I want to explicitly thank the organizers for putting up this competition. It is clear that everyone involved took a substantial risk by investing a lot of time in preparing and running the competition, compared to which the price money is probably a small item. There were difficult trade-offs to make between the approachability of the problem to a large enough audience and realism. From my limited perspective, I'd say the organizers made very good choices in this respect.
\n\nThe scientific outcome was (and is) uncertain. It is entirely possible that at the end the scientific bottom line might be zero (if one does not count negative results). If so, it will be a bitter pill to swallow, but I wouldn't see it as a failure of either the organizers or the participants. Innovation and exceptional ideas simply cannot be guaranteed to happen in a prescribed time and setting. Only speaking for my solution, I know that I did not manage to come up with some truly innovative, let alone game-changing ideas. It would have been great, but the prior probability for something like that is necessarily very low.
\n\nIt is also entirely possible that by analyzing results of this phase and the next one, some great ideas can be extracted. The big advantage of such a semi-realistic competition setting is that phony ideas which only work on toy problems are immediately filtered out. (In my field of expertise the academic literature is full of such ideas. I assume it is less so in HEP and ML, but still not zero.)
\n\nOne should also not overlook a potential \"long tail\" effect of this competition: The trackml dataset is now readily available and well documented, and it can serve as a benchmark for people in the HEP or ML communities who may come up with great ideas long after the competitions have concluded. In this context, the main disadvantage of the competition setting (relatively inexperienced people botching together things without the time for proper research) disappears.
\n\nFor me personally the competition was an eye-opener. I have been fascinated with physics for all of my life, and while I always had respect and admiration for the experimentalists, my interests focused mostly on the theoretical side. This competition taught me how incredibly challenging even a small single step in the data processing pipeline of a modern experiment can be. From now on, I will look at the nice plots in LHC papers with different eyes.
\n\nThank you!
", "messages": [ { "id": "370766", "postDate": "08/15/2018 12:32:11", "content": "I want to explicitly thank the organizers for putting up this competition. It is clear that everyone involved took a substantial risk by investing a lot of time in preparing and running the competition, compared to which the price money is probably a small item. There were difficult trade-offs to make between the approachability of the problem to a large enough audience and realism. From my limited perspective, I'd say the organizers made very good choices in this respect.
\n\nThe scientific outcome was (and is) uncertain. It is entirely possible that at the end the scientific bottom line might be zero (if one does not count negative results). If so, it will be a bitter pill to swallow, but I wouldn't see it as a failure of either the organizers or the participants. Innovation and exceptional ideas simply cannot be guaranteed to happen in a prescribed time and setting. Only speaking for my solution, I know that I did not manage to come up with some truly innovative, let alone game-changing ideas. It would have been great, but the prior probability for something like that is necessarily very low.
\n\nIt is also entirely possible that by analyzing results of this phase and the next one, some great ideas can be extracted. The big advantage of such a semi-realistic competition setting is that phony ideas which only work on toy problems are immediately filtered out. (In my field of expertise the academic literature is full of such ideas. I assume it is less so in HEP and ML, but still not zero.)
\n\nOne should also not overlook a potential \"long tail\" effect of this competition: The trackml dataset is now readily available and well documented, and it can serve as a benchmark for people in the HEP or ML communities who may come up with great ideas long after the competitions have concluded. In this context, the main disadvantage of the competition setting (relatively inexperienced people botching together things without the time for proper research) disappears.
\n\nFor me personally the competition was an eye-opener. I have been fascinated with physics for all of my life, and while I always had respect and admiration for the experimentalists, my interests focused mostly on the theoretical side. This competition taught me how incredibly challenging even a small single step in the data processing pipeline of a modern experiment can be. From now on, I will look at the nice plots in LHC papers with different eyes.
\n\nThank you!
