ChanServ changed the topic of #mlpack to: "mlpack: a fast, flexible machine learning library :: We don't always respond instantly, but we will respond; please be patient :: Logs at http://www.mlpack.org/irc/
ImQ009 has joined #mlpack
< chopper_inbound[> Hello everyone! Here is my weekly update blog for week 7 and 8. https://mrityunjay-tripathi.github.io/gsoc-with-mlpack/coding_period/week7_and_8.html :)
< nishantkr18[m]> hello everyone. Is there a way to get gradient wrt input in a Feed forward network(FFN)?
< nishantkr18[m]> I see that while doing `network.Backward(input, target, gradients)`, can give us gradients. But those gradients are wrt the weights and biases of the network. I was looking for a way to get gradient wrt the inputs.
< zoq> nishantkr18[m]: You can use layer.Delta()
ImQ009 has quit [Ping timeout: 272 seconds]
< zoq> nishantkr18[m]: https://github.com/mlpack/mlpack/blob/373a2a51d09e51bdb98e2e5a0d7959912cbd665b/src/mlpack/methods/ann/ffn_impl.hpp#L509-L523 is the code block that does the backward step in the FFN class.
< nishantkr18[m]> Thanks zoq
ImQ009 has joined #mlpack
< mlozhnikov[m]> Looks like I broke the App Veyor build. I thought it happened due to time limit. In fact it's not the case. I'll open a PR soon.
< himanshu_pathak[> Hey zoq: Is there a way I can do somthing like `data.shed_cols(where(label == 0))` in `arma::mat`
< zoq> himanshu_pathak[: Yes, data.shed_cols(arma::find(label == 0)), arma::find(label == 0) returns indices.
< himanshu_pathak[> zoq: thanks
< jeffin143[m]> @rcurtin:matrix.org: you there
< jeffin143[m]> How about a class interface for one hot encoding .??
< rcurtin> jeffin143[m]: why a class interface? seems to me like just a function would be fine
< rcurtin> I suppose a class could be useful if, e.g., there is some state and someone might like to, e.g., one-hot encode a new dataset or something
< sakshamb189[m]> himanshu_pathak: I think we missed our meet this week. When should we have that?
< himanshu_pathak[> Hey sakshamb189: I think today I will complete the implementation of multiclass svm can we have a meet tommorow??
< himanshu_pathak[> I forgot about the meet
< sakshamb189[m]> alright that should be fine :)
< rcurtin> jeffin143[m]: maybe we can get the existing one-hot encoding binding in, then file an issue for a follow-up fix or something? it would be great to get that into mlpack 3.4.0
< sakshamb189[m]> just make sure that the style and documentation build are passing 🙂
< himanshu_pathak[> Yeah I will sakshamb189: also the memory builds 😂
favre49 has joined #mlpack
favre49 has quit [Ping timeout: 260 seconds]
favre49 has joined #mlpack
< yashwants19[m]> Hi "rcurtin" (https://matrix.to/#/@freenode_rcurtin:matrix.org): Is version correct in https://github.com/mlpack/mlpack/blob/master/src/mlpack/core/util/version.hpp?
< rcurtin> yashwants19[m]: wow, it is not. that highlights the importance of getting a good automated release system in place that takes care of this automatically :-D
< rcurtin> and perhaps it also highlights the importance of not having me do releases, since I always seem to miss something :)
< rcurtin> I added that note to the list of things to be fixed in the release preparation guide: https://github.com/mlpack/mlpack/wiki/ReleasePreparationGuide
< yashwants19[m]> <rcurtin "I added that note to the list of"> This sounds good, thanks for adding it into the guide.
< yashwants19[m]> :)
pie3 has joined #mlpack
pie3 has quit [Ping timeout: 256 seconds]
< jjb[m]> Ryan Curtin outside of automation, maybe consider adding a “Release checklist” issue that acts in a manner similar to a grocery shopping list of what needs to happen for manual tasks?
< rcurtin> yeah, that's what https://github.com/mlpack/mlpack/wiki/ReleasePreparationGuide is supposed to be :-D
< rcurtin> it's changed a lot over the years and I've been removing things as the release scripts manage to handle them automatically
< rcurtin> but it looks like the update to version.hpp might have gotten accidentally removed or something
< favre49> https://github.com/mlpack/examples/blob/master/q_learning/cartpole_dqn.ipynb Not sure if this is just something on my side, but the Binder badge seems to fill the entire width of the page
< rcurtin> favre49: not for me, it seems like a regular "badge", maybe 150px wide
< rcurtin> I'm using firefox on debian, so, not sure if that is different than your setup
< zoq> favre49: Same for me, looks like a normal badge.
< favre49> I'm using firefox too, maybe it's something about using a custom userchrome then
favre49 has quit [Remote host closed the connection]
pie3 has joined #mlpack
pie3 has quit [Ping timeout: 240 seconds]
pie3 has joined #mlpack
ImQ009 has quit [Quit: Leaving]
< jeffin143[m]> 2535/36 could be merged
< jeffin143[m]> :)
pie3 has quit [Read error: Connection reset by peer]
< himanshu_pathak[> <zoq "himanshu_pathak: Yes, data.shed_"> Hey zoq: I am using this but getting this error ``` /usr/include/armadillo_bits/Row_bones.hpp:125:15: note: candidate: void arma::Row<eT>::shed_cols(arma::uword, arma::uword) [with eT = double; arma::uword = long long unsigned int]
< zoq> himanshu_pathak[: Can you post or link me to the code you are using?
< himanshu_pathak[> zoq: You can look here
< zoq> himanshu_pathak[: ahh, shed_cols that takes indices was introduced in 9.4
< jeffin143[m]> @rcurtin:matrix.org: whenever you get time , take a look at 2325
< jeffin143[m]> Attempted to solve the comments you left
< zoq> himanshu_pathak[: A workaround would be to use cols instead
< himanshu_pathak[> zoq: Ok I will try that :)
< zoq> himanshu_pathak[: so tempLables.cols(arma::find(tempLables == 1));
< himanshu_pathak[> Yup
< zoq> himanshu_pathak[: So we don't remove the cols we extrac the cols we want.
< zoq> himanshu_pathak[: -1 not 1
< himanshu_pathak[> <zoq "himanshu_pathak: -1 not 1"> What you mean by that??
< zoq> himanshu_pathak[: in the example above I used tempLables == 1 but I think it should be tempLables == -1
< himanshu_pathak[> zoq: Oh got it.
< jeffin143[m]> zoq (@freenode_zoq:matrix.org): do you have some bandwidth for 1895
< zoq> jeffin143[m]: I'll set it on my list for later today/tomorrow.
< jeffin143[m]> No issues , thank you :) zoq (@freenode_zoq:matrix.org)
< jenkins-mlpack2> Project mlpack - git commit test build #490: UNSTABLE in 1 hr 25 min: http://ci.mlpack.org/job/mlpack%20-%20git%20commit%20test/490/