each tree trains on different data 2. **Feature randomization** --- each split considers different features