rcurtin_irc changed the topic of #mlpack to: mlpack: a scalable machine learning library (https://www.mlpack.org/) -- channel logs: https://libera.irclog.whitequark.org/mlpack -- NOTE: messages sent here might not be seen by bridged users on matrix, gitter, or slack
<rcurtin[m]> mlpack discussed as part of the Anaconda dividend report: https://know.anaconda.com/rs/387-XNW-688/images/2021-Anaconda-DividendReport-.pdf
<rcurtin[m]> we worked with NumFOCUS to get those quotes and description there---hopefully this is another thing that can help drive visibility and interest :)
<Sirish07Sirish[m> Hey, can I use the current hyper parameter tuning methods on algos like SVDPlusPlus?
<shrit[m]> <rcurtin[m]> "mlpack discussed as part of..." <- Very good
<shrit[m]> Let us see what will happen in the next year
<rcurtin[m]> Sirish07 (Sirish): I *think* it could work; it depends on the API that `SVDPlusPlus` provides (I can't remember it offhand). I'm in a meeting now so I can't dig too much further, but if you take a look at the documentation for the hyperparameter tuning it should be pretty clear on the API that's needed
<rcurtin[m]> the whole hyperparameter tuner works basically by matching a certain API and doing template metaprogramming on it
<shrit[m]> 44k of donation is good start, hopefully in the near future this will increase
<zoq[m]1> Also, if you are trying to fix the test issue, it might be easier to find a random seed that breaks and debug from that point.
<zoq[m]1> Like you could run the test locally in a loop use a different test seed each time and use that as a starting point.
<Sirish07Sirish[m> yeah, I shall try it out. But I was a little confused as the hyper parameter tuning API requires responses (labels) as a parameter. In the recommendations dataset, I am not sure if there are features and labels explicitly.
<zoq[m]1> Sirish07Sirish[m: Right, you can't directly use it, you would have to wrap something around it.
<zoq[m]1> E.g. CF, for which you have some example tests.
<Sirish07Sirish[m> yeah 😭
<rcurtin[m]> you could, perhaps, also, if you wanted, use some of the categorical optimizers from ensmallen to do hyperparameter optimization
<rcurtin[m]> but it will be a little more work than using the hyperparameter optimization toolkit as-is
<Sirish07Sirish[m> Hey @marcusedel:matrix.org , I have a few queries regarding the cf tests, can I text you in person?
<zoq[m]1> I would prefer to use the public channels, so either here or on the github PR/issue.
<Sirish07Sirish[m> yeah that works.
<Sirish07Sirish[m> written on PR #3106