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/
sandinak has joined #mlpack
sandinak has quit [Remote host closed the connection]
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
uDdHsaWwk has joined #mlpack
uDdHsaWwk has quit [Remote host closed the connection]
bretfisher[m] has joined #mlpack
bretfisher[m] has quit [Remote host closed the connection]
niron has joined #mlpack
niron has quit [Remote host closed the connection]
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
SergioBZ has joined #mlpack
SergioBZ is now known as Guest84828
Guest84828 has quit [Remote host closed the connection]
andreagus15 has joined #mlpack
andreagus15 has quit [Remote host closed the connection]
EmleyMoor2 has joined #mlpack
EmleyMoor2 has quit [Remote host closed the connection]
Gizmokid200522 has joined #mlpack
Gizmokid200522 has quit [Remote host closed the connection]
schultmc4 has joined #mlpack
schultmc4 has quit [Remote host closed the connection]
WiseSon has joined #mlpack
WiseSon has quit [Remote host closed the connection]
gtank___ has quit [K-Lined]
braceletboy has quit [K-Lined]
Yen has joined #mlpack
Yen has quit [Remote host closed the connection]
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
ImQ009 has joined #mlpack
roryrjb has joined #mlpack
roryrjb has quit [Remote host closed the connection]
kitech120 has joined #mlpack
kitech120 has quit [Killed (Sigyn (Spam is off topic on freenode.))]
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
GaLaK`24 has joined #mlpack
GaLaK`24 has quit [Remote host closed the connection]
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
sysRPL has joined #mlpack
sysRPL has quit [Remote host closed the connection]
< jonpsy[m]>
zoq: Hi, so the main code of MOEA/D-DE is **done**. I'm working on templatizing the weight init policy. Part of it includes sampling from canonical unit simplex, I believe you've worked on it before, could you throw some light? Thanks
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
mindlifter has quit [Remote host closed the connection]
geceo[m|gr] has joined #mlpack
geceo[m|gr] has quit [Remote host closed the connection]
mindlifter has joined #mlpack
renato^^^ has joined #mlpack
renato^^^ has left #mlpack []
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
Thracky16 has joined #mlpack
Thracky16 has quit [Read error: Connection reset by peer]
Simvav has joined #mlpack
< zoq>
jonpsy[m]: on the policy design pattern?
< HarshVardhanKuma>
> @HarshVardhanKumar: if you compile with OpenMP, or if you are using OpenBLAS, this will happen automatically (depending on the algorithm) :)
< HarshVardhanKuma>
Hey, I've just installed using sudo apt. This default version is already compiled with OpenMP, right?
< HarshVardhanKuma>
* > @HarshVardhanKumar: if you compile with OpenMP, or if you are using OpenBLAS, this will happen automatically (depending on the algorithm) :)
< HarshVardhanKuma>
Hey, I've installed using sudo apt. This default version is already compiled with OpenMP, right?
Simvav has quit [Quit: Connection closed]
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
< rcurtin[m]1>
HarshVardhanKumar (Harsh Kumar): yeah, I think that should be using both OpenMP and OpenBLAS thunfisch
< rcurtin[m]1>
* HarshVardhanKumar (Harsh Kumar): yeah, I think that should be using both OpenMP and OpenBLAS 👍️
< RishabhGarg108Ri>
Hii, how can we explicitly define datatype for individual dimensions in DatasetInfo ?
< RishabhGarg108Ri>
Let's say I have to make a particular dimension categorical, then how can I do it ?
mindlifter has quit [Remote host closed the connection]
< jonpsy[m]>
> jonpsy: on the policy design pattern?
< jonpsy[m]>
yessir
mindlifter has joined #mlpack
< zoq>
RishabhGarg108Ri: That will be automatically infered from the type.
< zoq>
jonpsy[m]: Will comment on the PR.
< RishabhGarg108Ri>
@freenode_zoq:matrix.org , I thought the same. I used the `data::Load` overload that takes DatasetInfo, but it is assigning all dimensions to numeric
< RishabhGarg108Ri>
Although some of them should be categorical
dowdle has quit [Remote host closed the connection]
< zoq>
Is that for the house price; happy to take a look.
< zoq>
In this case you might want to remove the header.
< RishabhGarg108Ri>
Yes. it is for boston housing price dataset.
< RishabhGarg108Ri>
How can I remove the header?
< zoq>
I would just do it manually.
< RishabhGarg108Ri>
Ok. Let me try it out :+1:
< rcurtin[m]1>
I don't think I have the syntax exactly right there, but the general idea should work if I am not mistaken
< rcurtin[m]1>
if you like, I can spend a few minutes to construct a correct working example, just let me know. I don't seem to see any tests in `load_save_test.cpp` that easily demonstrate this functionality
< RishabhGarg108Ri>
Well, this has been itching me for too long and I just want to get rid of it as soon as possible. 😃
< rcurtin[m]1>
I'm glad you thought about the data loading here, I was still scratching my head thinking about what could be theoretically wrong with the approach... if the data is being loaded as all numeric, this could make a big difference in the quality of the tree
< RishabhGarg108Ri>
@freenode_zoq:matrix.org I did what you suggested but it is giving segmentation fault. By "header" you meant the header in the csv file right?
