07:50 < ShikharJ> zoq: I pushed in the missing file, sorry foe the delay.

11:25 < jenkins-mlpack> Yippee, build fixed!

11:25 < jenkins-mlpack> Project docker mlpack nightly build build #363: FIXED in 4 hr 11 min: http://masterblaster.mlpack.org/job/docker%20mlpack%20nightly%20build/363/

12:17 < sumedhghaisas> Atharva: Hi Atharva

12:17 < sumedhghaisas> sorry was little busy yesterday

12:17 < sumedhghaisas> I will take a look at the jocobian test today

12:20 < Atharva> sumedhghaisas: No problem. After the jacobian is taken care of, I will update the reconstruction PR. We need to merge both of them to start with the models. I have prepared the README.md

12:21 < Atharva> Or maybe, after we merge the normal dist PR, I will rebase the reconstruction PR and then we can get it merged

12:24 < sumedhghaisas> Atharva: It might take a while to merge them, I recommend working by rebasing on them

12:24 < Atharva> Okay, sure

12:24 < sumedhghaisas> in case they change, we can always rebase again

12:25 < sumedhghaisas> Regarding models, you will need to link models with the local copy of your code

12:27 < Atharva> Okay, so I will be able to start working on it before the PRs are merged

12:27 < sumedhghaisas> yes

12:29 < Atharva> As soon as we get the jacobian sorted, I will start working on the models with a local copy with all the changes

13:17 < sumedhghaisas> Atharva: Haven't taken a look yet though what is the issue you are facing? the approximate and real gradient for LogProb don't match right?

13:18 < Atharva> Yes

13:18 < Atharva> They dont

15:25 < manish7294> rcurtin: Please try to review the eval bounds as you get the chance https://github.com/mlpack/mlpack/issues/1449#issuecomment-401056900

15:26 < rcurtin> right, I saw them, I have it lined up for today

15:26 < rcurtin> intuitively, glancing at them they seem correct, I'll take a closer look though and verify

15:26 < rcurtin> you can always go ahead and implement as-is, and then see if it still gives the same results :)

15:26 < manish7294> Ya sure :)

15:27 < rcurtin> even if the bound is a little wrong, the code you implement could probably be easily adapted to a different version of the bound, so no need to wait until we have verified bounds to implement

15:29 < manish7294> you may not find the handwriting pleasant though :)

15:30 < rcurtin> that's ok, mine is not so great either :)

15:30 < rcurtin> I have to step out for lunch, back in a little while

18:58 < ShikharJ> zoq: Looks like the issue with enable_if is resolved. As of yet, we'll need to have three Evaluate and Gradient functions (one for both GAN and DCGAN, and one each for standard WGAN and Gradient-Penalized WGAN). Out of these functions, three are ready, and three more would be ready by Sunday. Then we can review and merge the PR.

19:07 < ShikharJ> zoq: Ah, I noticed that Gradient-Penalised WGAN also has a similar Gradient routine as Standard GAN and DCGAN. So now only the Evaluate functions for the WGANs are remaining to be implemented.

19:12 < zoq> ShikharJ: I think this are good news :)