![\"\"](\"https://media.springernature.com/lw900/springer-static/esm/art%3A10.1038%2Fs41586-018-0438-y/MediaObjects/41586_2018_438_Fig4_ESM.jpg\")
I found a paper in Nature Journal Deep learning of aftershock patterns following large earthquakes .
\n\nIt is different to what has been asked in this competition to predict but may be useful for generating ideas (in the paper they are predicting the probability on geospatial data).
\n\nI found their method pretty interesting. Excerpt from the paper
\n\n\n\n\nNegative training data (grid cells without aftershocks) are downsampled during training. During training, 10% of the positive training data samples were used for validation, along with an equal number of randomly selected negative samples.
\n
https://www.nature.com/articles/s41586-018-0438-y#Sec12
\n\nA free version can be read here https://www.researchgate.net/publication/327287427_Deep_learning_of_aftershock_patterns_following_large_earthquakes
\n\n
I found a paper in Nature Journal Deep learning of aftershock patterns following large earthquakes .
\n\nIt is different to what has been asked in this competition to predict but may be useful for generating ideas (in the paper they are predicting the probability on geospatial data).
\n\nI found their method pretty interesting. Excerpt from the paper
\n\n\n\n\nNegative training data (grid cells without aftershocks) are downsampled during training. During training, 10% of the positive training data samples were used for validation, along with an equal number of randomly selected negative samples.
\n
https://www.nature.com/articles/s41586-018-0438-y#Sec12
\n\nA free version can be read here https://www.researchgate.net/publication/327287427_Deep_learning_of_aftershock_patterns_following_large_earthquakes
\n\n
What do you find interesting in this paper, except for the topic?
", "rawMarkdown": "What do you find interesting in this paper, except for the topic?", "votes": null }, { "id": "534123", "postDate": "05/20/2019 18:10:36", "content": "Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.
", "rawMarkdown": "Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.", "votes": null }, { "id": "534128", "postDate": "05/20/2019 18:13:08", "content": "I wonder what would happen if I was down voting all questions people ask me...
", "rawMarkdown": "I wonder what would happen if I was down voting all questions people ask me...", "votes": null }, { "id": "534169", "postDate": "05/20/2019 20:26:26", "content": "> Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.
\n\n@cpmpml Cher Monsieur, It wasn't me who downvoted your post (I upvoted all your posts now just to make my point). I would leave the upvote and downvote thingy here.
\n\n> I wonder what would happen if I was down voting all questions people ask me…
\n\nYou are undoubtedly one of the most important asset to Kaggle community. Along the years you have helped many improve their tiers in multiple categories, including me.
\n\n> What do you find interesting in this paper, except for the topic?
\n\nI mentioned a tiny excerpt from the paper I found interesting. I will write a summary if time allows, didn't get much time to elaborate it further, it is very good work done by researchers at Google and Harvard.
\n\nI feel you wanted me to be more specific. I was thinking though people who are interested in it further should read the paper itself. If it is not highest h index journal I would hardly read it but in this case it's not, so worth time and effort I am sure there is a lot to learn such as they have compared/benchmarked NN to some domain specific tools they use to predict earthquakes.
\n\nJ'espère que ça aide
\n\nMerci
", "rawMarkdown": "> Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.\n\n@cpmpml Cher Monsieur, It wasn't me who downvoted your post (I upvoted all your posts now just to make my point). I would leave the upvote and downvote thingy here.\n\n> I wonder what would happen if I was down voting all questions people ask me…\n\nYou are undoubtedly one of the most important asset to Kaggle community. Along the years you have helped many improve their tiers in multiple categories, including me.\n\n> What do you find interesting in this paper, except for the topic?\n\nI mentioned a tiny excerpt from the paper I found interesting. I will write a summary if time allows, didn't get much time to elaborate it further, it is very good work done by researchers at Google and Harvard. \n\nI feel you wanted me to be more specific. I was thinking though people who are interested in it further should read the paper itself. If it is not highest h index journal I would hardly read it but in this case it's not, so worth time and effort I am sure there is a lot to learn such as they have compared/benchmarked NN to some domain specific tools they use to predict earthquakes. \n\nJ'espère que ça aide\n\nMerci", "votes": null }, { "id": "534200", "postDate": "05/20/2019 22:50:38", "content": "Don’t be too sensitive with downvoting. I did not downvote you. But that’s totally normal. Just live with it.
