arXiv subjects

Shanshan Yu

Publications and source records attributed to Shanshan Yu.

11 recordsLinked to original sources

arXiv2026

Fiber Nonlinearity Compensation of Coherent Signals Using Deep Photonic Reservoir Computer

Photonic reservoir computer (PRC) is a promising optical computing framework for high-speed optical signal processing, and various reports have shown its functionality of linear equalization for intensity-modulation direct-detection communication links. However, coherent communication links suffer more from nonlinear impairment, whereas its nonlinear equalization is very challenging. Here we demonstrate the nonlinear equalization of coherent 16-level quadrature amplitude modulation (16-QAM) signals using a deep PRC in experiment. The deep PRC consists of cascading injection-locked semiconductor lasers with optical feedback loops. For 16-QAM signals with a transmission rate of 240 Gbps and a launch power of 12 dBm, the single-channel PRC with 3 hidden layers raises the Q factor by as high as 0.58 dB, while the dual-channel PRC with 2 hidden layers raises the Q factor by 0.55 dB. In addition, we prove that the deep PRC is able to equalize 16-QAM signals of different launch powers and transmission distances.

physics.optics↗

arXiv2026

Arc-disjoint Steiner Cycles in Digraphs

Let $D=(V(D), A(D))$ be a digraph of order $n$ and let $S\subseteq V(D)$ with $2\leq |S|\leq n$. A directed cycle $C$ of $D$ is called a directed $S$-Steiner cycle (or, an $S$-cycle for short) if $S\subseteq V(C)$. Steiner cycles have applications in reliable designs for telecommunication and transportation networks. Two $S$-cycles are called arc-disjoint if they have no common arcs. We use $\lambda_{S}^{c}(D)$ to denote the maximum number of pairwise arc-disjoint $S$-cycles in $D$. The directed cycle $k$-arc-connectivity of $D$ is defined as $$\lambda_{k}^{c} (D)=\min\left \{ \lambda _{S}^{c}(D)\mid S\subseteq V(D),\left | S \right | =k,2\le k\le n \right \}.$$ In this paper, we determine the complexity for $\lambda_{S}^{c} (D)$ on Eulerian digraphs, planar digraphs and symmetric digraphs. We also obtain exact values of $\lambda_{k}^{c} (D)$ on complete digraphs, complete bipartite digraphs and complete regular multipartite digraphs.

math.CO↗

arXiv2026

Internally-disjoint directed pendant Steiner trees with three terminal vertices in Cartesian product digraphs

Let $D=(V(D),A(D))$ be a digraph with a terminal vertex subset $S\subseteq V(D)$ such that $|S|=k\geq 2$. An out-tree $T$ of $D$ rooted at $r$ is called a directed pendant $(S,r)$-Steiner tree (or, pendant $(S,r)$-tree for short) if $r\in S\subseteq V(T)$ and $d_{T}^{+}(r)=d_{T}^{-}(u)=1$ for each $u\in S\backslash \{r\}$. Two pendant $(S,r)$-trees $T_{1}$ and $T_{2}$ are internally-disjoint if $A(T_{1})\cap A(T_{2})=\varnothing$ and $V(T_{1})\cap V(T_{2})=S$. The pendant-tree $k$-connectivity $\tau_{k}(D)$ of $D$ is defined as $$\tau_{k}(D)=\min\{\tau_{S,r}(D)\mid S\subseteq V(D),|S|=k,r\in S\},$$ where $\tau_{S,r}(D)$ denotes the maximum number of pairwise internally-disjoint pendant $(S,r)$-trees in $D$. In this paper, we derive a sharp lower bound for the pendant-tree 3-connectivity of the Cartesian product digraph $D\square H$, where $D$ and $H$ are both strong digraphs. Specifically, we prove the lower bound $\tau_{3}(D\square H)\geq \tau_{3}(D)+\tau_{3}(H)$. Moreover, we propose a polynomial-time algorithm for finding internally-disjoint pendant $(S,r)$-trees which attain this lower bound.

math.CO↗

arXiv2026

Schr\"oMind: Mitigating Hallucinations in Multimodal Large Language Models via Solving the Schr\"odinger Bridge Problem

Recent advancements in Multimodal Large Language Models (MLLMs) have achieved significant success across various domains. However, their use in high-stakes fields like healthcare remains limited due to persistent hallucinations, where generated text contradicts or ignores visual input. We contend that MLLMs can comprehend images but struggle to produce accurate token sequences. Minor perturbations can shift attention from truthful to untruthful states, and the autoregressive nature of text generation often prevents error correction. To address this, we propose Schr\"oMind-a novel framework reducing hallucinations via solving the Schr\"odinger bridge problem. It establishes a token-level mapping between hallucinatory and truthful activations with minimal transport cost through lightweight training, while preserving the model's original capabilities. Extensive experiments on the POPE and MME benchmarks demonstrate the superiority of Schr\"odinger, which achieves state-of-the-art performance while introducing only minimal computational overhead.

