arXiv subjects

Cindy Zhang

Publications and source records attributed to Cindy Zhang.

9 recordsLinked to original sources

arXiv2026

Exotic Surfaces in 4-manifolds and Surface Corks

A fundamental result in 4-manifold topology asserts that any two exotic smooth structures on a simply-connected, closed 4-manifold differ by a cork twist: the operation of removing a compact, contractible, codimension-zero submanifold and regluing it by a diffeomorphism of its boundary. In this paper, we introduce the notion of a surface cork, an analogous object in the setting of smoothly embedded, closed surfaces $F$ in closed 4-manifolds $X$. This is a compact, contractible, codimension-zero submanifold intersecting $F$ in a controllable manner, whose removal and regluing via a diffeomorphism of its boundary changes the diffeomorphism type of $(X, F)$ as a pair while leaving its homeomorphism type unchanged. The way in which the surface $F$ interacts with the codimension-zero submanifold leads us to define three distinct notions of surface corks: enclosing surface corks, exterior surface corks, and transverse surface corks. We establish the existence of exterior surface corks for certain previously known examples of exotic pairs. Furthermore, we give the first explicit construction of a transverse surface cork for certain exotic families arising from Fintushel--Stern rim surgery. Notably, this transverse surface cork turns out to be diffeomorphic to a 4-ball.

math.GT↗

arXiv2025

Parsing Through Boundaries in Chinese Word Segmentation

Chinese word segmentation is a foundational task in natural language processing (NLP), with far-reaching effects on syntactic analysis. Unlike alphabetic languages like English, Chinese lacks explicit word boundaries, making segmentation both necessary and inherently ambiguous. This study highlights the intricate relationship between word segmentation and syntactic parsing, providing a clearer understanding of how different segmentation strategies shape dependency structures in Chinese. Focusing on the Chinese GSD treebank, we analyze multiple word boundary schemes, each reflecting distinct linguistic and computational assumptions, and examine how they influence the resulting syntactic structures. To support detailed comparison, we introduce an interactive web-based visualization tool that displays parsing outcomes across segmentation methods.

cs.CL↗

arXiv2025

Cross-Attention Fusion of MRI and Jacobian Maps for Alzheimer's Disease Diagnosis

Early diagnosis of Alzheimer's disease (AD) is critical for intervention before irreversible neurodegeneration occurs. Structural MRI (sMRI) is widely used for AD diagnosis, but conventional deep learning approaches primarily rely on intensity-based features, which require large datasets to capture subtle structural changes. Jacobian determinant maps (JSM) provide complementary information by encoding localized brain deformations, yet existing multimodal fusion strategies fail to fully integrate these features with sMRI. We propose a cross-attention fusion framework to model the intrinsic relationship between sMRI intensity and JSM-derived deformations for AD classification. Using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, we compare cross-attention, pairwise self-attention, and bottleneck attention with four pre-trained 3D image encoders. Cross-attention fusion achieves superior performance, with mean ROC-AUC scores of 0.903 (+/-0.033) for AD vs. cognitively normal (CN) and 0.692 (+/-0.061) for mild cognitive impairment (MCI) vs. CN. Despite its strong performance, our model remains highly efficient, with only 1.56 million parameters--over 40 times fewer than ResNet-34 (63M) and Swin UNETR (61.98M). These findings demonstrate the potential of cross-attention fusion for improving AD diagnosis while maintaining computational efficiency.

eess.IV↗

arXiv2025

OPTIC: Optimizing Patient-Provider Triaging & Improving Communications in Clinical Operations using GPT-4 Data Labeling and Model Distillation

The COVID-19 pandemic has accelerated the adoption of telemedicine and patient messaging through electronic medical portals (patient medical advice requests, or PMARs). While these platforms enhance patient access to healthcare, they have also increased the burden on healthcare providers due to the surge in PMARs. This study seeks to develop an efficient tool for message triaging to reduce physician workload and improve patient-provider communication. We developed OPTIC (Optimizing Patient-Provider Triaging & Improving Communications in Clinical Operations), a powerful message triaging tool that utilizes GPT-4 for data labeling and BERT for model distillation. The study used a dataset of 405,487 patient messaging encounters from Johns Hopkins Medicine between January and June 2020. High-quality labeled data was generated through GPT-4-based prompt engineering, which was then used to train a BERT model to classify messages as "Admin" or "Clinical." The BERT model achieved 88.85% accuracy on the test set validated by GPT-4 labeling, with a sensitivity of 88.29%, specificity of 89.38%, and an F1 score of 0.8842. BERTopic analysis identified 81 distinct topics within the test data, with over 80% accuracy in classifying 58 topics. The system was successfully deployed through Epic's Nebula Cloud Platform, demonstrating its practical effectiveness in healthcare settings.