", "rawMarkdown": "Don’t be too sensitive with downvoting. I did not downvote you. But that’s totally normal. Just live with it.", "votes": null }, { "id": "534387", "postDate": "05/21/2019 06:54:24", "content": "\n\n\nIt wasn't me who downvoted your post
\n
I didn't say it was you aither ;)
\n\n\n\n\nDon’t be too sensitive with downvoting
\n
I am not. But I am like everyone, reward impacts future actions. If I get down vote for asking questions or for answering questions (in another topic discussion) then why would I keep answering questions? I come here for the fun, if there is no fun anymore then I won't come back. I think everyone would react the same way.
\n\nI'll focus on winning this competition from now one and wish good luck to all.
", "rawMarkdown": "> It wasn't me who downvoted your post \n\nI didn't say it was you aither ;)\n\n> Don’t be too sensitive with downvoting\n\nI am not. But I am like everyone, reward impacts future actions. If I get down vote for asking questions or for answering questions (in another topic discussion) then why would I keep answering questions? I come here for the fun, if there is no fun anymore then I won't come back. I think everyone would react the same way.\n\nI'll focus on winning this competition from now one and wish good luck to all.", "votes": null }, { "id": "534395", "postDate": "05/21/2019 07:20:10", "content": "Wow. From this, I just want to thank you. You made me think: “age is just a number”. although I am now much younger than you :). However in a large community, you can’t expect everyone to react the way you want. That’s impossible. Anyway, good luck and wish you the best.
", "rawMarkdown": "Wow. From this, I just want to thank you. You made me think: “age is just a number”. although I am now much younger than you :). However in a large community, you can’t expect everyone to react the way you want. That’s impossible. Anyway, good luck and wish you the best.", "votes": null }, { "id": "534421", "postDate": "05/21/2019 08:14:54", "content": "\n\nDon't worry, someone once said that you can assess your importance by the number of your enemies. Given I'm the top down voted person on Kaggle I feel very important;)
\n\n@cyberia ,
\n\nthanks for the answer.
", "rawMarkdown": "khahuras ,\n\nDon't worry, someone once said that you can assess your importance by the number of your enemies. Given I'm the top down voted person on Kaggle I feel very important;)\n\n@cyberia ,\n\nthanks for the answer.", "votes": null }, { "id": "534499", "postDate": "05/21/2019 10:51:47", "content": "Nice approach to visualize different sample patterns and it's score. I am looking forward to this.
", "rawMarkdown": "Nice approach to visualize different sample patterns and it's score. I am looking forward to this.", "votes": null }, { "id": "534953", "postDate": "05/22/2019 05:39:48", "content": "Thanks for sharing this information.
", "rawMarkdown": "Thanks for sharing this information.", "votes": null } ], "comments": [ { "id": 534061, "author_name": "cpmpml", "author_url": "", "post_date": "05/20/2019 15:01:47", "content": "What do you find interesting in this paper, except for the topic?
", "votes": null, "replies": [ { "id": 534123, "author_name": "cpmpml", "author_url": "", "post_date": "05/20/2019 18:10:36", "content": "Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.
", "votes": null, "replies": [] }, { "id": 534128, "author_name": "cpmpml", "author_url": "", "post_date": "05/20/2019 18:13:08", "content": "I wonder what would happen if I was down voting all questions people ask me...
", "votes": null, "replies": [] }, { "id": 534169, "author_name": "cyberia", "author_url": "", "post_date": "05/20/2019 20:26:26", "content": "> Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.
\n\n@cpmpml Cher Monsieur, It wasn't me who downvoted your post (I upvoted all your posts now just to make my point). I would leave the upvote and downvote thingy here.
\n\n> I wonder what would happen if I was down voting all questions people ask me…
\n\nYou are undoubtedly one of the most important asset to Kaggle community. Along the years you have helped many improve their tiers in multiple categories, including me.
\n\n> What do you find interesting in this paper, except for the topic?
\n\nI mentioned a tiny excerpt from the paper I found interesting. I will write a summary if time allows, didn't get much time to elaborate it further, it is very good work done by researchers at Google and Harvard.
\n\nI feel you wanted me to be more specific. I was thinking though people who are interested in it further should read the paper itself. If it is not highest h index journal I would hardly read it but in this case it's not, so worth time and effort I am sure there is a lot to learn such as they have compared/benchmarked NN to some domain specific tools they use to predict earthquakes.
\n\nJ'espère que ça aide
\n\nMerci
", "votes": null, "replies": [] }, { "id": 534200, "author_name": "khahuras", "author_url": "", "post_date": "05/20/2019 22:50:38", "content": "Don’t be too sensitive with downvoting. I did not downvote you. But that’s totally normal. Just live with it.
", "votes": null, "replies": [] }, { "id": 534387, "author_name": "cpmpml", "author_url": "", "post_date": "05/21/2019 06:54:24", "content": "\n\n\nIt wasn't me who downvoted your post
\n
I didn't say it was you aither ;)
\n\n\n\n\nDon’t be too sensitive with downvoting
\n
I am not. But I am like everyone, reward impacts future actions. If I get down vote for asking questions or for answering questions (in another topic discussion) then why would I keep answering questions? I come here for the fun, if there is no fun anymore then I won't come back. I think everyone would react the same way.
\n\nI'll focus on winning this competition from now one and wish good luck to all.
", "votes": null, "replies": [] }, { "id": 534395, "author_name": "khahuras", "author_url": "", "post_date": "05/21/2019 07:20:10", "content": "Wow. From this, I just want to thank you. You made me think: “age is just a number”. although I am now much younger than you :). However in a large community, you can’t expect everyone to react the way you want. That’s impossible. Anyway, good luck and wish you the best.
", "votes": null, "replies": [] }, { "id": 534421, "author_name": "cpmpml", "author_url": "", "post_date": "05/21/2019 08:14:54", "content": "\n\nDon't worry, someone once said that you can assess your importance by the number of your enemies. Given I'm the top down voted person on Kaggle I feel very important;)
\n\n@cyberia ,
\n\nthanks for the answer.
", "votes": null, "replies": [] } ] }, { "id": 534499, "author_name": "himusoni", "author_url": "", "post_date": "05/21/2019 10:51:47", "content": "Nice approach to visualize different sample patterns and it's score. I am looking forward to this.
", "votes": null, "replies": [] }, { "id": 534953, "author_name": "venkatakkinapalli", "author_url": "", "post_date": "05/22/2019 05:39:48", "content": "Thanks for sharing this information.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "533958": "I found a paper in Nature Journal **Deep learning of aftershock patterns following large earthquakes** . \n\nIt is different to what has been asked in this competition to predict but may be useful for generating ideas (in the paper they are predicting the probability on geospatial data).\n\nI found their method pretty interesting. Excerpt from the paper\n\n> Negative training data (grid cells without aftershocks) are downsampled during training. During training, 10% of the positive training data samples were used for validation, along with an equal number of randomly selected negative samples.\n\nhttps://www.nature.com/articles/s41586-018-0438-y#Sec12\n\nA free version can be read here https://www.researchgate.net/publication/327287427_Deep_learning_of_aftershock_patterns_following_large_earthquakes\n\n", "534061": "What do you find interesting in this paper, except for the topic?", "534123": "Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.", "534128": "I wonder what would happen if I was down voting all questions people ask me...", "534169": "> Lol, downvoted because I ask a question. I'm still interested in the answer, to decide if I read it or not.\n\n@cpmpml Cher Monsieur, It wasn't me who downvoted your post (I upvoted all your posts now just to make my point). I would leave the upvote and downvote thingy here.\n\n> I wonder what would happen if I was down voting all questions people ask me…\n\nYou are undoubtedly one of the most important asset to Kaggle community. Along the years you have helped many improve their tiers in multiple categories, including me.\n\n> What do you find interesting in this paper, except for the topic?\n\nI mentioned a tiny excerpt from the paper I found interesting. I will write a summary if time allows, didn't get much time to elaborate it further, it is very good work done by researchers at Google and Harvard. \n\nI feel you wanted me to be more specific. I was thinking though people who are interested in it further should read the paper itself. If it is not highest h index journal I would hardly read it but in this case it's not, so worth time and effort I am sure there is a lot to learn such as they have compared/benchmarked NN to some domain specific tools they use to predict earthquakes. \n\nJ'espère que ça aide\n\nMerci", "534200": "Don’t be too sensitive with downvoting. I did not downvote you. But that’s totally normal. Just live with it.", "534387": "> It wasn't me who downvoted your post \n\nI didn't say it was you aither ;)\n\n> Don’t be too sensitive with downvoting\n\nI am not. But I am like everyone, reward impacts future actions. If I get down vote for asking questions or for answering questions (in another topic discussion) then why would I keep answering questions? I come here for the fun, if there is no fun anymore then I won't come back. I think everyone would react the same way.\n\nI'll focus on winning this competition from now one and wish good luck to all.", "534395": "Wow. From this, I just want to thank you. You made me think: “age is just a number”. although I am now much younger than you :). However in a large community, you can’t expect everyone to react the way you want. That’s impossible. Anyway, good luck and wish you the best.", "534421": "khahuras ,\n\nDon't worry, someone once said that you can assess your importance by the number of your enemies. Given I'm the top down voted person on Kaggle I feel very important;)\n\n@cyberia ,\n\nthanks for the answer.", "534499": "Nice approach to visualize different sample patterns and it's score. I am looking forward to this.", "534953": "Thanks for sharing this information." }, "source": "meta" }