arXiv subjects

Hao-Yun Hsu

Publications and source records attributed to Hao-Yun Hsu.

3 recordsLinked to original sources

arXiv2025

Dynamics of Information Exchange in Zebrafish: The Role of U-Turns in Visual Communication and Behavior Modulation

Motions of visually coupled zebrafish pairs are studied to understand the effects of information exchange on their behavior as a function of their minimal separation ($d$). We find that when $d$ is small, the pair can display a leader-follower relation (LFR) with trajectories of almost synchronized form. However, with larger $d$, although the same LFR is still maintained, the originally similar trajectories turn into different forms. Detailed analysis of their motion trajectories suggests that the pair might be using U-turns (UTs) to exchange information and to maintain a LFR at the same time. A simulation model based on UTs with inferred and proposed rules is able to reproduce prominent features of observed trajectories; indicating that the transition of trajectories can be understood as the result of a change in information exchange between the fish as $d$ increases. Our finding that UTs as important visual signals is consistent with the fact that UTs can induce a large amount of firings in retinas of observing fish.

physics.bio-ph↗

arXiv2024

Ensemble architecture in polyp segmentation

This study explored the architecture of semantic segmentation and evaluated models that excel in polyp segmentation. We present an integrated framework that harnesses the advantages of different models to attain an optimal outcome. Specifically, in this framework, we fuse the learned features from convolutional and transformer models for prediction, thus engendering an ensemble technique to enhance model performance. Our experiments on polyp segmentation revealed that the proposed architecture surpassed other top models, exhibiting improved learning capacity and resilience. The code is available at https://github.com/HuangDLab/EnFormer.

cs.CV↗

arXiv2022

The R Package HCV for Hierarchical Clustering from Vertex-links

The HCV package implements the hierarchical clustering for spatial data. It requires clustering results not only homogeneous in non-geographical features among samples but also geographically close to each other within a cluster. We modified typically used hierarchical agglomerative clustering algorithms to introduce the spatial homogeneity, by considering geographical locations as vertices and converting spatial adjacency into whether a shared edge exists between a pair of vertices. The main function HCV obeying constraints of the vertex links automatically enforces the spatial contiguity property at each step of iterations. In addition, two methods to find an appropriate number of clusters and to report cluster members are also provided.

stat.CO↗