Explainable GeoAI: can saliency maps help interpret artificial intelligence’s learning process? An empirical study on natural feature detection

Hsu, Chia‐Yu; Li, Wenwen

doi:10.1080/13658816.2023.2191256

Cited by 19 publications

(2 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…O Desvio Padrão (DP) foi empregado para indicar a variabilidade dos valores no conjunto de dados utilizado como referência para a análise comparativa entre as CNNs [Sokolova and Lapalme 2009]. Além disso, a explicabilidade (ou interpretabilidade) é outra técnica relevante adotada [Hsu and Li 2023]. Também consideramos a utilizac ¸ão da estratégia de Aprendizagem de Conjunto, também conhecida como Comitê ou Ensemble Learning em inglês, para combinar as previsões das CNNs que apresentaram os melhores resultados nos experimentos, a fim de gerar uma única previsão final [Moon et al 2020].…”

Section: Métricas De Avaliac ¸ãO E Ambiente De Experimentosunclassified

Explorando Arquiteturas de Redes Neurais Profundas na Classificação de Imagens de Cariótipos Humanos

Imperes Filho,

Machado,

Monte

et al. 2024

Anais Do XXIV Simpósio Brasileiro De Computação Aplicada À Saúde (SBCAS 2024)

View full text Add to dashboard Cite

A análise cromossômica, uma prática clinicamente crucial realizada tradicionalmente por geneticistas, pode ser suscetível à fadiga ao longo do tempo, afetando a qualidade dos diagnósticos. Neste artigo, exploramos a classificação automatizada de imagens de cromossomos por meio de diversas arquiteturas de redes neurais profundas. Avaliamos 23 pares de cromossomos humanos em uma tarefa multiclasse, revelando resultados promissores. Destacam-se os desempenhos superiores da arquitetura DenseNet169, alcançando uma acurácia, precisão, recall e F1-Score de 98,77%. O índice de concordância Kappa atingiu um nível ”Excelente”(0,99), enquanto um baixo desvio padrão (0,002) ressaltou a consistência das métricas, conferindo confiabilidade e previsibilidade ao modelo proposto.

show abstract

Section: Métricas De Avaliac ¸ãO E Ambiente De Experimentosunclassified

Explorando Arquiteturas de Redes Neurais Profundas na Classificação de Imagens de Cariótipos Humanos

Imperes Filho,

Machado,

Monte

et al. 2024

Anais Do XXIV Simpósio Brasileiro De Computação Aplicada À Saúde (SBCAS 2024)

View full text Add to dashboard Cite

show abstract

“…In addition, visualising the feature decision-making process also helps to determine which input features are more significant to the model predictions. Initial attempts to observe and explain the process were made by visualising various model components, such as filters and feature maps; however, the obtained information was difficult to map to meaningful geomorphological concepts (Hsu & Li, 2023). Hence, methods that can explain each decision via generating saliency maps (firstly compute gradients in the intermediate convolutional layer and then combine weight feature maps from different channels) have become popular, such as class activation maps (CAMs) (Zhou et al, 2016).…”

Section: Introductionmentioning

confidence: 99%

Intelligent classification and analysis of regional landforms based on automatic feature selection

Xu,

Zhu,

et al. 2023

Earth Surf Processes Landf

View full text Add to dashboard Cite

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

show abstract

Enhancing Explainability in Mobility Data Science Through a Combination of Methods

Makridis,

Koukos,

Fatouros

et al. 2024

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

Explainable GeoAI: can saliency maps help interpret artificial intelligence’s learning process? An empirical study on natural feature detection

Cited by 19 publications

References 30 publications

Explorando Arquiteturas de Redes Neurais Profundas na Classificação de Imagens de Cariótipos Humanos

Explorando Arquiteturas de Redes Neurais Profundas na Classificação de Imagens de Cariótipos Humanos

Intelligent classification and analysis of regional landforms based on automatic feature selection

Enhancing Explainability in Mobility Data Science Through a Combination of Methods

Contact Info

Product

Resources

About