cs.CV↗

arXiv2026

Scalpel: Fine-Grained Alignment of Attention Activation Manifolds via Mixture Gaussian Bridges to Mitigate Multimodal Hallucination

Rapid progress in large vision-language models (LVLMs) has achieved unprecedented performance in vision-language tasks. However, due to the strong prior of large language models (LLMs) and misaligned attention across modalities, LVLMs often generate outputs inconsistent with visual content - termed hallucination. To address this, we propose \textbf{Scalpel}, a method that reduces hallucination by refining attention activation distributions toward more credible regions. Scalpel predicts trusted attention directions for each head in Transformer layers during inference and adjusts activations accordingly. It employs a Gaussian mixture model to capture multi-peak distributions of attention in trust and hallucination manifolds, and uses entropic optimal transport (equivalent to Schr\"odinger bridge problem) to map Gaussian components precisely. During mitigation, Scalpel dynamically adjusts intervention strength and direction based on component membership and mapping relationships between hallucination and trust activations. Extensive experiments across multiple datasets and benchmarks demonstrate that Scalpel effectively mitigates hallucinations, outperforming previous methods and achieving state-of-the-art performance. Moreover, Scalpel is model- and data-agnostic, requiring no additional computation, only a single decoding step.

cs.CV↗

arXiv2025

Internally-disjoint Pendant Steiner Trees in Digraphs

For a digraph $D=(V(D),A(D))$ and a set $S\subseteq V(D)$ with $|S|\geq 2$ and $r\in S$, a directed pendant $(S,r)$-Steiner tree (or, simply, a pendant $(S,r)$-tree) is an out-tree $T$ rooted at $r$ such that $S\subseteq V(T)$ and each vertex of $S$ has degree one in $T$. Two pendant $(S,r)$-trees are called internally-disjoint if they are arc-disjoint and their common vertex set is exactly $S$. The goal of the {\sc Internally-disjoint Directed Pendant Steiner Tree Packing (IDPSTP)} problem is to find a largest collection of pairwise internally-disjoint pendant $(S,r)$-trees in $D$. Let $\tau_{k}(D)=\min\{\tau_{S,r}(D)\mid S\subseteq V(D),|S|=k,r\in S\}$, where $\tau_{S,r}(D)$ denotes the maximum number of pairwise internally-disjoint pendant $(S,r)$-trees in $D$. In this paper, we first completely determine the computational complexity for the decision version of IDPSTP on Eulerian digraphs and symmetric digraphs. We then show that, for any $\epsilon>0$, given an instance of IDPSTP with order $n$, it is NP-hard to approximate the solution within $O(n^{{1/3}-\epsilon})$. Finally, we get some sharp bounds for the parameter $\tau_{k}(D)$.

math.CO↗

arXiv2025

A Benchmark for Vision-Centric HD Mapping by V2I Systems

Autonomous driving faces safety challenges due to a lack of global perspective and the semantic information of vectorized high-definition (HD) maps. Information from roadside cameras can greatly expand the map perception range through vehicle-to-infrastructure (V2I) communications. However, there is still no dataset from the real world available for the study on map vectorization onboard under the scenario of vehicle-infrastructure cooperation. To prosper the research on online HD mapping for Vehicle-Infrastructure Cooperative Autonomous Driving (VICAD), we release a real-world dataset, which contains collaborative camera frames from both vehicles and roadside infrastructures, and provides human annotations of HD map elements. We also present an end-to-end neural framework (i.e., V2I-HD) leveraging vision-centric V2I systems to construct vectorized maps. To reduce computation costs and further deploy V2I-HD on autonomous vehicles, we introduce a directionally decoupled self-attention mechanism to V2I-HD. Extensive experiments show that V2I-HD has superior performance in real-time inference speed, as tested by our real-world dataset. Abundant qualitative results also demonstrate stable and robust map construction quality with low cost in complex and various driving scenes. As a benchmark, both source codes and the dataset have been released at OneDrive for the purpose of further study.

