Implicit Bias of Gradient Descent for Two-layer ReLU and Leaky ReLU Networks on Nearly-orthogonal Data.
Yiwen KouZixiang ChenQuanquan GuPublished in: NeurIPS (2023)
Keyphrases
- data sets
- data analysis
- experimental data
- original data
- raw data
- complex data
- database
- data distribution
- statistical analysis
- data processing
- image data
- xml documents
- training data
- data points
- knowledge discovery
- prior knowledge
- data collection
- social networks
- data objects
- cost function
- feature space
- high dimensional data
- synthetic data
- high quality
- feature selection
- noisy data
- search engine