Add new algorithm implemetation to the tasks

[peng yu]

Hi, 

We are trying to benchmark our implementation of  Random Extreme Hoeffding Tree, How can i add to the tasks? I did registered with the platform, but i can't receive the activation email.

[Joaquin Vanschoren]

Hi Peng, 

It looks like your account is active. Did you find the activation email after all?

When you say 'add to the tasks', do you mean that you want to create a new task (e.g. a classification task) on a dataset?

Note that OpenML doesn't run benchmark experiments by itself. It is a platform for sharing experiments which usually run on the client side (your own machine/cloud/...). In which library or programming language did you implement your algorithm? Then I could point you to the right API for that.

Cheers,

Joaquin

On Wed, 4 Jul 2018 at 20:28 peng yu <yup...@gmail.com> wrote:

Hi, 

We are trying to benchmark our implementation of  Random Extreme Hoeffding Tree, How can i add to the tasks? I did registered with the platform, but i can't receive the activation email.

--
You received this message because you are subscribed to the Google Groups "OpenML" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openml+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Thank you,

Joaquin

[peng yu]

Hi Joaquin, 

Thanks for replying!  I'm using tensorforest https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/tensor_forest for classification for now.

And yeah, that is still great if there is orchestration help of different benchmarks running with the right api and share the results with public on my own machine.

To unsubscribe from this group and stop receiving emails from it, send an email to openml+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Thank you,

Joaquin

[Joaquin Vanschoren]

Sadly, we don't have a native tensorflow API yet :(. 

Is there an easy way to wrap your algorithm as a scikit-learn estimator? (I remember there was scikit-flow, but this is now folded into TensorFlow itself, right?) If so, you can probably just piggyback on our sklearn API. I haven't tried that before, though. 

If not, this will be quite a bit of work, since you'd basically need to write a tensorflow extension for our Python API. 

Worst case, you could still use our Python API to download tasks (datasets + train/test splits), run experiments locally, and compare them to previous results on the same tasks. But you won't be able to upload your algorithms and experiments automatically yet.

If any of this sounds interesting, let me know.

Best,

Joaquin

To unsubscribe from this group and stop receiving emails from it, send an email to openml+un...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Thank you,

Joaquin

--

Thank you,

Joaquin

[peng yu]

Thank you! 

Yeah, I think i'll try to look into the sklearn API first.  Except for streaming algorithm,  how is the evaluation api designed ? i don't think sklearn API support that? 

To unsubscribe from this group and stop receiving emails from it, send an email to openml+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Thank you,

Joaquin

--

Thank you,

Joaquin

[Joaquin Vanschoren]

Dear Peng,

Sorry about the slow reply. I was on holidays. I looked into it, and indeed this is not yet supported in the Python API. It’s on our list, but it may take a little while ti support it. Any help is of course appreciated :). If we can help any other way, let us know.

Best,

Joaquin

To unsubscribe from this group and stop receiving emails from it, send an email to openml+un...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Thank you,

Joaquin

--

Thank you,

Joaquin

--

Thank you,

Joaquin