{
  "id": 307689,
  "title": "🤝🤝🤝 TEAM WORK 🤝🤝🤝 during Kaggle competition - summary ",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/307689",
  "author_name": "",
  "post_date": "2022-02-15T08:37:35.459888800Z",
  "votes": 31,
  "comment_count": 8,
  "views": 0,
  "content": "<p>In this post I would like to describe a few topics related to the organization of our work as a team. I think it might be interesting for people who would like to start as a team in the future.</p>\n<ul>\n<li><p>You have to think about forming a team early because Kaggle counts the number of submissions you have done so far when joining the team. If several people have made a large number of submissions, you have to wait to be able to team up.</p></li>\n<li><p>We created a team of people who had known each other before (from other competitions). We knew our strengths and weaknesses and used them by dividing the work accordingly.</p></li>\n<li><p>We used Slack for communication. We have created two channels - one general for continuous discussion and the other as a library of documents, links to code.</p></li>\n<li><p>Daily routines:<br>\n** We looked through the discussions forum - new topics and new posts.<br>\n** We were browsing notebooks - best score and new ones to check new ideas.<br>\n** We talked to each other every day - every day. We exchanged our observations about what was new in the competition. We were building hypotheses on how this might affect the competition result and test them locally (sometimes check using public LB).<br>\n** We divided our submissions in the team for example Remek 2, Lukasz 3. The next day we changed the order. If someone had better ideas, he would convince the other person to give up the submission.<br>\n** We made a lot of experiments - not by theorizing but by performing experiments and proving that something works or not.</p></li>\n<li><p>We created our environment on Goolge Colab Pro +. The environment has many advantages, but in such a competition, a definite advantage would be an unlimited access to a stronger environment, e.g. a strong 2xGPU machine (minimum) locally.</p></li>\n<li><p>We shared many parts of the code eg. <a href=\"https://www.kaggle.com/lukaszborecki\" target=\"_blank\">@lukaszborecki</a> wrote a function to validate f2  and we used this function in our scripts.</p></li>\n<li><p>We agreed in the team what could be shared and what we will wait for later. However, it turned out that we shared almost everything. We didn't see much sense in keeping something a secret as it was not a competitive advantage (anyone could do it).</p></li>\n<li><p>We had a lot of hard talks and stormy exchanges of views, but we always managed to come to constructive conclusions. Usually someone said - ok prove it.</p></li>\n<li><p>Now we know that this is marathon not sprint. You have to distribute energy very well during the competition. We had a lot of energy in the beginning but in the ending … we had less and less and it was not because we fell from GOLD to SILVER but we were just tired of implementing different ideas.</p></li>\n\n<li><p>[TO IMPROVE] We took a lot of notes - to plan next steps. I personally think that we shoud improve this part a lot in future competition. It seems to me now that despite writing downthe parameters of the experiment, we have repeated them.</p></li>\n<li><p>[TO IMPROVE] We have done a lot of experiments (submitting them). It is worth writing down in the submittal comments what the experiment was, in order to be able to compare the results later.</p></li>\n<li><p>[TO IMPROVE] Plan / Do / Check / Act as a scientific approach. Sometimes we tried to check many parameters at once. It was difficult to say later which parameter influenced the result. The experiment had to be repeated on a smaller number of parameters.</p></li>\n</ul>\n<p>I would like to thank my teammates <a href=\"https://www.kaggle.com/lukaszborecki\" target=\"_blank\">@lukaszborecki</a> and <a href=\"https://www.kaggle.com/marcinstasko\" target=\"_blank\">@marcinstasko</a> for absolutely fantastic work during this competition! </p>\n<p>Notes - December <br>\n<img src=\"https://i.ibb.co/wddhyYg/IMG-0597.jpg\" alt=\"Notes\"></p>",
  "messages": [
    {
      "id": "1691119",
      "postDate": "02/15/2022 08:37:35",
