💊 Data Talk: Fastai plus RoBERTa

To Vaccinate or Not to Vaccinate: It’s not a Question by #ZindiWeekendz

Helping Africa

$300 USD

Completed (~6 years ago)

Skills you will learn

Natural Language Processing

Classification

Sentiment Analysis

212 joined

126 active

Info Data Chat Leaderboard

Start

Apr 24, 20

Apr 26, 20

Reveal

Apr 26, 20

Lawrence_Moruye

Fastai plus RoBERTa

Notebooks · 26 Apr 2020, 21:26 · 7

I'm using Fastai plus HuggingFace Transformer RoBERTa.The classification approach gives a poor score of 0.58.Here is the link

https://github.com/LawrenceMoruye/To-Vaccinate-or-Not-to-Vaccinate-It-s-not-a-Question-by-ZindiWeekendz

Discussion 7 answers

Aymen_Soussi

Hello @Lawrence_Moruye thanks for chearing and you did really a great work, just if this challenge was a classification one you would get a far better score in the LB, again it is clearly that this challenge need to be classification one !!

26 Apr 2020, 22:28 (edited 1 minute later)

Upvotes 0

Nice one Lawrence.

Upvotes 0

Inveniam

Instead of predicting the classes, I think you should try predicting the probabilities. And then, following the argmax, you mulitply the given probability by a coefficient of :

* -1 if argmax is 0

* O if argmax is 1

* 1 if argmax is 2

26 Apr 2020, 23:09

Upvotes 0

Muhamed_Tuo

Inveniam

I think that the evaluation function (RMSE) doesn't stop us from solving this challenge as a classification problem. As long as at the end, we submit Probabilities instead of Classes/Labels.

replied to Muhamed_Tuo26 Apr 2020, 23:14

Upvotes 0