Skip to content

Taeu/NAVER-AI-RUSH-2019

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 

Repository files navigation

NAVER-AI-RUSH


  • 12 th (total score)

NAVER-AI-RUSH-1 / Image Classification


  • our solution : efficientNet + oct vNet Ensemble

  • 17 th / 100 teams (first round)

  • the period actually participate : Aug 6 to Aug 13 (about 7 days)

NAVER-AI-RUSH-2 / Click-Through Rate (CTR) Prediction


  • our solution : A single CAT model, (only use 7 different features : read_len, read_cnt, total_cnt, read_prob, gender, age_range, hh)

  • 11 th / 30 teams (final round)

  • the period actually participate : Aug 20 to Aug 28 (about 8 days)

  • tried : xDeepFM (feature, article_id, read_len, gender, age_range, hh, image_feature) - didn't go well..

  • tried : Embedding DNN network - didn't see it carefully..

  • tried : xgboost, lgbm for a lot of feature that we could make, (image_feature, catergory_id, cat_in, feature cross, means, std for row numeric values.. etc.. ) - trained well, got better score than our final submit score. But couldn't submit some memory issues, something wrong in features preprocess for test set..

  • Things we couldn't do because of not enough time :

    • submit for image feature, category feature, means, std of numeric cols that we made at training
    • Ensemble different models(xgboost, lgbm and CAT). even couldn't ensemble of result of k-fold of just single CAT, other model, respectively.
    • train item2vec of read articles id (history).. ( i think if we did this and make article_id feature, then we could get very good feature for the model... we just thought but didn't try even though it's not that difficult)

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages