representational power
Representational power refers to the ability of a model to capture and represent complex relationships and structures within data, which is crucial for effective learning and generalization.
- CAT: Circular-Convolutional Attention for Sub-Quadratic Transformers
- FlowFeat: Pixel-Dense Embedding of Motion Profiles
- Higher-Order Learning with Graph Neural Networks via Hypergraph Encodings
- Permutation Equivariant Neural Controlled Differential Equations for Dynamic Graph Representation Learning
- When Do Transformers Outperform Feedforward and Recurrent Networks? A Statistical Perspective