Hi,
\n\n37ish is not too glamoourous but we worked very hard and I wanted to share our aproach :)\nOur solution was based mainly in GNN which scored -2.317 provided by our team mate https://www.kaggle.com/jessky and some lgb ensemble that scores a bit less than -2.1. Then our best nnet (bagged) scores almost -1.9. For lgb and nnet models we benefit a lot from oof predictions of FC. We kept adding different prediction and the score improved by ~ -0.4 just adding those. Besides some ad-hoc FE, we obviously made use of all the amazing distances and amazing calculations available in kernels. Mainly: acsf descriptors and the sergii distances ( https://www.kaggle.com/criskiev/distance-is-all-you-need-lb-1-481)
\n\nOur final ensemble can be seen in here: https://www.kaggle.com/chechir/champs-ensemble \nThanks to the organisers, and all the awesome kernel authors in this competition !
", "messages": [ { "id": "611532", "postDate": "08/29/2019 10:05:19", "content": "Hi,
\n\n37ish is not too glamoourous but we worked very hard and I wanted to share our aproach :)\nOur solution was based mainly in GNN which scored -2.317 provided by our team mate https://www.kaggle.com/jessky and some lgb ensemble that scores a bit less than -2.1. Then our best nnet (bagged) scores almost -1.9. For lgb and nnet models we benefit a lot from oof predictions of FC. We kept adding different prediction and the score improved by ~ -0.4 just adding those. Besides some ad-hoc FE, we obviously made use of all the amazing distances and amazing calculations available in kernels. Mainly: acsf descriptors and the sergii distances ( https://www.kaggle.com/criskiev/distance-is-all-you-need-lb-1-481)
\n\nOur final ensemble can be seen in here: https://www.kaggle.com/chechir/champs-ensemble \nThanks to the organisers, and all the awesome kernel authors in this competition !
", "rawMarkdown": "Hi, \n\n37ish is not too glamoourous but we worked very hard and I wanted to share our aproach :)\nOur solution was based mainly in GNN which scored -2.317 provided by our team mate https://www.kaggle.com/jessky and some lgb ensemble that scores a bit less than -2.1. Then our best nnet (bagged) scores almost -1.9. For lgb and nnet models we benefit a lot from oof predictions of FC. We kept adding different prediction and the score improved by ~ -0.4 just adding those. Besides some ad-hoc FE, we obviously made use of all the amazing distances and amazing calculations available in kernels. Mainly: acsf descriptors and the sergii distances ( https://www.kaggle.com/criskiev/distance-is-all-you-need-lb-1-481) \n\nOur final ensemble can be seen in here: https://www.kaggle.com/chechir/champs-ensemble \nThanks to the organisers, and all the awesome kernel authors in this competition !", "votes": null }, { "id": "611546", "postDate": "08/29/2019 10:12:23", "content": "Congrats\nGreat Work...\nInformative..\nThanks for Sharing...!! @chechir
", "rawMarkdown": "Congrats\nGreat Work...\nInformative..\nThanks for Sharing...!! @chechir", "votes": null }, { "id": "611547", "postDate": "08/29/2019 10:12:38", "content": "Thanks for sharing. Good to know how much FC preds helped!
", "rawMarkdown": "Thanks for sharing. Good to know how much FC preds helped!", "votes": null }, { "id": "611558", "postDate": "08/29/2019 10:19:00", "content": "\"and some lgb ensemble that scores a over -3.1.\" Is that supposed to be -2.1?
", "rawMarkdown": "\"and some lgb ensemble that scores a over -3.1.\" Is that supposed to be -*2*.1?", "votes": null }, { "id": "611569", "postDate": "08/29/2019 10:22:25", "content": "oh yes, sorry! -2.1. I wish it was -3.1 :P
", "rawMarkdown": "oh yes, sorry! -2.1. I wish it was -3.1 :P", "votes": null } ], "comments": [ { "id": 611546, "author_name": "veeralakrishna", "author_url": "", "post_date": "08/29/2019 10:12:23", "content": "Congrats\nGreat Work...\nInformative..\nThanks for Sharing...!! @chechir
", "votes": null, "replies": [] }, { "id": 611547, "author_name": "taniaj", "author_url": "", "post_date": "08/29/2019 10:12:38", "content": "Thanks for sharing. Good to know how much FC preds helped!
", "votes": null, "replies": [] }, { "id": 611558, "author_name": "abendorf", "author_url": "", "post_date": "08/29/2019 10:19:00", "content": "\"and some lgb ensemble that scores a over -3.1.\" Is that supposed to be -2.1?
", "votes": null, "replies": [ { "id": 611569, "author_name": "chechir", "author_url": "", "post_date": "08/29/2019 10:22:25", "content": "oh yes, sorry! -2.1. I wish it was -3.1 :P
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "611532": "Hi, \n\n37ish is not too glamoourous but we worked very hard and I wanted to share our aproach :)\nOur solution was based mainly in GNN which scored -2.317 provided by our team mate https://www.kaggle.com/jessky and some lgb ensemble that scores a bit less than -2.1. Then our best nnet (bagged) scores almost -1.9. For lgb and nnet models we benefit a lot from oof predictions of FC. We kept adding different prediction and the score improved by ~ -0.4 just adding those. Besides some ad-hoc FE, we obviously made use of all the amazing distances and amazing calculations available in kernels. Mainly: acsf descriptors and the sergii distances ( https://www.kaggle.com/criskiev/distance-is-all-you-need-lb-1-481) \n\nOur final ensemble can be seen in here: https://www.kaggle.com/chechir/champs-ensemble \nThanks to the organisers, and all the awesome kernel authors in this competition !", "611546": "Congrats\nGreat Work...\nInformative..\nThanks for Sharing...!! @chechir", "611547": "Thanks for sharing. Good to know how much FC preds helped!", "611558": "\"and some lgb ensemble that scores a over -3.1.\" Is that supposed to be -*2*.1?", "611569": "oh yes, sorry! -2.1. I wish it was -3.1 :P" }, "source": "meta" }