Hi, all the fellow Kagglers! I want to help the frustrated beginners who may find their first competition daunting, by illustrating what I have undergone. There are lots of beginner guides on Kaggle, and I'd like to contribute with some of my fresh insights.
\nBackground: I know almost nothing about ML before the contest, and it's almost my first CS project. Yet, my team is placed 12 in this competition. So, what have I learnt?
\nEssential Tools: If you're serious about Kaggle, invest in a GPU - either Cloud GPU or better yet, a personal PC equipped with one. Kaggle's generous free resources are wonderful, but for prolonged training sessions, they won't suffice.
\nTime Commitment: While some ace Kagglers can grab medals within days, for most, it's a marathon of trial and error.
\nWhat's the right mindset to approach a Kaggle Competition?
\nI think this is rarely shed light on in the forum. The key is to get over the thresholds, and start engaging in the challenge. You will find your way easily with the help from fellow Kagglers! Yet, what is most dangerous is to never start, presumably for several reasons:
\nOvercome peer pressure: While it's true that many are experienced Data Scientists with vast knowledge, remember that Data Science spans a broad spectrum. Therefore, most competitors actually start from scratch. Besides, learning from pros is the crux of all the Kaggle Challenges!
Learn by Doing: Waiting to “know enough” to begin can be a trap. Dive in with a project-based approach. It's more effective than trying to connect the dots from scattered knowledge.
You will be good to go into any contest with these idea in mind. The next step varies and is dependent on the particulars. Yet, few things I feel like should be bear in mind:
\nTeamwork: I've learnt that hard lesson this time, and I feel sorry for not negotiating well with my teammates and instead rashly decided to committing to the contest alone. Please Avoid that! I certainly will learn to be a better collaborator next time, and I hope you can be one as well!
\"Sportsmanship\": My team was shaken down by one place and lost the gold medal this time. I was so annoyed, and hence performed poorly in the Physics Exam later🥲. If it happens, accept it with grace, as shake down/up is an integral part of Kaggle Competition.
Feel free to comment down below and share your thoughts🫡.
\nPS: the whole post may sound cliche, but if it's taken seriously, I guarentee it will help.
", "messages": [ { "id": "2488375", "postDate": "10/19/2023 08:30:22", "content": "Hi, all the fellow Kagglers! I want to help the frustrated beginners who may find their first competition daunting, by illustrating what I have undergone. There are lots of beginner guides on Kaggle, and I'd like to contribute with some of my fresh insights.
\nBackground: I know almost nothing about ML before the contest, and it's almost my first CS project. Yet, my team is placed 12 in this competition. So, what have I learnt?
\nEssential Tools: If you're serious about Kaggle, invest in a GPU - either Cloud GPU or better yet, a personal PC equipped with one. Kaggle's generous free resources are wonderful, but for prolonged training sessions, they won't suffice.
\nTime Commitment: While some ace Kagglers can grab medals within days, for most, it's a marathon of trial and error.
\nWhat's the right mindset to approach a Kaggle Competition?
\nI think this is rarely shed light on in the forum. The key is to get over the thresholds, and start engaging in the challenge. You will find your way easily with the help from fellow Kagglers! Yet, what is most dangerous is to never start, presumably for several reasons:
\nOvercome peer pressure: While it's true that many are experienced Data Scientists with vast knowledge, remember that Data Science spans a broad spectrum. Therefore, most competitors actually start from scratch. Besides, learning from pros is the crux of all the Kaggle Challenges!
Learn by Doing: Waiting to “know enough” to begin can be a trap. Dive in with a project-based approach. It's more effective than trying to connect the dots from scattered knowledge.
You will be good to go into any contest with these idea in mind. The next step varies and is dependent on the particulars. Yet, few things I feel like should be bear in mind:
\nTeamwork: I've learnt that hard lesson this time, and I feel sorry for not negotiating well with my teammates and instead rashly decided to committing to the contest alone. Please Avoid that! I certainly will learn to be a better collaborator next time, and I hope you can be one as well!
\"Sportsmanship\": My team was shaken down by one place and lost the gold medal this time. I was so annoyed, and hence performed poorly in the Physics Exam later🥲. If it happens, accept it with grace, as shake down/up is an integral part of Kaggle Competition.