      "content": "<p>In this post I would like to describe a few topics related to the organization of our work as a team. I think it might be interesting for people who would like to start as a team in the future.</p>\n<ul>\n<li><p>You have to think about forming a team early because Kaggle counts the number of submissions you have done so far when joining the team. If several people have made a large number of submissions, you have to wait to be able to team up.</p></li>\n<li><p>We created a team of people who had known each other before (from other competitions). We knew our strengths and weaknesses and used them by dividing the work accordingly.</p></li>\n<li><p>We used Slack for communication. We have created two channels - one general for continuous discussion and the other as a library of documents, links to code.</p></li>\n<li><p>Daily routines:<br>\n** We looked through the discussions forum - new topics and new posts.<br>\n** We were browsing notebooks - best score and new ones to check new ideas.<br>\n** We talked to each other every day - every day. We exchanged our observations about what was new in the competition. We were building hypotheses on how this might affect the competition result and test them locally (sometimes check using public LB).<br>\n** We divided our submissions in the team for example Remek 2, Lukasz 3. The next day we changed the order. If someone had better ideas, he would convince the other person to give up the submission.<br>\n** We made a lot of experiments - not by theorizing but by performing experiments and proving that something works or not.</p></li>\n<li><p>We created our environment on Goolge Colab Pro +. The environment has many advantages, but in such a competition, a definite advantage would be an unlimited access to a stronger environment, e.g. a strong 2xGPU machine (minimum) locally.</p></li>\n<li><p>We shared many parts of the code eg. <a href=\"https://www.kaggle.com/lukaszborecki\" target=\"_blank\">@lukaszborecki</a> wrote a function to validate f2  and we used this function in our scripts.</p></li>\n<li><p>We agreed in the team what could be shared and what we will wait for later. However, it turned out that we shared almost everything. We didn't see much sense in keeping something a secret as it was not a competitive advantage (anyone could do it).</p></li>\n<li><p>We had a lot of hard talks and stormy exchanges of views, but we always managed to come to constructive conclusions. Usually someone said - ok prove it.</p></li>\n<li><p>Now we know that this is marathon not sprint. You have to distribute energy very well during the competition. We had a lot of energy in the beginning but in the ending … we had less and less and it was not because we fell from GOLD to SILVER but we were just tired of implementing different ideas.</p></li>\n\n<li><p>[TO IMPROVE] We took a lot of notes - to plan next steps. I personally think that we shoud improve this part a lot in future competition. It seems to me now that despite writing downthe parameters of the experiment, we have repeated them.</p></li>\n<li><p>[TO IMPROVE] We have done a lot of experiments (submitting them). It is worth writing down in the submittal comments what the experiment was, in order to be able to compare the results later.</p></li>\n<li><p>[TO IMPROVE] Plan / Do / Check / Act as a scientific approach. Sometimes we tried to check many parameters at once. It was difficult to say later which parameter influenced the result. The experiment had to be repeated on a smaller number of parameters.</p></li>\n</ul>\n<p>I would like to thank my teammates <a href=\"https://www.kaggle.com/lukaszborecki\" target=\"_blank\">@lukaszborecki</a> and <a href=\"https://www.kaggle.com/marcinstasko\" target=\"_blank\">@marcinstasko</a> for absolutely fantastic work during this competition! </p>\n<p>Notes - December <br>\n<img src=\"https://i.ibb.co/wddhyYg/IMG-0597.jpg\" alt=\"Notes\"></p>",
