03:08 <rcurtin[m]> shrit: if you have `covertype.csv` somewhere, get a subset of it like `head -20000 covertype.csv > covertype-20k.csv`, then you can do: `mlpack_knn -k 3 -r covertype.csv -a naive -v`

03:08 <rcurtin[m]> (the `naive` approach scales quadratically with the number of data points, so the full covertype dataset is probably too much)

11:08 <jonpsy[m]> zoq: so i've a question, and this might be a potential bug

11:13 <jonpsy[m]> So here, what if the `nextState` is terminal? then `nextActionValues` aren't defined right?

11:13 <jonpsy[m]> * So [here, * here](https://github.com/mlpack/mlpack/blob/b2c14719fe245835973ed77fc426b05b7b55df71/src/mlpack/methods/reinforcement_learning/replay/prioritized_replay.hpp#L297-L298), what

12:58 <zoq[m]1> <jonpsy[m]> "So here, what if the `nextState`..." <- True, but if `nextState` you don't call `Update` right?

13:35 <zoq[m]1> * if `nextState` is terminal, you don't

13:35 <jonpsy[m]> lemme check

13:38 <jonpsy[m]> okay, i dont think so. can you fill me up with some code links?

15:10 <zoq[m]1> https://github.com/mlpack/mlpack/blob/b2c14719fe245835973ed77fc426b05b7b55df71/src/mlpack/methods/reinforcement_learning/q_learning_impl.hpp#L363-L364 that's the line I was thinking about

15:52 <jonpsy[m]> good point, and a common pittfal.

15:52 <jonpsy[m]> Let's imagine we're in a state, who's next state is terminal. That means we would still be inside while loop (the next loop will exit). So that means, `TrainAgent` is still called. And when you use `next_state` in that, it'll be `NULL`

15:52 <jonpsy[m]> s///, s/,//

15:53 <jonpsy[m]> we've handled it succintly [here](https://github.com/mlpack/mlpack/blob/b2c14719fe245835973ed77fc426b05b7b55df71/src/mlpack/methods/reinforcement_learning/q_learning_impl.hpp#L176-L178). But we haven't done the same for `Update`

16:58 <shrit[m]> rcurtin: I think there is a leaf size needs to be specific now for mlpack_knn?

16:59 <shrit[m]> I suppose it is 20

17:00 <shrit[m]> yeah, it just needs to be specified, figured out my self

17:01 <zoq[m]1> > <@jonpsy:matrix.org> good point, and a common pittfal.

17:01 <zoq[m]1> > Let's imagine we're in a state who's next state is terminal. That means we would still be inside while loop (the next loop will exit). So that means, `TrainAgent` is still called. And when you use `next_state` in that, it'll be `NULL`

17:01 <zoq[m]1> >

17:01 <zoq[m]1> So that means we have to check if the state is terminal.

17:01 <jonpsy[m]> yessir

17:02 <jonpsy[m]> well, unless im missing something. yes

17:02 <zoq[m]1> I checked it again with the information in mind, and to me it looks like you are right.

17:03 <zoq[m]1> We could manually validate that if we wanted.

17:03 <jonpsy[m]> we could go for a TDD approach

17:04 <jonpsy[m]> so we have to write a test case where our current impl fails, PR our fix, see if it fixes it

17:04 <zoq[m]1> Yes, that is a great idea!

17:05 <jonpsy[m]> perfect.

17:05 <jonpsy[m]> also heard of the NumFOCUS Project Incubator Community programme?

17:07 <rcurtin[m]> shrit: it should default to 20 so I don't think you need to specify it, but it doesn't hurt

17:08 <zoq[m]1> jonpsy[m]: Perhaps I have seen the idea on the list, but just read the blog post.

17:09 <jonpsy[m]> yeah, im applying. let's see what happens ;)

17:10 <zoq[m]1> jonpsy[m]: Nice!

17:11 <rcurtin[m]> jonpsy: do you have a link or anything? I haven't heard of the program

17:12 <zoq[m]1> https://numfocus.medium.com/

17:12 <jonpsy[m]> rcurtin[m]: aye aye captain, [here](https://numfocus.medium.com/call-for-volunteers-numfocus-project-incubator-3ec7b4567846)

17:12 <rcurtin[m]> thank you :)

17:12 <rcurtin[m]> I didn't think to check the blog

17:12 <jonpsy[m]> i've heard Paige Bailey would be one of the leads, so exciting!!

17:12 <rcurtin[m]> even though now that I look closer... zoq mentioned that it had a blog post. should have read more closely 😄

17:13 <jonpsy[m]> rcurtin: We could help budding organization, maybe they can even use our libraries as their base. That's the very spirit of open source

17:13 <rcurtin[m]> agreed, this could be a really neat opportunity

17:13 <jonpsy[m]> just like many orgs, use tensorflow/numpy as their base

17:14 <jonpsy[m]> s/,//

17:46 <shrit[m]> I think the Incubator is to help outside projects to enter NumFOCUS if I understand right?

18:11 <jonpsy[m]> I think its more of a "growth accelerator" for under-resourced, budding scientific open source projects.

18:11 <jonpsy[m]> hence the term "incubator"

18:15 <shrit[m]> What is the NumFOCUS Incubator?

18:15 <shrit[m]> The NumFOCUS Incubator was approved by the NumFOCUS Board in October to provide support for Open Source Scientific projects to grow, build a community, attract additional contributors, and more. At the end of the incubation period, projects will have the tools they need to either continue working towards meeting the Affiliated project requirements or directly apply for that status.

18:15 <shrit[m]> Feel free to reach NumFOCUS to verify I am not sure,

18:15 <shrit[m]> but that is what I would understand from the paragraph

18:43 <jonpsy[m]> zoq: Remember the "click" sound from my HDD. It's frequency increased today, I think its time has come

18:44 <shrit[m]> You should backup immediatly, before the last "click"

18:47 <jonpsy[m]> good for me, my local fortune teller warned me about this

19:17 <shrit[m]> I should do somebackups

19:17 <shrit[m]> I did not do any backup since I moved to ireland

19:18 <shrit[m]> It is on my long to do list

19:29 <jonpsy[m]> <jonpsy[m]> "zoq: Remember the "click" sound..." <- consider this an omen :D

20:19 <zoq[m]1> Did you already order a new hdd or are you now going for a new device?

20:40 <heisenbuugGopiMT> ```

20:40 <heisenbuugGopiMT> In file included from /home/heisenbug/heisenbug/ncf/mlpack/src/mlpack/methods/cf/ncf_main.cpp:21:

20:40 <heisenbuugGopiMT> 53 | #error "Unknown binding type! Be sure BINDING_TYPE is defined if you are " \

20:40 <heisenbuugGopiMT> /home/heisenbug/heisenbug/ncf/mlpack/src/mlpack/core/util/mlpack_main.hpp:53:4: error: #error "Unknown binding type! Be sure BINDING_TYPE is defined if you are " "including <mlpack/core/util/mlpack_main.hpp>.";

20:41 <heisenbuugGopiMT> Where do we have to define `BINDING_TYPE`?

20:41 <heisenbuugGopiMT> I dont see it in any other examples...