dimensionality
This term refers to the number of features or attributes in a dataset or the complexity of the space in which data points reside. High-dimensionality can create challenges in model training (the curse of dimensionality) and understanding relationships between variables.
- Degrees of Freedom for Linear Attention: Distilling Softmax Attention with Optimal Feature Efficiency
- H-SPLID: HSIC-based Saliency Preserving Latent Information Decomposition
- Infinite-Width Limit of a Single Attention Layer: Analysis via Tensor Programs
- Metropolis Adjusted Microcanonical Hamiltonian Monte Carlo
- PREAMBLE: Private and Efficient Aggregation via Block Sparse Vectors
- Scaling Laws for Gradient Descent and Sign Descent for Linear Bigram Models under Zipf’s Law