      "rawMarkdown": "In this post I would like to describe a few topics related to the organization of our work as a team. I think it might be interesting for people who would like to start as a team in the future.\n\n* You have to think about forming a team early because Kaggle counts the number of submissions you have done so far when joining the team. If several people have made a large number of submissions, you have to wait to be able to team up.\n\n* We created a team of people who had known each other before (from other competitions). We knew our strengths and weaknesses and used them by dividing the work accordingly.\n\n* We used Slack for communication. We have created two channels - one general for continuous discussion and the other as a library of documents, links to code.\n\n* Daily routines:\n** We looked through the discussions forum - new topics and new posts.\n** We were browsing notebooks - best score and new ones to check new ideas.\n** We talked to each other every day - every day. We exchanged our observations about what was new in the competition. We were building hypotheses on how this might affect the competition result and test them locally (sometimes check using public LB).\n** We divided our submissions in the team for example Remek 2, Lukasz 3. The next day we changed the order. If someone had better ideas, he would convince the other person to give up the submission.\n** We made a lot of experiments - not by theorizing but by performing experiments and proving that something works or not.\n\n* We created our environment on Goolge Colab Pro +. The environment has many advantages, but in such a competition, a definite advantage would be an unlimited access to a stronger environment, e.g. a strong 2xGPU machine (minimum) locally.\n* We shared many parts of the code eg. @lukaszborecki wrote a function to validate f2  and we used this function in our scripts.\n* We agreed in the team what could be shared and what we will wait for later. However, it turned out that we shared almost everything. We didn't see much sense in keeping something a secret as it was not a competitive advantage (anyone could do it).\n* We had a lot of hard talks and stormy exchanges of views, but we always managed to come to constructive conclusions. Usually someone said - ok prove it.\n* Now we know that this is marathon not sprint. You have to distribute energy very well during the competition. We had a lot of energy in the beginning but in the ending ... we had less and less and it was not because we fell from GOLD to SILVER but we were just tired of implementing different ideas.\n\n\n\n* [TO IMPROVE] We took a lot of notes - to plan next steps. I personally think that we shoud improve this part a lot in future competition. It seems to me now that despite writing downthe parameters of the experiment, we have repeated them.\n* [TO IMPROVE] We have done a lot of experiments (submitting them). It is worth writing down in the submittal comments what the experiment was, in order to be able to compare the results later.\n* [TO IMPROVE] Plan / Do / Check / Act as a scientific approach. Sometimes we tried to check many parameters at once. It was difficult to say later which parameter influenced the result. The experiment had to be repeated on a smaller number of parameters.\n\nI would like to thank my teammates @lukaszborecki and @marcinstasko for absolutely fantastic work during this competition! \n\nNotes - December \n![Notes](https://i.ibb.co/wddhyYg/IMG-0597.jpg)",
      "votes": null
    },
    {
      "id": "1691151",
      "postDate": "02/15/2022 08:59:54",
      "content": "<blockquote>\n  <p>Now we know that this is marathon not sprint. You have to distribute energy very well during the competition.</p>\n</blockquote>\n<p>Golden words! Thanks so much for the writeup and for sharing your secret sauces</p>\n<p>May I ask, what tools were you taking to track your experiments across the team? </p>\n<p>Were the ✍️and📝used throughout the comp?</p>",
      "rawMarkdown": "> Now we know that this is marathon not sprint. You have to distribute energy very well during the competition.\n\nGolden words! Thanks so much for the writeup and for sharing your secret sauces\n\nMay I ask, what tools were you taking to track your experiments across the team? \n\nWere the ✍️and📝used throughout the comp?",
      "votes": null
    },
    {
      "id": "1691172",
      "postDate": "02/15/2022 09:15:00",
      "content": "<p>Thank you very much!</p>\n<p>We used Slack for communication and sometimes we made telco (when we have to explain something hot and hard to describe). Then we use Google Sheet (and excel) to write down some experiments results. But to be honest we have to improve this area of collaboration. </p>\n<p>One of the sample sheet to track score<br>\n<img src=\"https://i.ibb.co/zVr8CWB/exc001.jpg\" alt=\"score\"></p>",
      "rawMarkdown": "Thank you very much!\n\nWe used Slack for communication and sometimes we made telco (when we have to explain something hot and hard to describe). Then we use Google Sheet (and excel) to write down some experiments results. But to be honest we have to improve this area of collaboration. \n\nOne of the sample sheet to track score\n![score](https://i.ibb.co/zVr8CWB/exc001.jpg)",
      "votes": null
    },
    {
      "id": "1692167",
      "postDate": "02/15/2022 21:40:35",
      "content": "<blockquote>\n  <p>Usually someone said - ok prove it.</p>\n</blockquote>\n<p>😄</p>\n<p>\"ok, prove it\" My PhD supervisor</p>",
      "rawMarkdown": "> Usually someone said - ok prove it.\n\n😄\n\n\"ok, prove it\" My PhD supervisor",
      "votes": null
    },
    {
      "id": "1692170",
      "postDate": "02/15/2022 21:45:23",
      "content": "<p>Yes, …. we have a lot of difficult discussions. Everybody has a lot of ideas. Sometimes we had problem what to submit (so many ideas so little submissions). Personally I thnink that this was due to our competition unexperinced mind … As I can see now (this is really honest) many TOP players moves like in chess play - they calculate next move. This is one of my lessons learned from this competition. </p>\n<p>But \"prove it\" is a scientific approach. Very often we just think that something is true … and it turns out to be a false assumption that needs to be proved.  </p>",
      "rawMarkdown": "Yes, .... we have a lot of difficult discussions. Everybody has a lot of ideas. Sometimes we had problem what to submit (so many ideas so little submissions). Personally I thnink that this was due to our competition unexperinced mind ... As I can see now (this is really honest) many TOP players moves like in chess play - they calculate next move. This is one of my lessons learned from this competition. \n\nBut \"prove it\" is a scientific approach. Very often we just think that something is true ... and it turns out to be a false assumption that needs to be proved.",
      "votes": null
    },
    {
      "id": "1692219",
      "postDate": "02/15/2022 22:41:46",
      "content": "<p>in chapter 4 from this book</p>\n<p><img src=\"https://i.postimg.cc/ZYFhR9R9/1617559696040.jpg\" alt=\"\"></p>\n<p><img src=\"https://i.postimg.cc/VLfPbYrK/Untitled.jpg\" alt=\"\"></p>",
      "rawMarkdown": "in chapter 4 from this book\n\n![](https://i.postimg.cc/ZYFhR9R9/1617559696040.jpg)\n\n![](https://i.postimg.cc/VLfPbYrK/Untitled.jpg)",
      "votes": null
    },
    {
      "id": "1693045",
      "postDate": "02/16/2022 12:08:56",
      "content": "<p>Thank you for sharing your experience! As a beginner with still no participations on rewarded competitions, I thought it would be interesting to ask for your opinion in whether to compete as a team or not. I frequently see individual competitors on top of the leaderbords, so would you say it is better to go solo or as a group?</p>",
      "rawMarkdown": "Thank you for sharing your experience! As a beginner with still no participations on rewarded competitions, I thought it would be interesting to ask for your opinion in whether to compete as a team or not. I frequently see individual competitors on top of the leaderbords, so would you say it is better to go solo or as a group?",
      "votes": null
    },
    {
      "id": "1693229",
      "postDate": "02/16/2022 14:35:53",
      "content": "<blockquote>\n  <p>However, it turned out that we shared almost everything.</p>\n</blockquote>\n<p>LOL! I am laughing 😄! Great team from Poland, great neighbors 😃!  </p>",
      "rawMarkdown": "> However, it turned out that we shared almost everything.\n\nLOL! I am laughing 😄! Great team from Poland, great neighbors 😃!",
      "votes": null
    },
    {
      "id": "1693571",
      "postDate": "02/16/2022 18:49:55",
      "content": "<p>Great job guys! 💪</p>",
      "rawMarkdown": "Great job guys! 💪",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1691151,
      "author_name": "init27",
      "author_url": "",
      "post_date": "02/15/2022 08:59:54",
      "content": "<blockquote>\n  <p>Now we know that this is marathon not sprint. You have to distribute energy very well during the competition.</p>\n</blockquote>\n<p>Golden words! Thanks so much for the writeup and for sharing your secret sauces</p>\n<p>May I ask, what tools were you taking to track your experiments across the team? </p>\n<p>Were the ✍️and📝used throughout the comp?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1691172,
