Summary of emerging machine learning and deep learning architectures, including their core concepts and key advantages
| Architecture | Core concept | Key advantages |
|---|---|---|
| Transformers / attention | Uses self-attention mechanisms to capture long-range dependencies and global context within data | Effective for modeling complex sequential and spatial relationships; scalable and flexible across data modalities |
| Graph neural networks (GNNs) | Processes data represented as graphs, where nodes and edges encode relationships among entities | Naturally handles irregular and non-Euclidean data while capturing complex relational structures |
| Physics-Informed neural networks (PINNs) | Incorporates governing physical laws and constraints into the learning process through the loss function | Improves physical consistency, interpretability and generalization, particularly when data are limited |
| Operator learning methods (DeepONet, FNO) | Learns mappings between function spaces rather than discrete input-output pairs | Enables rapid approximation of complex physical systems and efficient surrogate modeling |
| Explainable AI (XAI) | Provides methods to interpret and explain model predictions and decision-making processes | Improves transparency, trustworthiness and validation of machine learning models |
| Architecture | Core concept | Key advantages |
|---|---|---|
| Transformers / attention | Uses self-attention mechanisms to capture long-range dependencies and global context within data | Effective for modeling complex sequential and spatial relationships; scalable and flexible across data modalities |
| Graph neural networks (GNNs) | Processes data represented as graphs, where nodes and edges encode relationships among entities | Naturally handles irregular and non-Euclidean data while capturing complex relational structures |
| Physics-Informed neural networks (PINNs) | Incorporates governing physical laws and constraints into the learning process through the loss function | Improves physical consistency, interpretability and generalization, particularly when data are limited |
| Operator learning methods (DeepONet, | Learns mappings between function spaces rather than discrete input-output pairs | Enables rapid approximation of complex physical systems and efficient surrogate modeling |
| Explainable | Provides methods to interpret and explain model predictions and decision-making processes | Improves transparency, trustworthiness and validation of machine learning models |
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.