming method for synthetic aperture radar (SAR) based on multiple receivers is proposed. This deals with the parameters of hostile SAR as a whole jamming. By measuring range differences between different receivers, higher accuracy can be achieved than detecting each directly. The squint angle, which hard to obtained electronic reconnaissance, also included in multireceiver system gets, so this works SAR. algorithm and noise are analyzed. Simulation results verify effectiveness...

This paper addresses the issue of deceptive jamming against synthetic aperture radar (SAR) by using 1-bit sampling and time-varying threshold (TVT). With intercepted SAR signal, multipliers involved in a convolution is replaced xnor gates, which considerably simplify signal generation. Moreover, TVT used for quantization before retransmission to retain relative amplitude information signal. As result, proposed schemes are superior their conventional counterpart terms realization. Effects...

Recently, multifeature learning in collaborative representation classification (CRC) for hyperspectral images has generated promising performance. In this paper, two novel algorithms that update dictionary directly and indirectly are proposed. order to offer the complementarity of multifeature, four different types features-global feature (i.e., Gabor feature), local binary pattern), shape extended multiattribute profiles), spectral feature-are adopted paper. Under hypothesis most features...

Spatial information has widely been used in hyperspectral image (HSI) classification to improve accuracy. However, the structural may not be fully explored when using spatial information, this paper proposes joint collaborative representation with correlation matrix (CRC-CM) for HSI by features patches, which could keep local intrinsic structure band images. Considering heterogeneity a patch, matrices of target neighborhood patch and training are improved binary weight shape-adaptive...

Preparation of high-efficiency dual-functional catalysts remains the bottleneck for electrochemical water splitting. To prepare a non-precious metal catalyst with high activity and stability, here, we present seaweed-like structure consisting transition-metal sulfide nanoplates self-assembled on carbon nanotube sponge networks (SW-CoS@CNT). By adjusting key parameters during synthesis (e.g., loading amount ratio Co S precursors), microstructure can be tailored in wide range, sulfur defects...

Multimodal Large Language Models (MLLMs) have demonstrated notable capabilities in general visual understanding and reasoning tasks. However, their deployment is hindered by substantial computational costs both training inference, limiting accessibility to the broader research user communities. A straightforward solution leverage smaller pre-trained vision language models, which inevitably causes significant performance drop. In this paper, we demonstrate possibility beat scaling law train a...

This paper addresses a novel SAR imaging scheme based on 1-bit sampling assisted with single-frequency threshold. The approach is able to considerably reduce the quantization cost. However, when technique simplifies system by reducing word length of each sample only 1 bit, amplitude information echo lost and high-order harmonics are introduced, degrading quality. strategy threshold linearly maintain shifts spectra away from component caused intermodulation. Hence, quality using sampled data...

This paper considers the problem of reconstructing true targets in a single-channel synthetic aperture radar (SAR) imaging system, which has been disturbed by deceptive jammings. Since jammings are usually confined to main lobe an SAR antenna, their time-frequency distributions different from those echoes. enables us utilize dynamic (DSA) scheme extract characteristics and false targets. Dictionaries about constructed taking interactions between scatterers into account. Then sparsity-driven...

The effect of the primary user traffic on performance spectrum sensing is investigated. investigation considers both local and collaborative sensing. Numerical results show that can be significantly degraded if channel state changes frequently, effective in alleviating deleterious caused by traffic.

This paper proposes a new approach to improve the error performance of synergy netting deception jamming against synthetic aperture radar (SAR). The existing methods cannot tackle issue amplification induced by linear equations constructed for coefficients calculation. To handle this problem, we develop an optimal layout receivers minimize effect minimizing condition number, which measures sensitivity system perturbation. It is revealed that devised relies little on SAR kinematic parameters....

This paper proposed a novel algorithm to suppress the deception jamming against single channel synthetic aperture radar (SAR), which provides an effective and economic way improve anti-deception-jamming ability of conventional SAR system. The dynamic apertures are employed provide multiple observations about time-frequency distributions real echoes. amplitude loss in different is modeled formulate optimization problem for suppressing. Super-resolution scheme also involved tighten feasible...

Aerogels and foams are promising electrode materials owing to their lightweight, high porosity, large surface area for creating abundant active/catalytic sites. Tailoring porous structure is essential toward maximum performance yet remains challenging in the field. Here, by modifying a pristine carbon nanotube (CNT) sponge with random internal distribution, we present CNT platform consisting of regular, orthogonally intercrossed through-channels centered at suitable lateral size (around 5...

We propose a new deceptive jamming method using networked receivers to perform against multimode synthetic aperture radar. Using multiple measure the time difference of beam arrival, we solve instantaneous modulation parameters precisely via linear equations. then delays and frequency offsets signal without approximation avoid accumulating electronic reconnaissance errors with traditional methods. Additionally, design distribution minimize condition number, reducing our method's sensitivity...

A deceptive jamming method against terrain observation by progressive scans synthetic aperture radar (TOPSAR) that considers the effect of changing relative position between a jammer and TOPSAR platform is presented. The can generate flexible false targets with varying Doppler centroids. problem decomposed into simpler subproblems same type, according to satellite. partial modulation function be pregenerated before parameter reconnaissance. partitioned scheme produces multiple batches in...

A fast image stitching algorithm based on improved speeded up robust feature (SURF) is proposed to overcome the real-time performance and robustness of original SURF algorithms. The machine learning method adopted build a binary classifier, which identify key points extracted by remove non-key points. In addition, RELIEF-F used for dimension reduction simplification descriptor achieve registration. threshold-based weighted fusion seamless stitching. Finally, several experiments are conducted...

Matrix completion (MC) is a promising technique which able to recover an intact matrix with low-rank property from sub-sampled/incomplete data. Its application varies computer vision, signal processing wireless network, and thereby receives much attention in the past several years. There are plenty of works addressing behaviors applications MC methodologies. This work provides comprehensive review for approaches perspective processing. In particular, problem first grouped into six...

Recently, collaborative representation for hyperspectral image analysis has received great interest. Due to the effectiveness of local manifold in a tangent space, this letter extends classification (CRC) mechanism into space. Specifically, uses simplified distance and new regularization term designs modified classifier innovatively. Moreover, two variants with weighted diagonal matrices adaptively adjust terms are developed further improve performance. In experiments, real images were...