< RishabhGarg108Ri>
@ryan:ratml.org yes. It was hard to figure out. :)
< RishabhGarg108Ri>
@ryr
< rcurtin[m]1>
👍️ of course, the data type might not be the only issue... it might be worth also comparing the performance of the iris dataset
< rcurtin[m]1>
in fact, let me try that quickly... hopefully we should be able to produce 0.0 MSE on the training set
ew73_ has joined #mlpack
ew73_ has quit [Remote host closed the connection]
< RishabhGarg108Ri>
> in fact, let me try that quickly... hopefully we should be able to produce 0.0 MSE on the training set
< RishabhGarg108Ri>
Great. Thanks!
< rcurtin[m]1>
the `datasetInfo` holds, for each categorical dimension, a mapping between input values (the strings "0", "1", "2", "3") and the integer values used to represent them. That code snippet there basically adds "0", "1", "2", and "3" as valid values for dimension 2, and "0" and "1" as a valid value for dimension 3
< rcurtin[m]1>
I think that the interface to `DatasetInfo` could be cleaned up a bit for sure...
< heisenbuugGopiMT>
I agree with you @ryan:ratml.org...
< heisenbuugGopiMT>
I went through most of the code while working on parser...
< heisenbuugGopiMT>
If it's okay I can clean it up a bit when I will be integrating the new parser with `DatasetInfo`
< rcurtin[m]1>
absolutely heisenbuug (Gopi M Tatiraju), that would be great! one task we should definitely make easier is for a user to do things like specify which dimensions are categorical before load 👍️
< RishabhGarg108Ri>
One thing that can surely be added is that a user can pass a list of categorical features and then the function can itself scan and find the valid values for those dimensions
< rcurtin[m]1>
RishabhGarg108 (RishabhGarg108): I'll let you know how it does on the iris dataset, but I have to handle one or two other things first
< RishabhGarg108Ri>
Sure, do it at your own time. No hurries :)
< heisenbuugGopiMT>
Yes, we can have a discussion on it, I will let you all know when I will start working on it...
< heisenbuugGopiMT>
Yea, do we have any auto detect implemented for attribute type?
< shrit[m]>
heisenbuug (Gopi M Tatiraju): you mean for the categorical data?
< heisenbuugGopiMT>
No...
< heisenbuugGopiMT>
Can there be a way to detect at the time of loading if attribute is numeric or categorical?
< heisenbuugGopiMT>
Ahh, never mind...
< rcurtin[m]1>
yeah, so we do autodetection already simply by trying to load the data as a float64, and if that fails, then we assume it is categorical
< rcurtin[m]1>
but, this is not perfect... if a user specifies their categories as 0, 1, 2, 3, 4, ..., we have no way of knowing just from the file whether that should be categorical
< heisenbuugGopiMT>
What can be a better way to do it?
< rcurtin[m]1>
there is no better way to do autodetection... it is ambiguous; we will have to provide the user with a clean way to specify, at load time, if any dimensions should be categorical 👍️
< rcurtin[m]1>
the current way of "create a `DatasetInfo` and manually set dimensions" works... but it's not documented and maybe could be improved upon :)
< heisenbuugGopiMT>
Yea, implementation is good, we just need to clean the interface and document it...
Michael-M120 has joined #mlpack
Michael-M120 has quit [Remote host closed the connection]
mindlifter has quit [Remote host closed the connection]
mindlifter has joined #mlpack
< Aakash-kaushikAa>
Hey @kartikdutt18, @zoq in y recent code or thr resent PR i am facing a segfault while trying to run the code is there a better way to debug why this could be happening ?
< swaingotnochill[>
zoq I would be taking today off...Having Viral Fever 🤒 due to weather changes... The first week notebook is half done just stuck with matplotlibcpp...will complete it by tomorrow.
< rcurtin[m]1>
hope you feel better swaingotnochill! I glanced through the notebook while looking into #2977 and it looks really nice 😃 (I am not sure if that's the same one you were referring to though)
< swaingotnochill[>
As for the python bindings one..guess will have to wait for it to be solved, though it's pretty much done
< swaingotnochill[>
ryan No, that one is python bindings one. The thing is I want to make same visualizations for the cpp notebook too, but matplotlibcpp is not that flexible..
< swaingotnochill[>
I am having difficulty creating a histogram in matplotlibcpp..they don't have any documentation too😔
< swaingotnochill[>
ryan Did you find any way to fix or work around that issue in notebook, that solve() solution not found error? It's not urgent, but whenever you are free, please take a look.
< rcurtin[m]1>
I just left a comment on the issue a few moments ago---take a look and let me know what you think when you have a chance :)
ScorchedMuffin23 has joined #mlpack
ScorchedMuffin23 has quit [Read error: Connection reset by peer]
< swaingotnochill[>
Sure
< jonpsy[m]>
> zoq I would be taking today off...Having Viral Fever 🤒 due to weather changes... The first week notebook is half done just stuck with matplotlibcpp...will complete it by tomorrow.
< jonpsy[m]>
Actually, have yourself covid checked. I thought similar when I had normal fever, but it soon turned out to be covid symptoms and it quickly went downhill
< zoq>
jonpsy[m]: Totally fine, take the time to fully recover, no need to work over the weekend.
< zoq>
Ahh wrong person.
< zoq>
swaingotnochill[: Totally fine, take the time to fully recover, no need to work over the weekend.
< zoq>
swaingotnochill[: About matplotlibcpp not flexible enough, you can also just write the visualization in python and use the python C API to call that one to generate the plot.
< zoq>
That's what we did for some plots in the past.
< zoq>
Like the movie lens cf example uses a wordcloud python script to generate the wordcloud.
< swaingotnochill[>
Took a paracetamol....it's heavily raining here in Mumbai....the transition of weather always hits me😅