Feel free to comment down below and share your thoughts🫡.
\nPS: the whole post may sound cliche, but if it's taken seriously, I guarentee it will help.
", "rawMarkdown": "Hi, all the fellow Kagglers! I want to help the frustrated beginners who may find their first competition daunting, by illustrating what I have undergone. There are lots of beginner guides on Kaggle, and I'd like to contribute with some of my fresh insights.\n\n**Background:** I know almost nothing about **ML** before the contest, and it's almost my **first CS project**. Yet, my team is placed 12 in this competition. So, what have I learnt?\n\n---\n\n**Essential Tools:** If you're serious about Kaggle, invest in a **GPU** - either Cloud GPU or better yet, a personal PC equipped with one. Kaggle's generous free resources are wonderful, but for prolonged training sessions, they won't suffice.\n\n**Time Commitment:** While some ace Kagglers can grab medals within days, for most, it's a marathon of trial and error. \n\n--- \n\nWhat's the right mindset to approach a Kaggle Competition?\n\nI think this is rarely shed light on in the forum. The key is to get over the **thresholds**, and start engaging in the challenge. You will find your way easily with the help from fellow Kagglers! Yet, what is most dangerous is to **never start**, presumably for several reasons:\n\n - **Overcome peer pressure:** While it's true that many are experienced Data Scientists with vast knowledge, remember that Data Science spans a broad spectrum. Therefore, **most competitors** actually start from **scratch**. Besides, learning from pros is the crux of all the Kaggle Challenges!\n\n - **Learn by Doing:** Waiting to “know enough” to begin can be a trap. Dive in with a **project-based** approach. It's more effective than trying to connect the dots from scattered knowledge.\n\n---\n\nYou will be good to go into any contest with these idea in mind. The next step varies and is dependent on the particulars. Yet, few things I feel like should be bear in mind:\n\n- **Teamwork:** I've learnt that **hard lesson** this time, and I feel sorry for not negotiating well with my teammates and instead rashly decided to committing to the contest **alone**. Please **Avoid that!** I certainly will learn to be a better collaborator next time, and I hope you can be one as well!\n\n- **\"Sportsmanship\":** My team was shaken down by one place and lost the gold medal this time. I was so annoyed, and hence performed poorly in the Physics Exam later🥲. If it happens, accept it with **grace**, as **shake down/up** is an integral part of Kaggle Competition. \n\n---\n\nFeel free to comment down below and share your thoughts🫡.\n\nPS: the whole post may sound cliche, but if it's taken seriously, I guarentee it will help.", "votes": null }, { "id": "2488377", "postDate": "10/19/2023 08:30:59", "content": "Is here the right place to post such a kind of thing?🫠
", "rawMarkdown": "Is here the right place to post such a kind of thing?🫠", "votes": null }, { "id": "2489181", "postDate": "10/19/2023 18:56:20", "content": "Absolutely. I agree with your points. By the way, you are going up 😉
", "rawMarkdown": "Absolutely. I agree with your points. By the way, you are going up 😉", "votes": null }, { "id": "2489359", "postDate": "10/19/2023 23:48:23", "content": "Is it because that one team has used a second account to submit for more times? This all looks so surreal 🥹
", "rawMarkdown": "Is it because that one team has used a second account to submit for more times? This all looks so surreal 🥹", "votes": null }, { "id": "2489389", "postDate": "10/20/2023 01:01:02", "content": "Great advice for beginners! I recommend adding a Beginner tag to this post so that it will gain more visibility.
\n\n\nLearn by Doing: Waiting to “know enough” to begin can be a trap. Dive in with a project-based approach. It's more effective than trying to connect the dots from scattered knowledge.
\n
I would like to understand what you meant by \"project-based approach\"?
", "rawMarkdown": "Great advice for beginners! I recommend adding a Beginner tag to this post so that it will gain more visibility.\n\n>Learn by Doing: Waiting to “know enough” to begin can be a trap. Dive in with a project-based approach. It's more effective than trying to connect the dots from scattered knowledge.\n\nI would like to understand what you meant by \"project-based approach\"?", "votes": null }, { "id": "2489430", "postDate": "10/20/2023 02:31:50", "content": "It means that it's better to have a concrete goal in mind, like increasing the accuracy of your ASR model. To improve on the project, you will learn a lot in the process. And each Kaggle competition in my opinion is a challenging project, so I believe it's a good place to kick off.
