Ignite ML status update

[Yuriy Babak]

Hi Igniters,

Here is the some update about Ignite ML. Last 2 month we working on a few new cool features.

First of all it's decision tree algorithm with next step - random forest. Currently this feature under development by Artem Malykh.

The same for the second major change - BLAS. BLAS is high efficient lib for linear algebra and it's "must have" for any ML lib. We already have this feature but now it works only for local on-heap matrices/vectors. So the next step is add support for off-heap matrices/vectors and Oleg Ignatenko working on it. Also we will set BLAS as default for any math operations instead of current naive implementations.

And last but not least - new distributed block matrix. This version is more efficient rather than current distributed row/col matrix. It's already have better performance for multiplication. Also this matrix is suitable for Strassen algorithm which have lower complexity - O(n^2.8).

All this features will be relesed with AI 2.2. Also for AI 2.2 we will deffetly implement log regression and maybe something else.

Stay tuned for further updateds.

Regards,
Yury.

[dsetrakyan]

Yury, thanks for the update! It would be nice if you wrote a blog about
IgniteML where you would demonstrate a few examples. This would make it
easier for the community to digest the new features.

On Wed, Jul 19, 2017 at 11:18 AM, Yury Babak <[hidden email]> wrote:

> Hi Igniters,
>
> Here is the some update about Ignite ML. Last 2 month we working on a few
> new cool features.
>
> First of all it's decision tree algorithm with next step - random forest.
> Currently this feature under development by Artem Malykh.
>
> The same for the second major change - BLAS. BLAS is high efficient lib for
> linear algebra and it's "must have" for any ML lib. We already have this
> feature but now it works only for local on-heap matrices/vectors. So the
> next step is add support for off-heap matrices/vectors and Oleg Ignatenko
> working on it. Also we will set BLAS as default for any math operations
> instead of current naive implementations.
>
> And last but not least - new distributed block matrix. This version is more
> efficient rather than current distributed row/col matrix. It's already have
> better performance for multiplication. Also this matrix is suitable for
> Strassen algorithm which have lower complexity - O(n^2.8).
>
> All this features will be relesed with AI 2.2. Also for AI 2.2 we will
> deffetly implement log regression and maybe something else.
>
> Stay tuned for further updateds.
>
> Regards,
> Yury.
>
>
>
> --
> View this message in context: http://apache-ignite-
> developers.2346864.n4.nabble.com/Ignite-ML-status-update-tp19829.html
> Sent from the Apache Ignite Developers mailing list archive at Nabble.com.
>

[Yuriy Babak]

Dmitriy, I can do it. But as far as I know Oleg Ignatenko already wrote blog about Ignite ML. But it only about AI 2.0 and only on Russian(Habrahabr). So we could translate this post in English.

And also I can write a blog about new features ML in AI 2.1 and mentioned features which we plan for 2.2.

Regards,
Yury

[dsetrakyan]

Sounds good.

On Jul 20, 2017, 1:56 PM, at 1:56 PM, Yury Babak <[hidden email]> wrote:

>Dmitriy, I can do it. But as far as I know Oleg Ignatenko already wrote
>blog
>about Ignite ML. But it only about AI 2.0 and only on
>Russian(Habrahabr). So
>we could translate this post in English.
>
>And also I can write a blog about new features ML in AI 2.1 and
>mentioned
>features which we plan for 2.2.
>
>Regards,
>Yury
>
>
>
>--
>View this message in context:
>http://apache-ignite-developers.2346864.n4.nabble.com/Ignite-ML-status-update-tp19829p19875.html
>Sent from the Apache Ignite Developers mailing list archive at
>Nabble.com.