03:30 < ayesdie> Heya, I updated #1546 and `zoq` restarted the Travis build for it, but it seems to have failed at `sudo apt-key adv` part for one of the build.

06:56 < vedantbassi> Hi, for the FFN class, could someone explain the ordering of the mat parameters returned by the Parameters() method and how to make sense of it, thanks :)

13:50 < zoq> ayesdie: Let's restart the build one more time.

15:01 < ayesdie> I think it's built successfully now.

15:23 < zoq> ayesdie: Agreed, looks good.

18:25 < davida> zoq: If I wanted to print a cost curve as I am training, I believe I need to run model.Evaluate(). If correct, my question is how do I pass in the data to evaluate as there doesn't seem to be any parameter for the data in the function definition. This is particularly important with mini-batch gradient descent since I would want to calculate the loss on the full dataset and not just the last bactch I ran.

19:12 < zoq> davida: You have to manually run model.Evaluate(..) on the dataset, there is an open issue that adds the feature to do that.

19:13 < davida> zoq: How do I manually run it and add the dataset? This is where I am a bit lost,

19:24 < zoq> davida: The FFN class implements: double Evaluate(arma::mat predictors, arma::mat responses);

19:24 < zoq> davida: Do you think that could work?

19:27 < davida> zoq: I don't see that function. I see 3 versions of Evaluate all need to pass parameters.

19:30 < zoq> FFN or RNN?

19:30 < zoq> https://github.com/mlpack/mlpack/blob/master/src/mlpack/methods/ann/ffn.hpp#L145

19:31 < davida> zoq: FFN for now. I just went into the code and I see the function you are referring to. It is not documented on the docs page and that is why I missed it.

19:31 < zoq> ahh, have to step out, back later

20:01 < davida> zoq: I am at the ResNET stage of the Coursera examples. I have built the ResNET50 structure in Keras. I can build separate FFNs representing the Identity blocks and the Convolutional blocks, but I am not sure how to link them together in one mode. In fact, I need to link about 17 blocks together with pre and post stages. What is the right way to build this in MLPACK?

20:02 < davida> *mode = model