\nConversely, another approach is to learn for the sake of learning. I have taken some online ML courses, that will just skim through NLP, image recognition, and various loosely related content. For me, I will lose the motive easily, because I am not doing it for a specific purpose.
\nHave I clarified my point? @kieranyogaraj
", "rawMarkdown": "It means that it's better to have a concrete goal in mind, like increasing the accuracy of your ASR model. To improve on the **project**, you will learn a lot in the process. And each Kaggle competition in my opinion is a challenging project, so I believe it's a good place to kick off.\n\nConversely, another approach is to **learn for the sake of learning**. I have taken some online ML courses, that will just skim through NLP, image recognition, and various loosely related content. For me, I will lose the motive easily, because I am not doing it for a specific purpose. \n\nHave I clarified my point? @kieranyogaraj", "votes": null }, { "id": "2489499", "postDate": "10/20/2023 03:25:55", "content": "Thank you, your answer is very helpful
", "rawMarkdown": "Thank you, your answer is very helpful", "votes": null }, { "id": "2490122", "postDate": "10/20/2023 12:54:53", "content": "Something like this, somebody cheated and got banned. Congratulations!
", "rawMarkdown": "Something like this, somebody cheated and got banned. Congratulations!", "votes": null }, { "id": "2517289", "postDate": "11/08/2023 11:07:08", "content": "说得好！！！我是beginner，可以将您的3090捐献给我吗？
", "rawMarkdown": "说得好！！！我是beginner，可以将您的3090捐献给我吗？", "votes": null }, { "id": "2518177", "postDate": "11/09/2023 05:41:19", "content": "谢谢您的丰富遗产！！！我现在已经成为马克扎克伯格了！！！
", "rawMarkdown": "谢谢您的丰富遗产！！！我现在已经成为马克扎克伯格了！！！", "votes": null } ], "comments": [ { "id": 2488377, "author_name": "renyiwei", "author_url": "", "post_date": "10/19/2023 08:30:59", "content": "Is here the right place to post such a kind of thing?🫠
", "votes": null, "replies": [ { "id": 2489181, "author_name": "manwithaflower", "author_url": "", "post_date": "10/19/2023 18:56:20", "content": "Absolutely. I agree with your points. By the way, you are going up 😉
", "votes": null, "replies": [ { "id": 2489359, "author_name": "renyiwei", "author_url": "", "post_date": "10/19/2023 23:48:23", "content": "Is it because that one team has used a second account to submit for more times? This all looks so surreal 🥹
", "votes": null, "replies": [ { "id": 2490122, "author_name": "manwithaflower", "author_url": "", "post_date": "10/20/2023 12:54:53", "content": "Something like this, somebody cheated and got banned. Congratulations!
", "votes": null, "replies": [] } ] } ] } ] }, { "id": 2489389, "author_name": "kieranyogaraj", "author_url": "", "post_date": "10/20/2023 01:01:02", "content": "Great advice for beginners! I recommend adding a Beginner tag to this post so that it will gain more visibility.
\n\n\nLearn by Doing: Waiting to “know enough” to begin can be a trap. Dive in with a project-based approach. It's more effective than trying to connect the dots from scattered knowledge.
\n
I would like to understand what you meant by \"project-based approach\"?
", "votes": null, "replies": [ { "id": 2489430, "author_name": "renyiwei", "author_url": "", "post_date": "10/20/2023 02:31:50", "content": "It means that it's better to have a concrete goal in mind, like increasing the accuracy of your ASR model. To improve on the project, you will learn a lot in the process. And each Kaggle competition in my opinion is a challenging project, so I believe it's a good place to kick off.
\nConversely, another approach is to learn for the sake of learning. I have taken some online ML courses, that will just skim through NLP, image recognition, and various loosely related content. For me, I will lose the motive easily, because I am not doing it for a specific purpose.
