Curse or Redemption? How Data Heterogeneity Affects the Robustness of Federated Learning.
Syed ZawadAhsan AliPin-Yu ChenAli AnwarYi ZhouNathalie BaracaldoYuan TianFeng YanPublished in: AAAI (2021)
Keyphrases
- data sets
- image data
- database
- high dimensional data
- data processing
- data sources
- learning algorithm
- background knowledge
- data quality
- prior knowledge
- raw data
- learning systems
- statistical analysis
- knowledge discovery
- training data
- high quality
- data structure
- probability distribution
- learning tasks
- original data
- missing data
- synthetic data
- reinforcement learning
- training examples
- computer systems
- data analysis
- distributed data
- supervised learning
- noisy data
- hidden variables
- highly heterogeneous
- experimental data
- information sources
- data collection
- online learning
- active learning
- database systems