expressivity
Expressivity refers to the capacity of a model to represent a wide range of functions or behaviors, indicating its ability to learn complex patterns and relationships in data.
- A compressive-expressive communication framework for compositional representations
- Better NTK Conditioning: A Free Lunch from (ReLU) Nonlinear Activation in Wide Neural Networks
- DeltaFormer: Unlock the state space of Transformer
- Embeddings as Probabilistic Equivalence in Logic Programs
- Enhancing Visual Prompting through Expanded Transformation Space and Overfitting Mitigation
- Fixed-Point RNNs: Interpolating from Diagonal to Dense
- Generalizable Insights for Graph Transformers in Theory and Practice
- How to Learn a Star: Binary Classification with Starshaped Polyhedral Sets
- Improving Model Representation and Reducing KV Cache via Skip Connections with First Value Heads
- Mixture of Scope Experts at Test: Generalizing Deeper Graph Neural Networks with Shallow Variants
- Normalizing Flows are Capable Models for Continuous Control
- On Universality Classes of Equivariant Networks
- On topological descriptors for graph products
- PaTH Attention: Position Encoding via Accumulating Householder Transformations
- Prompt Tuning Transformers for Data Memorization
- Random Search Neural Networks for Efficient and Expressive Graph Learning
- Sound Logical Explanations for Mean Aggregation Graph Neural Networks