03:24 < ricklly> hello, rcrutin, can you give me some feedback for my draft if available?

03:57 < ricklly> hello, I do cmake source with BUILD_PYTHON_BINDS on, but later I still can't import mlpack in python. What's wrong with it maybe?

04:10 < ricklly> ok, I solve it

06:52 < ricklly> "w/ a linearly dependent set of constraints", hello, may I ask what is "w/ a"?

09:44 < Atharva> sumedhghaisas2: I saw your NTM pr but it is huge and I can’t find the LayerTypes implementation anywhere, can you please point me to the file you changed or something like that?

09:57 < Atharva> I found it, I had some doubts, Are you saying that we get rid of the Model template parameter and pass encoder/decoder as LayerTypes<> which will have all the layers of the encoder/decoder network?

10:47 < manthan> rcurtin : zoq : kindly review my proposal if you get time. I have updated it according to zoq's suggestions. It will be very helpful for me. Thanks.

12:37 < zoq> ckeshavabs: Yeah it could be similar to Gradient clipping, and you are right, Reward clipping is probably only useful for the RL methods, so if you like to place it inside the RL related folder please feel free.

12:38 < zoq> hassanmahmo: See the answer on the mailing list.

13:03 < Atharva> Is the discrete distribution in core/dists a generalisation of the bernoulli distribution?

14:03 < haritha1313> @zoq: Hi, I guess it was you who reviewed my draft as Anonymus. I had replied to the comments regarding ANN layers to be used and need of another merge layer. Should I repeat them here so we can resolve it, in case there is any point which needs more clarity?

14:19 < rajiv_> @zoq: Hi Marcus, I have added major changes to the timeline and also included a few pseudo-codes to my proposal. It would be of great help if you could review it and give your feedback so that I can work on improving it

15:44 < enowak> Hi. I have been posting on mailing list with GSOC'2018. Unfortunatelly due to my job duties I cannot afford enough time for this project during June-September. I am still interested in contributing to project without GSoC program in a free time. Should I post message on mailing list, that I will not participate in GSOC but I am interested in developing proposed architectures in a free time?

17:24 < sourabhvarshney1> enowak: Hello there!!! Feel free to contribute. You can visit http://www.mlpack.org/involved.html and http://www.mlpack.org/gsoc.html for getting started.

17:59 < hassanmahmood> Hello! I commented on this opened issue 2 days ago and haven't got any response. I had a query. It would be great if someone can answer it. https://github.com/mlpack/mlpack/issues/1021

20:52 < sourabhvarshney1> @zoq I would like work on tutorial of rnn if no one else is working

20:54 < Atharva> sumedhghaisas2: I tried to reproduce the changes from your PR by changing the layer_types.hpp file but have been facing compilation errors during make. Am I missing something? Does it require changes other that that file? I am trying to define LayerType for FFN models.

21:21 < sujith> Hello, I have a query regarding the issue #406. Why can’t we assume that d(X.col(i),X.col(j) = d(H.col(i),H.col(j)) as multiplying with W will increase the euclidean distance of all the distance pairs by some constant factor, not changing the sorted order?