Low Dimensional Trajectory Hypothesis is True: DNNs Can Be Trained in Tiny Subspaces.
Tao LiLei TanZhehao HuangQinghua TaoYipeng LiuXiaolin HuangPublished in: IEEE Trans. Pattern Anal. Mach. Intell. (2023)
Keyphrases
- low dimensional
- high dimensional
- high dimensional data
- dimensionality reduction
- linear subspace
- principal component analysis
- lower dimensional
- data points
- manifold learning
- dimension reduction
- feature space
- euclidean space
- nonlinear dimensionality reduction
- input space
- high dimensional data space
- multidimensional scaling
- manifold structure
- subspace learning
- training set
- multilayer perceptron
- locally linear embedding
- image processing
- clustering algorithm
- linear dimensionality reduction