cs.LG↗

arXiv2024

VEPerform: a web resource for evaluating the performance of variant effect predictors

Computational variant effect predictors (VEPs) are providing increasingly strong evidence to classify the pathogenicity of missense variants. Precision vs. recall analysis is useful in evaluating VEP performance, especially when adjusted for imbalanced test sets. Here, we describe VEPerform, a web-based tool for evaluating the performance of VEPs at the gene level using balanced precision vs. recall curve (BPRC) analysis.

q-bio.GN↗

arXiv2024

Fixed-point Grover Adaptive Search for Quadratic Binary Optimization Problems

We study a Grover-type method for Quadratic Unconstrained Binary Optimization (QUBO) problems. For an $n$-dimensional QUBO problem with $m$ nonzero terms, we construct a marker oracle for such problems with a tuneable parameter, $Λ\in \left[ 1, m \right] \cap \mathbb{Z}$. At $d \in \mathbb{Z}_+$ precision, the oracle uses $O (n + Λd)$ qubits, has total depth of $O \left( \tfrac{m}Λ \log_2 (n) + \log_2 (d) \right)$, and non-Clifford depth of $O \left( \tfrac{m}Λ \right)$. Moreover, each qubit required to be connected to at most $O \left( \log_2 (Λ+ d) \right)$ other qubits. In the case of a maximum graph cuts, as $d = 2 \left\lceil \log_2 (n) \right\rceil$ always suffices, the depth of the marker oracle can be made as shallow as $O (\log_2 (n))$. For all values of $Λ$, the non-Clifford gate count of these oracles is strictly lower (at least by a factor of $\sim 2$) than previous constructions. Furthermore, we introduce a novel \textit{Fixed-point Grover Adaptive Search for QUBO Problems}, using our oracle design and a hybrid Fixed-point Grover Search, motivated by the works of Boyer et al. and Li et al. This method has better performance guarantees than previous Grover Adaptive Search methods. Some of our results are novel and useful for any method based on Fixed-point Grover Search. Finally, we give a heuristic argument that, with high probability and in $O \left( \tfrac{\log_2 (n)}{\sqrtε} \right)$ time, this adaptive method finds a configuration that is among the best $ε2^n$ ones.

quant-ph↗

arXiv2024

Novel oracle constructions for quantum random access memory

We present new designs for quantum random access memory. More precisely, for each function, $f : \mathbb{F}_2^n \rightarrow \mathbb{F}_2^d$, we construct oracles, $\mathcal{O}_f$, with the property \begin{equation} \mathcal{O}_f \left| x \right\rangle_n \left| 0 \right\rangle_d = \left| x \right\rangle_n \left| f(x) \right\rangle_d. \end{equation} Our methods are based on the Walsh-Hadamard Transform of $f$, viewed as an integer valued function. In general, the complexity of our method scales with the sparsity of the Walsh-Hadamard Transform and not the sparsity of $f$, yielding more favorable constructions in cases such as binary optimization problems and function with low-degree Walsh-Hadamard Transforms. Furthermore, our design comes with a tuneable amount of ancillas that can trade depth for size. In the ancilla-free design, these oracles can be $ε$-approximated so that the Clifford + $T$ depth is $O \left( \left( n + \log_2 \left( \tfrac{d}ε \right) \right) \mathcal{W}_f \right)$, where $\mathcal{W}_f$ is the number of nonzero components in the Walsh-Hadamard Transform. The depth of the shallowest version is $O \left( n + \log_2 \left( \tfrac{d}ε \right) \right)$, using $n + d \mathcal{W}_f$ qubit. The connectivity of these circuits is also only logarithmic in $\mathcal{W}_f$. As an application, we show that for boolean functions with low approximate degrees (as in the case of read-once formulas) the complexities of the corresponding QRAM oracles scale only as $2^{\widetilde{O} \left( \sqrt{n} \log_2 \left( n \right) \right)}$.

quant-ph↗

arXiv2022

Bounds for Kirby-Thompson invariants of knotted surfaces

We provide sharp lower bounds for two versions of the Kirby-Thompson invariants for knotted surfaces, one of which was originally defined by Blair, Campisi, Taylor, and Tomova. The second version introduced in this paper measures distances in the dual curve complex instead of the pants complex. We compute the exact values of both KT-invariants for infinitely many knotted surfaces with bridge number at most six.

math.GT↗

arXiv2021

Bounding the Kirby-Thompson invariant of spun knots

A bridge trisection of a smooth surface in $S^4$ is a decomposition analogous to a bridge splitting of a link in $S^3$. The Kirby-Thompson invariant of a bridge trisection measures its complexity in terms of distances between disc sets in the pants complex of the trisection surface. We give the first significant bounds for the Kirby-Thompson invariant of spun knots. In particular, we show that the Kirby-Thompson invariant of the spun trefoil is 15.

math.GT↗