# Tournament NN baseline with the new massive data

**URL:** <https://forum.numer.ai/t/tournament-nn-baseline-with-the-new-massive-data/4305>\
**Category:** Tournament\
**Created:** [October 10, 2021, 12:04pm UTC](https://forum.numer.ai/t/tournament-nn-baseline-with-the-new-massive-data/4305 "2021-10-10T12:04:05Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![katsu1110](https://yyz1.discourse-cdn.com/flex009/user_avatar/forum.numer.ai/katsu1110/32/2830_2.png) [@katsu1110](https://forum.numer.ai/u/katsu1110)\
**Post date:** [October 10, 2021, 12:04pm UTC](https://forum.numer.ai/t/tournament-nn-baseline-with-the-new-massive-data/4305/1 "2021-10-10T12:04:05Z")

</div>

Hi, I published an end-to-end (from data loading to submission) kaggle notebook for the Numerai Tournament. Although a kaggle notebook has only ca. 16GB RAM, a bit of tricks make it possible to use the massive data in it.

This notebook uses a multi-layer perceptron with multi-outputs to take advantage of multiple targets of the new data.

[[Numerai] NN baseline with new massive data](https://www.kaggle.com/code1110/numerai-nn-baseline-with-new-massive-data)

Hopefully this notebook helps those who have troubles with dealing with the massive data. Have fun!

---

<div class="post-metadata">

**Author:** ![autratec](https://avatars.discourse-cdn.com/v4/letter/a/6bbea6/32.png) [@autratec](https://forum.numer.ai/u/autratec)\
**Post date:** [October 18, 2021, 11:25am UTC](https://forum.numer.ai/t/tournament-nn-baseline-with-the-new-massive-data/4305/2 "2021-10-18T11:25:14Z")

</div>

Katsu1110 san, thanks for sharing the source code again. Btw if feel using xgb will be more reliable than NN.

---

<div class="post-metadata">

**Author:** ![katsu1110](https://yyz1.discourse-cdn.com/flex009/user_avatar/forum.numer.ai/katsu1110/32/2830_2.png) [@katsu1110](https://forum.numer.ai/u/katsu1110)\
**Post date:** [October 18, 2021, 2:24pm UTC](https://forum.numer.ai/t/tournament-nn-baseline-with-the-new-massive-data/4305/3 "2021-10-18T14:24:52Z")

</div>

Thanks for your feedback. It is nice for Numerai to have different model predictions, so go for your idea!

---

<div class="post-metadata">

**Author:** ![rdr91h](https://yyz1.discourse-cdn.com/flex009/user_avatar/forum.numer.ai/rdr91h/32/1520_2.png) [@rdr91h](https://forum.numer.ai/u/rdr91h)\
**Post date:** [October 19, 2021, 4:52pm UTC](https://forum.numer.ai/t/tournament-nn-baseline-with-the-new-massive-data/4305/4 "2021-10-19T16:52:08Z")

</div>

The two top numerai signals models are based in NN