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/
< rcurtin>
only the changes to decision_tree_impl.hpp need to be applied really
< rcurtin>
the rest is helpful but not needed just to see if the problem is fixed
< gmanlan>
perfect
< gmanlan>
thanks for taking a look at this
< rcurtin>
sure, it turned out to be pretty easy, I think
< rcurtin>
assuming it's fixed :)
< gmanlan>
rcurtin: (I'm still building to test)
< gmanlan>
BTW, do you think it would make sense to support more hyperparameters? (i.e. max features, max depth, min samples split, min samples leaf...)
< rcurtin>
I certainly wouldn't have a problem adding more
< gmanlan>
right, I'm thinking about "feature parity" with other frameworks (i.e. sci-kit)
< gmanlan>
ok so I have tested using the sample app
< gmanlan>
now the training accuracy, size of the model and other metrics have drastically changed
< gmanlan>
but the resulting model is still static (i.e. it doesn't change every time you run)
< zoq>
set mlpack::math::RandomSeed(time(NULL)); at the beginning
< gmanlan>
let me try
< gmanlan>
ahhh that's it zoq
< gmanlan>
that fixes the problem
< gmanlan>
interesting, I thought a random seed would have been something embedded into the core for random forest
< rcurtin>
gmanlan: mlpack doesn't automatically set the seed; if you were to run, e.g., mlpack_random_forest, by default it'll set the seed to std::time(NULL)
< rcurtin>
but in a run on your own you'd have to set it
< gmanlan>
uhm ok, that's probably something we could document then
< gmanlan>
by default it'll set the seed to std::time(NULL) --> you mean the algorithm is doing it?
< gmanlan>
ah no, it's the test app that is doing it
< rcurtin>
oh, you're right... there's no --seed. that's an irritating oversight, let me fix that in my PR too
pd09041999 has joined #mlpack
< chandramouli_r>
rcurtin: can you help me with this error .
< chandramouli_r>
mlpack/core/dists/gaussian_distribution.cpp line="20" id="issue" severity="warning" msg="It is possible that not all members of a class are initialized inside the constructor. Consider inspecting: logDetCov.
< chandramouli_r>
It is mentioned in 6th point. Whenever I pull the code the static code analyser always gives this error. It should be ignored or is there any changes to be made.
< zoq>
chandramouli_r: Looks like adding logDetCov(0.0) to the GaussianDistribution(const arma::vec& mean, const arma::mat& covariance) constructor would fix the the warning.
< Mulx10>
Zoq: ah yes.
< zoq>
chandramouli_r: Btw. feel free to work on optimizing the code.
Mulx10 has quit [Ping timeout: 256 seconds]
< rcurtin>
gmanlan: added a --seed parameter to mlpack_random_forest as part of #1891
pd09041999 has joined #mlpack
jeffin143 has joined #mlpack
jeffin143 has quit [Client Quit]
< chandramouli_r>
zoq: definitely yes I will work on that too.