\nHave I clarified my point? @kieranyogaraj
", "votes": null, "replies": [ { "id": 2489499, "author_name": "kieranyogaraj", "author_url": "", "post_date": "10/20/2023 03:25:55", "content": "Thank you, your answer is very helpful
", "votes": null, "replies": [] } ] } ] }, { "id": 2517289, "author_name": "tongyanjyut", "author_url": "", "post_date": "11/08/2023 11:07:08", "content": "说得好！！！我是beginner，可以将您的3090捐献给我吗？
", "votes": null, "replies": [] }, { "id": 2518177, "author_name": "tongyanjyut", "author_url": "", "post_date": "11/09/2023 05:41:19", "content": "谢谢您的丰富遗产！！！我现在已经成为马克扎克伯格了！！！
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2488375": "Hi, all the fellow Kagglers! I want to help the frustrated beginners who may find their first competition daunting, by illustrating what I have undergone. There are lots of beginner guides on Kaggle, and I'd like to contribute with some of my fresh insights.\n\n**Background:** I know almost nothing about **ML** before the contest, and it's almost my **first CS project**. Yet, my team is placed 12 in this competition. So, what have I learnt?\n\n---\n\n**Essential Tools:** If you're serious about Kaggle, invest in a **GPU** - either Cloud GPU or better yet, a personal PC equipped with one. Kaggle's generous free resources are wonderful, but for prolonged training sessions, they won't suffice.\n\n**Time Commitment:** While some ace Kagglers can grab medals within days, for most, it's a marathon of trial and error. \n\n--- \n\nWhat's the right mindset to approach a Kaggle Competition?\n\nI think this is rarely shed light on in the forum. The key is to get over the **thresholds**, and start engaging in the challenge. You will find your way easily with the help from fellow Kagglers! Yet, what is most dangerous is to **never start**, presumably for several reasons:\n\n - **Overcome peer pressure:** While it's true that many are experienced Data Scientists with vast knowledge, remember that Data Science spans a broad spectrum. Therefore, **most competitors** actually start from **scratch**. Besides, learning from pros is the crux of all the Kaggle Challenges!\n\n - **Learn by Doing:** Waiting to “know enough” to begin can be a trap. Dive in with a **project-based** approach. It's more effective than trying to connect the dots from scattered knowledge.\n\n---\n\nYou will be good to go into any contest with these idea in mind. The next step varies and is dependent on the particulars. Yet, few things I feel like should be bear in mind:\n\n- **Teamwork:** I've learnt that **hard lesson** this time, and I feel sorry for not negotiating well with my teammates and instead rashly decided to committing to the contest **alone**. Please **Avoid that!** I certainly will learn to be a better collaborator next time, and I hope you can be one as well!\n\n- **\"Sportsmanship\":** My team was shaken down by one place and lost the gold medal this time. I was so annoyed, and hence performed poorly in the Physics Exam later🥲. If it happens, accept it with **grace**, as **shake down/up** is an integral part of Kaggle Competition. \n\n---\n\nFeel free to comment down below and share your thoughts🫡.\n\nPS: the whole post may sound cliche, but if it's taken seriously, I guarentee it will help.", "2488377": "Is here the right place to post such a kind of thing?🫠", "2489181": "Absolutely. I agree with your points. By the way, you are going up 😉", "2489359": "Is it because that one team has used a second account to submit for more times? This all looks so surreal 🥹", "2489389": "Great advice for beginners! I recommend adding a Beginner tag to this post so that it will gain more visibility.\n\n>Learn by Doing: Waiting to “know enough” to begin can be a trap. Dive in with a project-based approach. It's more effective than trying to connect the dots from scattered knowledge.\n\nI would like to understand what you meant by \"project-based approach\"?", "2489430": "It means that it's better to have a concrete goal in mind, like increasing the accuracy of your ASR model. To improve on the **project**, you will learn a lot in the process. And each Kaggle competition in my opinion is a challenging project, so I believe it's a good place to kick off.\n\nConversely, another approach is to **learn for the sake of learning**. I have taken some online ML courses, that will just skim through NLP, image recognition, and various loosely related content. For me, I will lose the motive easily, because I am not doing it for a specific purpose. \n\nHave I clarified my point? @kieranyogaraj", "2489499": "Thank you, your answer is very helpful", "2490122": "Something like this, somebody cheated and got banned. Congratulations!", "2517289": "说得好！！！我是beginner，可以将您的3090捐献给我吗？", "2518177": "谢谢您的丰富遗产！！！我现在已经成为马克扎克伯格了！！！" }, "source": "meta" }