Reply on RC2 DOI Creative Commons
Wenwen Li

Published: March 15, 2024

Abstract. The revolutionary advances of Artificial Intelligence (AI) in the past decade have brought transformative innovation across science and engineering disciplines. Also field Arctic science, we witnessed an increasing trend adoption AI, especially deep learning, to support analysis big data facilitate new discoveries. In this paper, provide a comprehensive review applications learning sea ice remote sensing domains, focusing on problems such as lead detection, thickness estimation, concentration, extent forecasting motion detection well type classification. addition discussing these applications, also summarize technological that customized solutions, including loss functions strategies better understand dynamics. To promote growth exciting interdisciplinary field, further explore several research areas where community can benefit from cutting-edge AI technology. These include improving multi-modal capabilities, enhancing model accuracy measuring prediction uncertainty, leveraging foundation models, deepening integration with physics-based models. We hope paper serve cornerstone progress using inspire field.

Language: Английский

Reasoning cartographic knowledge in deep learning-based map generalization with explainable AI DOI Creative Commons
Cheng Fu, Zhiyong Zhou, Yanan Xin

et al.

International Journal of Geographical Information Science, Journal Year: 2024, Volume and Issue: 38(10), P. 2061 - 2082

Published: June 20, 2024

Cartographic map generalization involves complex rules, and a full automation has still not been achieved, despite many efforts over the past few decades. Pioneering studies show that some tasks can be partially automated by deep neural networks (DNNs). However, DNNs are used as black-box models in previous studies. We argue integrating explainable AI (XAI) into DL-based process give more insights to develop refine understanding what cartographic knowledge exactly is learned. Following an XAI framework for empirical case study, visual analytics quantitative experiments were applied explain importance of input features regarding prediction pre-trained ResU-Net model. This experimental study finds XAI-based visualization results easily interpreted human experts. With proposed workflow, we further find DNN pays attention building boundaries than interior parts buildings. thus suggest boundary intersection union better evaluation metric commonly qualifying raster-based results. Overall, this shows necessity feasibility part future development frameworks.

Language: Английский

Citations

3
Advancing Arctic Sea Ice Remote Sensing with AI and Deep Learning: Opportunities and Challenges DOI Creative Commons
Wenwen Li, Chia-Yu Hsu, Marco Tedesco

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(20), P. 3764 - 3764

Published: Oct. 10, 2024

Revolutionary advances in artificial intelligence (AI) the past decade have brought transformative innovation across science and engineering disciplines. In field of Arctic science, we witnessed an increasing trend adoption AI, especially deep learning, to support analysis big data facilitate new discoveries. this paper, provide a comprehensive review applications learning sea ice remote sensing domains, focusing on problems such as lead detection, thickness estimation, concentration extent forecasting, motion type classification. addition discussing these applications, also summarize technological that customized solutions, including loss functions strategies better understand dynamics. To promote growth exciting interdisciplinary field, further explore several research areas where community can benefit from cutting-edge AI technology. These include improving multimodal capabilities, enhancing model accuracy measuring prediction uncertainty, leveraging foundation models, deepening integration with physics-based models. We hope paper serve cornerstone progress using inspire field.

Language: Английский

Citations

3
Global or local modeling for XGBoost in geospatial studies upon simulated data and German COVID-19 infection forecasting DOI Creative Commons
Ximeng Cheng, Jackie Ma

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 14, 2025

Abstract Methods from artificial intelligence (AI) and, in particular, machine learning and deep learning, have advanced rapidly recent years been applied to multiple fields including geospatial analysis. Due the spatial heterogeneity fact that conventional methods can not mine large data, studies typically model homogeneous regions locally within entire study area. However, AI models process amounts of theoretically, more diverse train robust a well-trained will be. In this paper, we typical method XGBoost, with question: Is it better build single global or local for XGBoost studies? To compare modeling, is first studied on simulated data then also forecast daily infection cases COVID-19 Germany. The results indicate if under different relationships between independent dependent variables are balanced corresponding value ranges similar, i.e., low variation, modeling most cases; otherwise, stable better, especially secondary data. Besides, has potential using parallel computing because each sub-model trained independently, but partition requires extra attention affect results.

Language: Английский

Citations

0
Applicability evaluation of selected xAI methods for machine learning algorithms for signal parameters extraction DOI Open Access
Kalina Dimitrova, V. Kozhuharov, P. Petkov

et al.

Journal of Physics Conference Series, Journal Year: 2025, Volume and Issue: 3002(1), P. 012005 - 012005

Published: April 1, 2025

Abstract Machine learning methods find growing application in the reconstruction and analysis of data high energy physics experiments. A modified convolutional autoencoder model was employed to identify reconstruct pulses from scintillating crystals. The further investigated using four xAI for deeper understanding underlying mechanism. results are discussed detail, underlining importance knowledge gain improvement algorithms.

Language: Английский

Citations

0
Applying machine learning to understand water security and water access inequality in underserved colonia communities DOI Creative Commons
Zhining Gu, Wenwen Li, W. Michael Hanemann

et al.