          "author_name": "remekkinas",
          "author_url": "",
          "post_date": "02/15/2022 09:15:00",
          "content": "<p>Thank you very much!</p>\n<p>We used Slack for communication and sometimes we made telco (when we have to explain something hot and hard to describe). Then we use Google Sheet (and excel) to write down some experiments results. But to be honest we have to improve this area of collaboration. </p>\n<p>One of the sample sheet to track score<br>\n<img src=\"https://i.ibb.co/zVr8CWB/exc001.jpg\" alt=\"score\"></p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1692167,
      "author_name": "faisalalsrheed",
      "author_url": "",
      "post_date": "02/15/2022 21:40:35",
      "content": "<blockquote>\n  <p>Usually someone said - ok prove it.</p>\n</blockquote>\n<p>😄</p>\n<p>\"ok, prove it\" My PhD supervisor</p>",
      "votes": null,
      "replies": [
        {
          "id": 1692170,
          "author_name": "remekkinas",
          "author_url": "",
          "post_date": "02/15/2022 21:45:23",
          "content": "<p>Yes, …. we have a lot of difficult discussions. Everybody has a lot of ideas. Sometimes we had problem what to submit (so many ideas so little submissions). Personally I thnink that this was due to our competition unexperinced mind … As I can see now (this is really honest) many TOP players moves like in chess play - they calculate next move. This is one of my lessons learned from this competition. </p>\n<p>But \"prove it\" is a scientific approach. Very often we just think that something is true … and it turns out to be a false assumption that needs to be proved.  </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1692219,
          "author_name": "faisalalsrheed",
          "author_url": "",
          "post_date": "02/15/2022 22:41:46",
          "content": "<p>in chapter 4 from this book</p>\n<p><img src=\"https://i.postimg.cc/ZYFhR9R9/1617559696040.jpg\" alt=\"\"></p>\n<p><img src=\"https://i.postimg.cc/VLfPbYrK/Untitled.jpg\" alt=\"\"></p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1693045,
      "author_name": "tomasroldanborjas",
      "author_url": "",
      "post_date": "02/16/2022 12:08:56",
      "content": "<p>Thank you for sharing your experience! As a beginner with still no participations on rewarded competitions, I thought it would be interesting to ask for your opinion in whether to compete as a team or not. I frequently see individual competitors on top of the leaderbords, so would you say it is better to go solo or as a group?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1693229,
      "author_name": "vad13irt",
      "author_url": "",
      "post_date": "02/16/2022 14:35:53",
      "content": "<blockquote>\n  <p>However, it turned out that we shared almost everything.</p>\n</blockquote>\n<p>LOL! I am laughing 😄! Great team from Poland, great neighbors 😃!  </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1693571,
      "author_name": "michau96",
      "author_url": "",
      "post_date": "02/16/2022 18:49:55",
      "content": "<p>Great job guys! 💪</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1691119": "In this post I would like to describe a few topics related to the organization of our work as a team. I think it might be interesting for people who would like to start as a team in the future.\n\n* You have to think about forming a team early because Kaggle counts the number of submissions you have done so far when joining the team. If several people have made a large number of submissions, you have to wait to be able to team up.\n\n* We created a team of people who had known each other before (from other competitions). We knew our strengths and weaknesses and used them by dividing the work accordingly.\n\n* We used Slack for communication. We have created two channels - one general for continuous discussion and the other as a library of documents, links to code.\n\n* Daily routines:\n** We looked through the discussions forum - new topics and new posts.\n** We were browsing notebooks - best score and new ones to check new ideas.\n** We talked to each other every day - every day. We exchanged our observations about what was new in the competition. We were building hypotheses on how this might affect the competition result and test them locally (sometimes check using public LB).