low-dimensional subspace
A reduced-dimensional representation of data that retains essential features while removing redundancy and noise, enhancing the efficiency and effectiveness of AI algorithms in tasks like classification and clustering.
- Better Language Model Inversion by Compactly Representing Next-Token Distributions
- Does Object Binding Naturally Emerge in Large Pretrained Vision Transformers?
- Multiresolution Analysis and Statistical Thresholding on Dynamic Networks
- RefLoRA: Refactored Low-Rank Adaptation for Efficient Fine-Tuning of Large Models
- Shallow Diffuse: Robust and Invisible Watermarking through Low-Dim Subspaces in Diffusion Models
- Subspace Networks: Scaling Decentralized Training with Communication-Efficient Model Parallelism