Predict multiple outcomes

[talegari]

One usecase where I reach out to NN is when I require to predict multiple outcomes. This means y should be allowed to be a matrix ($d \times n$) where $d$ is the number of outcomes and $n$ is the number of rows of X (dataframe of predictors). (Idea is to not reduce this to $d$ different models)

I like the brulee API. Is there a possibility extending it to predict multiple outcomes? (assuming user provides the loss function)