1: \begin{abstract} 

2: We consider the prediction of weak effects in a multiple-output regression setup, when covariates are expected

3: to explain a small amount, less than $\approx 1\%$, of the variance of the target variables. To facilitate the

4: prediction of the weak effects, we constrain our model structure by introducing a novel Bayesian approach of

5: sharing information between the regression model and the noise model. Further reduction of the effective

6: number of parameters is achieved by introducing an infinite shrinkage prior and group sparsity in the context

7: of the Bayesian reduced rank regression, and using the Bayesian infinite factor model as a flexible low-rank

8: noise model. In our experiments the model incorporating the novelties outperformed alternatives in genomic

9: prediction of rich phenotype data. In particular, the information sharing between the noise and regression

10: models led to significant improvement in prediction accuracy.%

11: %We introduce a new method for modeling weak effects in a multiple-output regression setup, i.e. when covariates explain $\approx 1\%$ of the variance of the target variables. 

12: %With weak effects we refer to prediction tasks where covariates explain $\approx 1\%$ of the variance of the target variables.

13: %

14: %An infinite shrinkage prior and group sparsity are for the first time presented in the context of Bayesian reduced regression. A Bayesian infinite factor model is used as a flexible low-rank noise model in our model, to account for any latent confounding factors.

15: %

16: %Further reduction of the effective number of parameters is needed and this is achieved by introducing a conceptially new approach of sharing information between the regression model and the noise model.

17: %

18: %In our experiment the method outperformed comparison methods in genomic prediction of rich phenotype data. The new principle is shown to  statistically significantly improve predictive performance.

19: \end{abstract}