", "rawMarkdown": "I want to explicitly thank the organizers for putting up this competition. It is clear that everyone involved took a substantial risk by investing a lot of time in preparing and running the competition, compared to which the price money is probably a small item. There were difficult trade-offs to make between the approachability of the problem to a large enough audience and realism. From my limited perspective, I'd say the organizers made very good choices in this respect.\n\nThe scientific outcome was (and is) uncertain. It is entirely possible that at the end the scientific bottom line might be zero (if one does not count negative results). If so, it will be a bitter pill to swallow, but I wouldn't see it as a failure of either the organizers or the participants. Innovation and exceptional ideas simply cannot be guaranteed to happen in a prescribed time and setting. Only speaking for my solution, I know that I did not manage to come up with some truly innovative, let alone game-changing ideas. It would have been great, but the prior probability for something like that is necessarily very low.\n\nIt is also entirely possible that by analyzing results of this phase and the next one, some great ideas can be extracted. The big advantage of such a semi-realistic competition setting is that phony ideas which only work on toy problems are immediately filtered out. (In my field of expertise the academic literature is full of such ideas. I assume it is less so in HEP and ML, but still not zero.)\n\nOne should also not overlook a potential \"long tail\" effect of this competition: The trackml dataset is now readily available and well documented, and it can serve as a benchmark for people in the HEP or ML communities who may come up with great ideas long after the competitions have concluded. In this context, the main disadvantage of the competition setting (relatively inexperienced people botching together things without the time for proper research) disappears.\n\nFor me personally the competition was an eye-opener. I have been fascinated with physics for all of my life, and while I always had respect and admiration for the experimentalists, my interests focused mostly on the theoretical side. This competition taught me how incredibly challenging *even a small single step* in the data processing pipeline of a modern experiment can be. From now on, I will look at the nice plots in LHC papers with different eyes.\n\nThank you!", "votes": null }, { "id": "370768", "postDate": "08/15/2018 12:37:14", "content": "Well said Edwin and thanks to the organizers. I have learned a lot from this competition.
", "rawMarkdown": "Well said Edwin and thanks to the organizers. I have learned a lot from this competition.", "votes": null } ], "comments": [ { "id": 370768, "author_name": "jackvial", "author_url": "", "post_date": "08/15/2018 12:37:14", "content": "Well said Edwin and thanks to the organizers. I have learned a lot from this competition.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "370766": "I want to explicitly thank the organizers for putting up this competition. It is clear that everyone involved took a substantial risk by investing a lot of time in preparing and running the competition, compared to which the price money is probably a small item. There were difficult trade-offs to make between the approachability of the problem to a large enough audience and realism. From my limited perspective, I'd say the organizers made very good choices in this respect.\n\nThe scientific outcome was (and is) uncertain. It is entirely possible that at the end the scientific bottom line might be zero (if one does not count negative results). If so, it will be a bitter pill to swallow, but I wouldn't see it as a failure of either the organizers or the participants. Innovation and exceptional ideas simply cannot be guaranteed to happen in a prescribed time and setting. Only speaking for my solution, I know that I did not manage to come up with some truly innovative, let alone game-changing ideas. It would have been great, but the prior probability for something like that is necessarily very low.\n\nIt is also entirely possible that by analyzing results of this phase and the next one, some great ideas can be extracted. The big advantage of such a semi-realistic competition setting is that phony ideas which only work on toy problems are immediately filtered out. (In my field of expertise the academic literature is full of such ideas. I assume it is less so in HEP and ML, but still not zero.)\n\nOne should also not overlook a potential \"long tail\" effect of this competition: The trackml dataset is now readily available and well documented, and it can serve as a benchmark for people in the HEP or ML communities who may come up with great ideas long after the competitions have concluded. In this context, the main disadvantage of the competition setting (relatively inexperienced people botching together things without the time for proper research) disappears.\n\nFor me personally the competition was an eye-opener. I have been fascinated with physics for all of my life, and while I always had respect and admiration for the experimentalists, my interests focused mostly on the theoretical side. This competition taught me how incredibly challenging *even a small single step* in the data processing pipeline of a modern experiment can be. From now on, I will look at the nice plots in LHC papers with different eyes.\n\nThank you!", "370768": "Well said Edwin and thanks to the organizers. I have learned a lot from this competition." }, "source": "meta" }