cs.CV↗

arXiv2021

Multi-Relation Aware Temporal Interaction Network Embedding

Temporal interaction networks are formed in many fields, e.g., e-commerce, online education, and social network service. Temporal interaction network embedding can effectively mine the information in temporal interaction networks, which is of great significance to the above fields. Usually, the occurrence of an interaction affects not only the nodes directly involved in the interaction (interacting nodes), but also the neighbor nodes of interacting nodes. However, existing temporal interaction network embedding methods only use historical interaction relations to mine neighbor nodes, ignoring other relation types. In this paper, we propose a multi-relation aware temporal interaction network embedding method (MRATE). Based on historical interactions, MRATE mines historical interaction relations, common interaction relations, and interaction sequence similarity relations to obtain the neighbor based embeddings of interacting nodes. The hierarchical multi-relation aware aggregation method in MRATE first employs graph attention networks (GATs) to aggregate the interaction impacts propagated through a same relation type and then combines the aggregated interaction impacts from multiple relation types through the self-attention mechanism. Experiments are conducted on three public temporal interaction network datasets, and the experimental results show the effectiveness of MRATE.

cs.LG↗

arXiv2021

Multi-Level Visual Similarity Based Personalized Tourist Attraction Recommendation Using Geo-Tagged Photos

Geo-tagged photo based tourist attraction recommendation can discover users' travel preferences from their taken photos, so as to recommend suitable tourist attractions to them. However, existing visual content based methods cannot fully exploit the user and tourist attraction information of photos to extract visual features, and do not differentiate the significances of different photos. In this paper, we propose multi-level visual similarity based personalized tourist attraction recommendation using geo-tagged photos (MEAL). MEAL utilizes the visual contents of photos and interaction behavior data to obtain the final embeddings of users and tourist attractions, which are then used to predict the visit probabilities. Specifically, by crossing the user and tourist attraction information of photos, we define four visual similarity levels and introduce a corresponding quintuplet loss to embed the visual contents of photos. In addition, to capture the significances of different photos, we exploit the self-attention mechanism to obtain the visual representations of users and tourist attractions. We conducted experiments on a dataset crawled from Flickr, and the experimental results proved the advantage of this method.

cs.MM↗

arXiv2014

Herschel observations of EXtra-Ordinary Sources: Analysis of the HIFI 1.2 THz Wide Spectral Survey Toward Orion KL I. Methods

We present a comprehensive analysis of a broad band spectral line survey of the Orion Kleinmann-Low nebula (Orion KL), one of the most chemically rich regions in the Galaxy, using the HIFI instrument on board the Herschel Space Observatory. This survey spans a frequency range from 480 to 1907 GHz at a resolution of 1.1 MHz. These observations thus encompass the largest spectral coverage ever obtained toward this high-mass star-forming region in the sub-mm with high spectral resolution, and include frequencies $>$ 1 THz where the Earth's atmosphere prevents observations from the ground. In all, we detect emission from 39 molecules (79 isotopologues). Combining this dataset with ground based mm spectroscopy obtained with the IRAM 30 m telescope, we model the molecular emission from the mm to the far-IR using the XCLASS program which assumes local thermodynamic equilibrium (LTE). Several molecules are also modeled with the MADEX non-LTE code. Because of the wide frequency coverage, our models are constrained by transitions over an unprecedented range in excitation energy. A reduced $\chi^{2}$ analysis indicates that models for most species reproduce the observed emission well. In particular, most complex organics are well fit by LTE implying gas densities are high ($>$10$^6$ cm$^{-3}$) and excitation temperatures and column densities are well constrained. Molecular abundances are computed using H$_{2}$ column densities also derived from the HIFI survey. The distribution of rotation temperatures, $T_{\rm rot}$, for molecules detected toward the hot core is significantly wider than the compact ridge, plateau, and extended ridge $T_{\rm rot}$ distributions, indicating the hot core has the most complex thermal structure.

astro-ph.GA↗

arXiv2014

Herschel observations of EXtraordinary Sources: Analysis of the full Herschel/HIFI molecular line survey of Sagittarius B2(N)

A sensitive broadband molecular line survey of the Sagittarius B2(N) star-forming region has been obtained with the HIFI instrument on the Herschel Space Observatory, offering the first high-spectral resolution look at this well-studied source in a wavelength region largely inaccessible from the ground (625-157 um). From the roughly 8,000 spectral features in the survey, a total of 72 isotopologues arising from 44 different molecules have been identified, ranging from light hydrides to complex organics, and arising from a variety of environments from cold and diffuse to hot and dense gas. We present an LTE model to the spectral signatures of each molecule, constraining the source sizes for hot core species with complementary SMA interferometric observations, and assuming that molecules with related functional group composition are cospatial. For each molecule, a single model is given to fit all of the emission and absorption features of that species across the entire 480-1910 GHz spectral range, accounting for multiple temperature and velocity components when needed to describe the spectrum. As with other HIFI surveys toward massive star forming regions, methanol is found to contribute more integrated line intensity to the spectrum than any other species. We discuss the molecular abundances derived for the hot core, where the local thermodynamic equilibrium approximation is generally found to describe the spectrum well, in comparison to abundances derived for the same molecules in the Orion KL region from a similar HIFI survey.

astro-ph.GA↗