Search

The modeling of users' behaviors is crucial in modern recommendation systems. A lot of research focuses on modeling users' lifelong sequences, which can be extremely long and sometimes exceed thousands of items. These models use the target item to search for the most relevant items from the historical sequence. However, training lifelong sequences in click through rate (CTR) prediction or personalized search ranking (PSR) is extremely difficult due to the insufficient learning problem of ID embedding, especially when the IDs in the lifelong sequence features do not exist in the samples of training dataset. Additionally, existing target attention mechanisms struggle to learn the multi-modal representations of items in the sequence well. The distribution of multi-modal embedding (text, image and attributes) output of user's interacted items are not properly aligned and there exist divergence across modalities. We also observe that users' search query sequences and item browsing sequences can fully depict users' intents and benefit from each other. To address these challenges, we propose a unified lifelong multi-modal sequence model called SEMINAR-Search Enhanced Multi-Modal Interest Network and Approximate Retrieval. Specifically, a network called Pretraining Search Unit (PSU) learns the lifelong sequences of multi-modal query-item pairs in a pretraining-finetuning manner with multiple objectives: multi-modal alignment, next query-item pair prediction, query-item relevance prediction, etc. After pretraining, the downstream model restores the pretrained embedding as initialization and finetunes the network. To accelerate the online retrieval speed of multi-modal embedding, we propose a multi-modal codebook-based product quantization strategy to approximate the exact attention calculati, Comment: 9 pages,code released

In this paper, we are concerned with the global existence of small data weak solutions to the $n-$dimensional semilinear wave equation $\partial_t^2u-\Delta u+\frac{\mu}{t}\partial_tu=|u|^p$ with time-dependent scale-invariant damping, where $n\geq 2$, $t\geq 1$, $\mu\in(0,1)\cup(1,2]$ and $p>1$. This equation can be changed into the semilinear generalized Tricomi equation $\partial_t^2u-t^m\Delta u=t^{\alpha(m)}|u|^p$, where $m=m(\mu)>0$ and $\alpha(m)\in\Bbb R$ are two suitable constants. At first, for the more general semilinear Tricomi equation $\partial_t^2v-t^m\Delta v=t^{\alpha}|v|^p$ with any fixed constant $m>0$ and arbitrary parameter $\alpha\in\Bbb R$, we shall show that in the case of $\alpha\leq -2$, $n\geq 3$ and $p>1$, the small data weak solution $v$ exists globally; in the case of $\alpha>-2$, through determining the conformal exponent $p_{conf}(n,m,\alpha)>1$, the global small data weak solution $v$ exists when some extra restrictions of $p\geq p_{conf}(n,m,\alpha)$ are given. Returning to the original equation $\partial_t^2u-\Delta u+\frac{\mu}{t}\partial_tu=|u|^p$, the corresponding global existence results on the small data solution $u$ can be obtained., Comment: 32 pages, 2 figures

Social network plays an important role in propagating people's viewpoints, emotions, thoughts, and fears. Notably, following lockdown periods during the COVID-19 pandemic, the issue of depression has garnered increasing attention, with a significant portion of individuals resorting to social networks as an outlet for expressing emotions. Using deep learning techniques to discern potential signs of depression from social network messages facilitates the early identification of mental health conditions. Current efforts in detecting depression through social networks typically rely solely on analyzing the textual content, overlooking other potential information. In this work, we conduct a thorough investigation that unveils a strong correlation between depression and negative emotional states. The integration of such associations as external knowledge can provide valuable insights for detecting depression. Accordingly, we propose a multi-task training framework, DeSK, which utilizes shared sentiment knowledge to enhance the efficacy of depression detection. Experiments conducted on both Chinese and English datasets demonstrate the cross-lingual effectiveness of DeSK.

Electric field detection has been widely utilized in many fields such as scientific research and integrated circuits. To enhance the tuning sensitivity of E-field sensor, in this paper, we theoretically proposed a highly sensitive E-field sensor composed of etchless lithium niobate (LN) material and hybrid coupling-grating systems in the visible near-infrared regime. Such configuration supports high-quality factor quasi-BIC resonance, which generates strong localized field confinement. Due to the large electro-optic coefficient of LN material, one can shift the wavelength and reflection ratio of resonance by tuning the refractive index of LN material. An analytical theory is carried out to explain the relationship between the refractive index variations and the applied voltages, so we successfully obtained a tuning sensitivity of 40.8 nm/V and a minimum detectable electric field amplitude of 24.5 mV with wavelength resolution of 1 nm. Due to the low parasitic capacitance of LN material and high conductivity of gold film and ITO layer which are utilized as the electrodes, the 3dB bandwidth of the devices should exceed 154 GHz. And we believe that such a surface-normal E-field sensor has extensive potential for the extremely weak electric field detection., Comment: 10 pages, 5 figures

With a changing climate, the frequency and intensity of extreme weather events are likely to increase, posing a threat to infrastructure systems' resilience. The response of infrastructure systems to localised failures depends on whether assets are affected randomly, in a targeted strategic way, or any way in between. More than that, infrastructure decisions today, including new routes or improvements to existing assets, will underpin the behaviour of the systems over the next century. It is important to separate and analyse the case of climate-based disruptions and how they affect systems' resilience. This paper presents a probabilistic resilience assessment framework where failure scenarios and network disruptions are generated using weather profile data from climate prediction models with component-level fragility functions. A case study is then carried out to quantify the resilience of Great Britain's railway passenger transport system to high-temperature-related track buckling under the Representative Concentration Pathway 8.5 (RCP8.5) climate change scenario. A 95-year horizon on the resilience of the railway system is drawn. The results also reveal the non-linear responses of the railway system to the increasing temperature and show that models considering random asset failures overestimate the system's resilience.

Multiwalled carbon nanotubes (MWCNTs) are decorated with Pt nanoparticles by a "layer-by-layer" approach using poly (sodium 4-styrene sulfonate) (PSS) and poly (diallyl dimethylammonium chloride) (PDDA). Transmission electron microscopy (TEM) images and Energy Dispersive X-Ray (EDX) analysis of the samples confirm Pt deposition on surfaces of CNTs. Dispersibility and dispersion stability of MWCNTs in the solvents are enhanced when MWCNTs are coated with Pt nanoparticles. Mg AZ91 composites reinforced with MWCNTs are then produced by a melt stirring process. Compression tests of the composites show that adding 0.05\% wt Pt-coated MWCNTs in AZ91 improves the composite's mechanical properties compared to the pure AZ91 and pristine MWCNT/AZ91. Fracture surface analysis of the composite using a scanning electron microscope (SEM) shows individuals pulled out MWCNTs in the case of the Pt-coated MWCNT/AZ91 composites. We attribute this finding to the uniform dispersion of Pt-coated MWCNTs in Mg due to the improved wettability of Pt-coated MWCNTs in Mg melts. Molecular dynamics (MD) simulations of the interaction between Pt-coated MWCNTs and Mg support this interpretation., Comment: 20 pages,7 figures, one table