Computers Environment and Urban Systems, Journal Year: 2023, Volume and Issue: 102, P. 101969 - 101969

Published: April 18, 2023

Language: Английский

Citations

7
Intelligent classification and analysis of regional landforms based on automatic feature selection DOI
Yuexue Xu, Hongchun Zhu, Zhiwei Lu

et al.

Earth Surface Processes and Landforms, Journal Year: 2023, Volume and Issue: 49(2), P. 787 - 803

Published: Nov. 1, 2023

Abstract Terrain features are an important basis for realizing high‐precision landform classification, and feature selection is a key step of machine learning knowledge mining. However, the process facing challenges due to multidimensionality correlation multisource terrain datasets factors. Traditional methods lack enough consideration interpretability transparency factors, but transparent decision‐making precisely determines modelling effect reliability model application results. Current research urgently needs work out black holes visual representation during selection. In intelligent multiple effective essential factor in enhancing performance generalisation ability network. Therefore, we initially selected 40 parameters, including basic factors digital elevation (DEM) textures, calculate contribution degree sort parameter importance based on SHapley Additive exPlanations (SHAP) method, then reserved 10%, 20%, 30%, 40% 50% turn constructing classification dataset. Because traditional UNet network cannot completely capture abrupt features, convolutional block attention module (CBAM) was integrated into UNet, deep established fine‐grained regional landforms. Considering calculation rate, even though there large spatial differences genetic mechanisms, it appropriate retain 20% classification. The accuracy typical regions, namely, Hanzhong Basin, North China Plain, Yunnan–Guizhou Plateau Tibetan Plateau, reached 98.76%, 97.36%, 96.3% 92.78%, respectively, what's more, some accuracies went up higher level under other combinations. Meanwhile, given different combinations corresponding types, combinative stability orderliness characteristics were explored explain variation trend.

Language: Английский

Citations

7
Assessment of noise pollution-prone areas using an explainable geospatial artificial intelligence approach DOI
Seyed Vahid Razavi-Termeh, Abolghasem Sadeghi‐Niaraki, Xiaobai Yao

et al.

Journal of Environmental Management, Journal Year: 2024, Volume and Issue: 370, P. 122361 - 122361

Published: Sept. 9, 2024

Language: Английский

Citations

2
Understanding of the predictability and uncertainty in population distributions empowered by visual analytics DOI Creative Commons
Peng Luo, Chuan Chen, Song Gao

et al.

International Journal of Geographical Information Science, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 31

Published: Nov. 19, 2024

Understanding the intricacies of fine-grained population distribution, including both predictability and uncertainty, is crucial for urban planning, social equity, environmental sustainability. The spatial processes associated with distribution populations are complex, enhancing their involves revealing nonlinear interactions among various explanatory variables. Additionally, influenced by factors that often challenging to quantify, thereby introducing uncertainty into predictive models. Although development explainable artificial intelligence (XAI) helps identify underlying factors, complex geographical special nature data present challenges purely statistical-based explanation methods, leading incomplete or incorrect explanations. To address these challenges, we introduce GeoVisX, a geospatial visual analytics framework integrated XAI. GeoVisX integrates XAI dissect processes. Through case study Munich, demonstrates its utility in analyzing identifying key impacting at 100 m grid level. Our findings highlight GeoVisX's capability enhance understanding phenomena, contributing more informed policy planning strategies. This not only validates effectiveness but also emphasizes importance incorporating methodologies addressing issues.

Language: Английский

Citations

2
Enhancing Explainability in Mobility Data Science Through a Combination of Methods DOI
Georgios Makridis, Vasileios Koukos, Georgios Fatouros

et al.

Lecture notes in networks and systems, Journal Year: 2024, Volume and Issue: unknown, P. 45 - 60

Published: Jan. 1, 2024

Language: Английский

Citations

1
Enhancing explainability of deep learning models for point cloud analysis: a focus on semantic segmentation DOI Creative Commons
Francesca Matrone, Marina Paolanti, Emanuele Frontoni

et al.

International Journal of Digital Earth, Journal Year: 2024, Volume and Issue: 17(1)

Published: Sept. 2, 2024

Semantic segmentation of point clouds plays a critical role in various applications, such as urban planning, infrastructure management, environmental analyses and autonomous navigation. Understanding the behaviour deep neural networks (DNNs) analysing cloud data is essential for improving accuracy developing effective network architectures acquisition strategies. In this paper, we investigate traits some state-of-the-art using indoor outdoor datasets. We compare PointNet, DGCNN, BAAF-Net on specifically selected datasets, including synthetic real-world environments. The chosen datasets are S3DIS, SynthCity, Semantic3D, KITTI. analyse impact different factors dataset type (synthetic vs. real), scene (indoor outdoor), system (static mobile sensors). Through detailed comparisons, provide insights into strengths limitations not only handling but also their structure. This study contributes to going beyond mere unconditional use AI algorithms, trying explain DNNs analysis paving way future research enhance develop possible guidelines both design geomatics field.

Language: Английский

Citations

1