\n** We divided our submissions in the team for example Remek 2, Lukasz 3. The next day we changed the order. If someone had better ideas, he would convince the other person to give up the submission.\n** We made a lot of experiments - not by theorizing but by performing experiments and proving that something works or not.\n\n* We created our environment on Goolge Colab Pro +. The environment has many advantages, but in such a competition, a definite advantage would be an unlimited access to a stronger environment, e.g. a strong 2xGPU machine (minimum) locally.\n* We shared many parts of the code eg. @lukaszborecki wrote a function to validate f2  and we used this function in our scripts.\n* We agreed in the team what could be shared and what we will wait for later. However, it turned out that we shared almost everything. We didn't see much sense in keeping something a secret as it was not a competitive advantage (anyone could do it).\n* We had a lot of hard talks and stormy exchanges of views, but we always managed to come to constructive conclusions. Usually someone said - ok prove it.\n* Now we know that this is marathon not sprint. You have to distribute energy very well during the competition. We had a lot of energy in the beginning but in the ending ... we had less and less and it was not because we fell from GOLD to SILVER but we were just tired of implementing different ideas.\n\n\n\n* [TO IMPROVE] We took a lot of notes - to plan next steps. I personally think that we shoud improve this part a lot in future competition. It seems to me now that despite writing downthe parameters of the experiment, we have repeated them.\n* [TO IMPROVE] We have done a lot of experiments (submitting them). It is worth writing down in the submittal comments what the experiment was, in order to be able to compare the results later.\n* [TO IMPROVE] Plan / Do / Check / Act as a scientific approach. Sometimes we tried to check many parameters at once. It was difficult to say later which parameter influenced the result. The experiment had to be repeated on a smaller number of parameters.\n\nI would like to thank my teammates @lukaszborecki and @marcinstasko for absolutely fantastic work during this competition! \n\nNotes - December \n![Notes](https://i.ibb.co/wddhyYg/IMG-0597.jpg)",
    "1691151": "> Now we know that this is marathon not sprint. You have to distribute energy very well during the competition.\n\nGolden words! Thanks so much for the writeup and for sharing your secret sauces\n\nMay I ask, what tools were you taking to track your experiments across the team? \n\nWere the ✍️and📝used throughout the comp?",
    "1691172": "Thank you very much!\n\nWe used Slack for communication and sometimes we made telco (when we have to explain something hot and hard to describe). Then we use Google Sheet (and excel) to write down some experiments results. But to be honest we have to improve this area of collaboration. \n\nOne of the sample sheet to track score\n![score](https://i.ibb.co/zVr8CWB/exc001.jpg)",
    "1692167": "> Usually someone said - ok prove it.\n\n😄\n\n\"ok, prove it\" My PhD supervisor",
    "1692170": "Yes, .... we have a lot of difficult discussions. Everybody has a lot of ideas. Sometimes we had problem what to submit (so many ideas so little submissions). Personally I thnink that this was due to our competition unexperinced mind ... As I can see now (this is really honest) many TOP players moves like in chess play - they calculate next move. This is one of my lessons learned from this competition. \n\nBut \"prove it\" is a scientific approach. Very often we just think that something is true ... and it turns out to be a false assumption that needs to be proved.",
    "1692219": "in chapter 4 from this book\n\n![](https://i.postimg.cc/ZYFhR9R9/1617559696040.jpg)\n\n![](https://i.postimg.cc/VLfPbYrK/Untitled.jpg)",
    "1693045": "Thank you for sharing your experience! As a beginner with still no participations on rewarded competitions, I thought it would be interesting to ask for your opinion in whether to compete as a team or not. I frequently see individual competitors on top of the leaderbords, so would you say it is better to go solo or as a group?",
    "1693229": "> However, it turned out that we shared almost everything.\n\nLOL! I am laughing 😄! Great team from Poland, great neighbors 😃!",
    "1693571": "Great job guys! 💪"
  },
  "source": "meta"
}