1: \begin{abstract}

2: We investigate and compare the fundamental performance of several distributed 

3: learning methods that have 

4: been proposed recently. 

5: We do this in the context of a  distributed version of the classical 

6: signal-in-Gaussian-white-noise model, which  serves as a benchmark model for 

7: studying performance in this setting.

8: The results show how the design and tuning of a distributed method

9:  can have great impact on convergence rates and validity of uncertainty quantification.

10:   Moreover, we highlight the difficulty of  designing 

11:  nonparametric distributed procedures that automatically adapt to smoothness.

12: \end{abstract}

13: