Interesting question, just one of the few you can find about Kaggle in here.
\n\nI tried to look for it in Quora, but no luck.
\n\nPS: Questions with Kaggle in them are pretty safe.
", "messages": [ { "id": "417873", "postDate": "11/08/2018 22:57:45", "content": "Interesting question, just one of the few you can find about Kaggle in here.
\n\nI tried to look for it in Quora, but no luck.
\n\nPS: Questions with Kaggle in them are pretty safe.
", "rawMarkdown": "Interesting question, just one of the few you can find about Kaggle in [here][1].\n\nI tried to look for it in Quora, but no luck.\n\nPS: Questions with Kaggle in them are pretty safe.\n\n\n [1]: https://www.kaggle.com/carloshuertas/quora-and-kaggle", "votes": null }, { "id": "417953", "postDate": "11/09/2018 03:44:27", "content": "What I have learned:
\n\nPractical experience of applying machine learning and deep learning algorithms.
Pipelines of data cleaning, model training/hyper-parameter tuning and ensemble for various types of problems.
What I want to learn in the future:
\n\nWhat I have learnt:\n Everything practical I have learnt and applied so far, which includes EDA, Visualization, feature engineering intuitions, feature selection, creating pipelines, hyper parameter tuning, validation approaches.
", "rawMarkdown": "What I have learnt:\n Everything practical I have learnt and applied so far, which includes EDA, Visualization, feature engineering intuitions, feature selection, creating pipelines, hyper parameter tuning, validation approaches.", "votes": null }, { "id": "418078", "postDate": "11/09/2018 08:58:19", "content": "Learning By Doing is the best thing about Kaggle. Solving problems while learning is the best way to solidify the concepts that we are trying to learn. For example, It would always be easier for a guy to fix a car who knows how to drive rather than a guy who just knows how a car works but have never actually driven one. Maybe a bad analogy but I hope you get what I am saying.
", "rawMarkdown": "**Learning By Doing** is the best thing about Kaggle. Solving problems while learning is the best way to solidify the concepts that we are trying to learn. For example, It would always be easier for a guy to fix a car who knows how to drive rather than a guy who just knows how a car works but have never actually driven one. Maybe a bad analogy but I hope you get what I am saying.", "votes": null }, { "id": "418406", "postDate": "11/09/2018 21:32:17", "content": "Two things among all, in every competition I took part up so far (they are about 132):\n 1. best practices of applying machine learning to real world problems from all competitors\n 2. bleeding edge techniques from top ones
", "rawMarkdown": "Two things among all, in every competition I took part up so far (they are about 132):\n 1. best practices of applying machine learning to real world problems from all competitors\n 2. bleeding edge techniques from top ones", "votes": null }, { "id": "418471", "postDate": "11/10/2018 00:33:55", "content": "I learn ML through kaggle.
\n\nStarting with the basic concepts like: CV, random forest and feature engineering.\nthen hyper param optimisation and boosting.\nNow each competition I try to learn at least one new subject.\nIn this competition I'm going over the functional API of keras and embedding vectors.
", "rawMarkdown": "I learn ML through kaggle.\n\nStarting with the basic concepts like: CV, random forest and feature engineering.\nthen hyper param optimisation and boosting.\nNow each competition I try to learn at least one new subject.\nIn this competition I'm going over the functional API of keras and embedding vectors.", "votes": null }, { "id": "418530", "postDate": "11/10/2018 03:38:51", "content": "I love these questions:
\n\n'How would researchers like Yann Lecun and the like do in a Kaggle competition?'\nA great answer was given by Geoff Hinton and his team who crushed the first and only Kaggle competition they entered: http://blog.kaggle.com/2012/11/01/deep-learning-how-i-did-it-merck-1st-place-interview/
\n\n'Will Kaggle cease to exist when Auto-ML comes into full force?'
\n\nThe trick is 'when'. It could take decades before ML automation makes top data scientists useless.
", "rawMarkdown": "I love these questions:\n\n'How would researchers like Yann Lecun and the like do in a Kaggle competition?'\nA great answer was given by Geoff Hinton and his team who crushed the first and only Kaggle competition they entered: http://blog.kaggle.com/2012/11/01/deep-learning-how-i-did-it-merck-1st-place-interview/\n\n'Will Kaggle cease to exist when Auto-ML comes into full force?'\n\nThe trick is 'when'. It could take decades before ML automation makes top data scientists useless.", "votes": null }, { "id": "418544", "postDate": "11/10/2018 04:51:58", "content": "Oh.... those days without Kaggle Kernels :)
", "rawMarkdown": "Oh.... those days without Kaggle Kernels :)", "votes": null }, { "id": "418863", "postDate": "11/10/2018 18:49:51", "content": "I have learnt : Fork, Commit, Submit, Don't Upvote, Climb the LB, Get a medal
Update: I forgot... Don't get tracked down by GM NxGTR eagle eyes
In that case, we find Hinton's account: https://www.kaggle.com/jeff20
", "rawMarkdown": "In that case, we find Hinton's account: https://www.kaggle.com/jeff20", "votes": null }, { "id": "418901", "postDate": "11/10/2018 20:45:35", "content": "I'm always amazed when someone comes out of nowhere and finishes in the money. Here are some more good questions from the Quora site:
\n\nLinks to these questions and their answers at this topic: https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70867
", "rawMarkdown": "I'm always amazed when someone comes out of nowhere and finishes in the money. Here are some more good questions from the Quora site:\n\n - Is Kaggle dead?\n - Does winning a Kaggle competition matter outside of Kaggle?\n - What does it feel like to be addicted to Kaggle?\n - Why do most kaggle users use various ML and DL frameworks and libraries instead of implementing algorithms from scratch? Are they all incompetent?\n\nLinks to these questions and their answers at this topic: https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70867", "votes": null }, { "id": "418934", "postDate": "11/10/2018 22:48:53", "content": "I have learned that 95% of methods and techniques described in academic papers are not applicable to real life (/kaggle) data sets.
\n\nAnd for the 5% that do work, well, you need to find out if they actually work the hard way - By trying it out. There is no free lunch.
", "rawMarkdown": "I have learned that 95% of methods and techniques described in academic papers are not applicable to real life (/kaggle) data sets.\n\nAnd for the 5% that do work, well, you need to find out if they actually work the hard way - By trying it out. There is no free lunch.", "votes": null }, { "id": "418935", "postDate": "11/10/2018 23:01:38", "content": "It is a great environment to practice ML algorithms on real problems it improved my ML practical skills as you learn by doing
", "rawMarkdown": "It is a great environment to practice ML algorithms on real problems it improved my ML practical skills as you learn by doing", "votes": null }, { "id": "418936", "postDate": "11/10/2018 23:04:26", "content": "When i started competing on kaggle i wasn't able to create a simple neural network from scratch, I started with simple problems and by the help of kernels and practicing i learned too much so i can now build my own models, So i worked on other problems so i improved my skills in frameworks like tensorflow and keras also i engaged with problems categories i haven't worked with before like segmentation and NLP problems.
", "rawMarkdown": "When i started competing on kaggle i wasn't able to create a simple neural network from scratch, I started with simple problems and by the help of kernels and practicing i learned too much so i can now build my own models, So i worked on other problems so i improved my skills in frameworks like tensorflow and keras also i engaged with problems categories i haven't worked with before like segmentation and NLP problems.", "votes": null }, { "id": "419300", "postDate": "11/11/2018 17:05:00", "content": "Kaggle has helped me tremendously by helping me practice, practice, practice. Of course a lot of learning from others who publish top notch kernels/topics.
", "rawMarkdown": "Kaggle has helped me tremendously by helping me practice, practice, practice. Of course a lot of learning from others who publish top notch kernels/topics.", "votes": null }, { "id": "419313", "postDate": "11/11/2018 17:34:42", "content": "Writing reproducible code and logging everything is extremely important.
", "rawMarkdown": "Writing reproducible code and logging everything is extremely important.", "votes": null }, { "id": "419336", "postDate": "11/11/2018 18:35:05", "content": "I shared few things I learned here: https://www.kaggle.com/c/PLAsTiCC-2018/discussion/70908
\n\nSome are specific to the Plasticc competition, but most are pretty generic.
", "rawMarkdown": "I shared few things I learned here: https://www.kaggle.com/c/PLAsTiCC-2018/discussion/70908\n\nSome are specific to the Plasticc competition, but most are pretty generic.", "votes": null }, { "id": "419549", "postDate": "11/12/2018 07:08:41", "content": "I read the book 'American Gods' this summer. It talks about the old gods like Odin, Ra, Zeus, etc and the modern gods which are the internet, the television, money, etc. Well, in Kaggle, I met the ML gods: CPMP, NxGTR, SRK, Kazanova, Olivier, Bojan, etc...Respect the ML gods - they might just have turned you into what you are today, or will be tomorrow.
", "rawMarkdown": "I read the book 'American Gods' this summer. It talks about the old gods like Odin, Ra, Zeus, etc and the modern gods which are the internet, the television, money, etc. Well, in Kaggle, I met the ML gods: CPMP, NxGTR, SRK, Kazanova, Olivier, Bojan, etc...Respect the ML gods - they might just have turned you into what you are today, or will be tomorrow.", "votes": null }, { "id": "419565", "postDate": "11/12/2018 08:02:37", "content": "You have low standards for Gods hahaha :P
", "rawMarkdown": "You have low standards for Gods hahaha :P", "votes": null }, { "id": "419570", "postDate": "11/12/2018 08:10:38", "content": "It is a paradox but it is actually true: The normal thing to do is to first dig into data, perform EDA to get some insight and then apply a model of some kind on those data. But being eager to see a model 'alive', most of us, as newbies, first copied and deployed a model to see a rank on the leaderboard and then started gaining actual knowledge on the data we are working with!
", "rawMarkdown": "It is a paradox but it is actually true: The normal thing to do is to first dig into data, perform EDA to get some insight and then apply a model of some kind on those data. But being eager to see a model 'alive', most of us, as newbies, first copied and deployed a model to see a rank on the leaderboard and then started gaining actual knowledge on the data we are working with!", "votes": null }, { "id": "419642", "postDate": "11/12/2018 10:54:44", "content": "@NxGTR I have to agree I'm not a God by any standards. I'm too incompetent to grade the others listed ;-)
", "rawMarkdown": "NxGTR I have to agree I'm not a God by any standards. I'm too incompetent to grade the others listed ;-)", "votes": null }, { "id": "419665", "postDate": "11/12/2018 11:21:51", "content": "Amen ;)
\n\nI think there are stronger ML gods than me that you did not cite, like giba, bestfitting, etc. But thanks a lot, very honored to be seen as a god! May the strength be with you!
", "rawMarkdown": "Amen ;)\n\nI think there are stronger ML gods than me that you did not cite, like giba, bestfitting, etc. But thanks a lot, very honored to be seen as a god! May the strength be with you!", "votes": null }, { "id": "419701", "postDate": "11/12/2018 12:25:11", "content": "Ahhh yeah Giba! He found the leak in Sandanter Challenge (among other great things). Yes Kudos also. But I was not meaning to make an exhaustive list here so apologies to all altrouistic contributors that I didn't mention...
", "rawMarkdown": "Ahhh yeah Giba! He found the leak in Sandanter Challenge (among other great things). Yes Kudos also. But I was not meaning to make an exhaustive list here so apologies to all altrouistic contributors that I didn't mention...", "votes": null }, { "id": "419713", "postDate": "11/12/2018 12:51:01", "content": "I found a new motivation now. Why be a grandmaster when you can be a god?
", "rawMarkdown": "I found a new motivation now. Why be a grandmaster when you can be a god?", "votes": null }, { "id": "419885", "postDate": "11/12/2018 17:45:42", "content": "This was interesting to read.
", "rawMarkdown": "This was interesting to read.", "votes": null }, { "id": "420028", "postDate": "11/13/2018 00:08:35", "content": "Absolutely true in my case :-)
", "rawMarkdown": "Absolutely true in my case :-)", "votes": null }, { "id": "420123", "postDate": "11/13/2018 05:09:05", "content": "I do some EDA myself, but I want to learn to do it in a more systematic way.
", "rawMarkdown": "I do some EDA myself, but I want to learn to do it in a more systematic way.", "votes": null }, { "id": "420644", "postDate": "11/13/2018 23:50:46", "content": "\"Doing is better than reading\"
\n\nWhat you learn by doing it yourself can't be replaced by anything. No matter how many books/articles/blogs you read about Data Science but the way you understand and retain concepts while doing it on your own is simply amazing. I have been into Analytics Consulting, however, what I learned within 6 months of Kaggling is simply beyond words.
", "rawMarkdown": "\"Doing is better than reading\"\n\nWhat you learn by doing it yourself can't be replaced by anything. No matter how many books/articles/blogs you read about Data Science but the way you understand and retain concepts while doing it on your own is simply amazing. I have been into Analytics Consulting, however, what I learned within 6 months of Kaggling is simply beyond words.", "votes": null } ], "comments": [ { "id": 417953, "author_name": "naivelamb", "author_url": "", "post_date": "11/09/2018 03:44:27", "content": "What I have learned:
\n\nPractical experience of applying machine learning and deep learning algorithms.
Pipelines of data cleaning, model training/hyper-parameter tuning and ensemble for various types of problems.
What I want to learn in the future:
\n\nIt is a paradox but it is actually true: The normal thing to do is to first dig into data, perform EDA to get some insight and then apply a model of some kind on those data. But being eager to see a model 'alive', most of us, as newbies, first copied and deployed a model to see a rank on the leaderboard and then started gaining actual knowledge on the data we are working with!
", "votes": null, "replies": [] }, { "id": 420028, "author_name": "ggopalan", "author_url": "", "post_date": "11/13/2018 00:08:35", "content": "Absolutely true in my case :-)
", "votes": null, "replies": [] }, { "id": 420123, "author_name": "naivelamb", "author_url": "", "post_date": "11/13/2018 05:09:05", "content": "I do some EDA myself, but I want to learn to do it in a more systematic way.
", "votes": null, "replies": [] } ] }, { "id": 418061, "author_name": "karthikdutt", "author_url": "", "post_date": "11/09/2018 08:11:13", "content": "What I have learnt:\n Everything practical I have learnt and applied so far, which includes EDA, Visualization, feature engineering intuitions, feature selection, creating pipelines, hyper parameter tuning, validation approaches.
", "votes": null, "replies": [] }, { "id": 418078, "author_name": "satian", "author_url": "", "post_date": "11/09/2018 08:58:19", "content": "Learning By Doing is the best thing about Kaggle. Solving problems while learning is the best way to solidify the concepts that we are trying to learn. For example, It would always be easier for a guy to fix a car who knows how to drive rather than a guy who just knows how a car works but have never actually driven one. Maybe a bad analogy but I hope you get what I am saying.
", "votes": null, "replies": [] }, { "id": 418406, "author_name": "lucamassaron", "author_url": "", "post_date": "11/09/2018 21:32:17", "content": "Two things among all, in every competition I took part up so far (they are about 132):\n 1. best practices of applying machine learning to real world problems from all competitors\n 2. bleeding edge techniques from top ones
", "votes": null, "replies": [] }, { "id": 418471, "author_name": "mrbeer", "author_url": "", "post_date": "11/10/2018 00:33:55", "content": "I learn ML through kaggle.
\n\nStarting with the basic concepts like: CV, random forest and feature engineering.\nthen hyper param optimisation and boosting.\nNow each competition I try to learn at least one new subject.\nIn this competition I'm going over the functional API of keras and embedding vectors.
", "votes": null, "replies": [] }, { "id": 418530, "author_name": "cpmpml", "author_url": "", "post_date": "11/10/2018 03:38:51", "content": "I love these questions:
\n\n'How would researchers like Yann Lecun and the like do in a Kaggle competition?'\nA great answer was given by Geoff Hinton and his team who crushed the first and only Kaggle competition they entered: http://blog.kaggle.com/2012/11/01/deep-learning-how-i-did-it-merck-1st-place-interview/
\n\n'Will Kaggle cease to exist when Auto-ML comes into full force?'
\n\nThe trick is 'when'. It could take decades before ML automation makes top data scientists useless.
", "votes": null, "replies": [ { "id": 418544, "author_name": "carloshuertas", "author_url": "", "post_date": "11/10/2018 04:51:58", "content": "Oh.... those days without Kaggle Kernels :)
", "votes": null, "replies": [] }, { "id": 418881, "author_name": "shujian", "author_url": "", "post_date": "11/10/2018 19:33:00", "content": "In that case, we find Hinton's account: https://www.kaggle.com/jeff20
", "votes": null, "replies": [] }, { "id": 418901, "author_name": "jpmiller", "author_url": "", "post_date": "11/10/2018 20:45:35", "content": "I'm always amazed when someone comes out of nowhere and finishes in the money. Here are some more good questions from the Quora site:
\n\nLinks to these questions and their answers at this topic: https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70867
", "votes": null, "replies": [] }, { "id": 419885, "author_name": "gonnel", "author_url": "", "post_date": "11/12/2018 17:45:42", "content": "This was interesting to read.
", "votes": null, "replies": [] } ] }, { "id": 418863, "author_name": "ogrellier", "author_url": "", "post_date": "11/10/2018 18:49:51", "content": "I have learnt : Fork, Commit, Submit, Don't Upvote, Climb the LB, Get a medal
Update: I forgot... Don't get tracked down by GM NxGTR eagle eyes
I have learned that 95% of methods and techniques described in academic papers are not applicable to real life (/kaggle) data sets.
\n\nAnd for the 5% that do work, well, you need to find out if they actually work the hard way - By trying it out. There is no free lunch.
", "votes": null, "replies": [] }, { "id": 418935, "author_name": "mohamedramzy", "author_url": "", "post_date": "11/10/2018 23:01:38", "content": "It is a great environment to practice ML algorithms on real problems it improved my ML practical skills as you learn by doing
", "votes": null, "replies": [] }, { "id": 418936, "author_name": "mohamedramzy", "author_url": "", "post_date": "11/10/2018 23:04:26", "content": "When i started competing on kaggle i wasn't able to create a simple neural network from scratch, I started with simple problems and by the help of kernels and practicing i learned too much so i can now build my own models, So i worked on other problems so i improved my skills in frameworks like tensorflow and keras also i engaged with problems categories i haven't worked with before like segmentation and NLP problems.
", "votes": null, "replies": [] }, { "id": 419300, "author_name": "ggopalan", "author_url": "", "post_date": "11/11/2018 17:05:00", "content": "Kaggle has helped me tremendously by helping me practice, practice, practice. Of course a lot of learning from others who publish top notch kernels/topics.
", "votes": null, "replies": [] }, { "id": 419313, "author_name": "artgor", "author_url": "", "post_date": "11/11/2018 17:34:42", "content": "Writing reproducible code and logging everything is extremely important.
", "votes": null, "replies": [] }, { "id": 419336, "author_name": "cpmpml", "author_url": "", "post_date": "11/11/2018 18:35:05", "content": "I shared few things I learned here: https://www.kaggle.com/c/PLAsTiCC-2018/discussion/70908
\n\nSome are specific to the Plasticc competition, but most are pretty generic.
", "votes": null, "replies": [] }, { "id": 419549, "author_name": "georsara1", "author_url": "", "post_date": "11/12/2018 07:08:41", "content": "I read the book 'American Gods' this summer. It talks about the old gods like Odin, Ra, Zeus, etc and the modern gods which are the internet, the television, money, etc. Well, in Kaggle, I met the ML gods: CPMP, NxGTR, SRK, Kazanova, Olivier, Bojan, etc...Respect the ML gods - they might just have turned you into what you are today, or will be tomorrow.
", "votes": null, "replies": [ { "id": 419565, "author_name": "carloshuertas", "author_url": "", "post_date": "11/12/2018 08:02:37", "content": "You have low standards for Gods hahaha :P
", "votes": null, "replies": [] }, { "id": 419642, "author_name": "ogrellier", "author_url": "", "post_date": "11/12/2018 10:54:44", "content": "@NxGTR I have to agree I'm not a God by any standards. I'm too incompetent to grade the others listed ;-)
", "votes": null, "replies": [] }, { "id": 419665, "author_name": "cpmpml", "author_url": "", "post_date": "11/12/2018 11:21:51", "content": "Amen ;)
\n\nI think there are stronger ML gods than me that you did not cite, like giba, bestfitting, etc. But thanks a lot, very honored to be seen as a god! May the strength be with you!
", "votes": null, "replies": [] }, { "id": 419701, "author_name": "georsara1", "author_url": "", "post_date": "11/12/2018 12:25:11", "content": "Ahhh yeah Giba! He found the leak in Sandanter Challenge (among other great things). Yes Kudos also. But I was not meaning to make an exhaustive list here so apologies to all altrouistic contributors that I didn't mention...
", "votes": null, "replies": [] }, { "id": 419713, "author_name": "satian", "author_url": "", "post_date": "11/12/2018 12:51:01", "content": "I found a new motivation now. Why be a grandmaster when you can be a god?
", "votes": null, "replies": [] } ] }, { "id": 420644, "author_name": "rp1611", "author_url": "", "post_date": "11/13/2018 23:50:46", "content": "\"Doing is better than reading\"
\n\nWhat you learn by doing it yourself can't be replaced by anything. No matter how many books/articles/blogs you read about Data Science but the way you understand and retain concepts while doing it on your own is simply amazing. I have been into Analytics Consulting, however, what I learned within 6 months of Kaggling is simply beyond words.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "417873": "Interesting question, just one of the few you can find about Kaggle in [here][1].\n\nI tried to look for it in Quora, but no luck.\n\nPS: Questions with Kaggle in them are pretty safe.\n\n\n [1]: https://www.kaggle.com/carloshuertas/quora-and-kaggle", "417953": "What I have learned:\n\n - Practical experience of applying machine learning and deep learning algorithms. \n\n - Pipelines of data cleaning, model training/hyper-parameter tuning and ensemble for various types of problems. \n\nWhat I want to learn in the future:\n\n - Get insights and conclusions from data based on EDA.", "418061": "What I have learnt:\n Everything practical I have learnt and applied so far, which includes EDA, Visualization, feature engineering intuitions, feature selection, creating pipelines, hyper parameter tuning, validation approaches.", "418078": "**Learning By Doing** is the best thing about Kaggle. Solving problems while learning is the best way to solidify the concepts that we are trying to learn. For example, It would always be easier for a guy to fix a car who knows how to drive rather than a guy who just knows how a car works but have never actually driven one. Maybe a bad analogy but I hope you get what I am saying.", "418406": "Two things among all, in every competition I took part up so far (they are about 132):\n 1. best practices of applying machine learning to real world problems from all competitors\n 2. bleeding edge techniques from top ones", "418471": "I learn ML through kaggle.\n\nStarting with the basic concepts like: CV, random forest and feature engineering.\nthen hyper param optimisation and boosting.\nNow each competition I try to learn at least one new subject.\nIn this competition I'm going over the functional API of keras and embedding vectors.", "418530": "I love these questions:\n\n'How would researchers like Yann Lecun and the like do in a Kaggle competition?'\nA great answer was given by Geoff Hinton and his team who crushed the first and only Kaggle competition they entered: http://blog.kaggle.com/2012/11/01/deep-learning-how-i-did-it-merck-1st-place-interview/\n\n'Will Kaggle cease to exist when Auto-ML comes into full force?'\n\nThe trick is 'when'. It could take decades before ML automation makes top data scientists useless.", "418544": "Oh.... those days without Kaggle Kernels :)", "418863": "I have learnt : `Fork`, `Commit`, `Submit`, `Don't Upvote`, `Climb the LB`, `Get a medal`\n\nUpdate: I forgot... `Don't get tracked down by GM NxGTR eagle eyes`", "418881": "In that case, we find Hinton's account: https://www.kaggle.com/jeff20", "418901": "I'm always amazed when someone comes out of nowhere and finishes in the money. Here are some more good questions from the Quora site:\n\n - Is Kaggle dead?\n - Does winning a Kaggle competition matter outside of Kaggle?\n - What does it feel like to be addicted to Kaggle?\n - Why do most kaggle users use various ML and DL frameworks and libraries instead of implementing algorithms from scratch? Are they all incompetent?\n\nLinks to these questions and their answers at this topic: https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70867", "418934": "I have learned that 95% of methods and techniques described in academic papers are not applicable to real life (/kaggle) data sets.\n\nAnd for the 5% that do work, well, you need to find out if they actually work the hard way - By trying it out. There is no free lunch.", "418935": "It is a great environment to practice ML algorithms on real problems it improved my ML practical skills as you learn by doing", "418936": "When i started competing on kaggle i wasn't able to create a simple neural network from scratch, I started with simple problems and by the help of kernels and practicing i learned too much so i can now build my own models, So i worked on other problems so i improved my skills in frameworks like tensorflow and keras also i engaged with problems categories i haven't worked with before like segmentation and NLP problems.", "419300": "Kaggle has helped me tremendously by helping me practice, practice, practice. Of course a lot of learning from others who publish top notch kernels/topics.", "419313": "Writing reproducible code and logging everything is extremely important.", "419336": "I shared few things I learned here: https://www.kaggle.com/c/PLAsTiCC-2018/discussion/70908\n\nSome are specific to the Plasticc competition, but most are pretty generic.", "419549": "I read the book 'American Gods' this summer. It talks about the old gods like Odin, Ra, Zeus, etc and the modern gods which are the internet, the television, money, etc. Well, in Kaggle, I met the ML gods: CPMP, NxGTR, SRK, Kazanova, Olivier, Bojan, etc...Respect the ML gods - they might just have turned you into what you are today, or will be tomorrow.", "419565": "You have low standards for Gods hahaha :P", "419570": "It is a paradox but it is actually true: The normal thing to do is to first dig into data, perform EDA to get some insight and then apply a model of some kind on those data. But being eager to see a model 'alive', most of us, as newbies, first copied and deployed a model to see a rank on the leaderboard and then started gaining actual knowledge on the data we are working with!", "419642": "NxGTR I have to agree I'm not a God by any standards. I'm too incompetent to grade the others listed ;-)", "419665": "Amen ;)\n\nI think there are stronger ML gods than me that you did not cite, like giba, bestfitting, etc. But thanks a lot, very honored to be seen as a god! May the strength be with you!", "419701": "Ahhh yeah Giba! He found the leak in Sandanter Challenge (among other great things). Yes Kudos also. But I was not meaning to make an exhaustive list here so apologies to all altrouistic contributors that I didn't mention...", "419713": "I found a new motivation now. Why be a grandmaster when you can be a god?", "419885": "This was interesting to read.", "420028": "Absolutely true in my case :-)", "420123": "I do some EDA myself, but I want to learn to do it in a more systematic way.", "420644": "\"Doing is better than reading\"\n\nWhat you learn by doing it yourself can't be replaced by anything. No matter how many books/articles/blogs you read about Data Science but the way you understand and retain concepts while doing it on your own is simply amazing. I have been into Analytics Consulting, however, what I learned within 6 months of Kaggling is simply beyond words." }, "source": "meta" }