if there were the next G2Net Competition, the organizer can consider:
\nsome improvement points for this challenge
\nActually, there are many experts in kagglers, with more labels (target and auxiliary) for training, I think you can see more magical solution.
\nSome kagglers have \"super computers\" and can create more data for training as well.
", "messages": [ { "id": "1528324", "postDate": "09/29/2021 14:55:15", "content": "if there were the next G2Net Competition, the organizer can consider:
\nsome improvement points for this challenge
\nActually, there are many experts in kagglers, with more labels (target and auxiliary) for training, I think you can see more magical solution.
\nSome kagglers have \"super computers\" and can create more data for training as well.
", "rawMarkdown": "if there were the next G2Net Competition, the organizer can consider:\n- estimate the GW parameters (not just detection)\n- use additional data like radio, IR, UV, etc ...\n- we want to see real data! not just simulated ones \n\n----\n\nsome improvement points for this challenge\n- the organizer can actually release the simulation parameters (and maybe also the simulation software) in the train data (the parameters can remain hidden in the test data) at the start of the challenge. With the parameters known, one can do multi-task learning, denoising (since ground truth wave is known).\n\nActually, there are many experts in kagglers, with more labels (target and auxiliary) for training, I think you can see more magical solution.\n\nSome kagglers have \"super computers\" and can create more data for training as well.", "votes": null }, { "id": "1528366", "postDate": "09/29/2021 15:31:14", "content": "It'd also be nice for the organizers to stay connected in the community throughout the course of the competition. G2Net seemed active for a month or two, which was actually better than most other hosts who disappear entirely after the first week. If I was running a comp that I paid for, I'd be egging the contestants along until the last second.
", "rawMarkdown": "It'd also be nice for the organizers to stay connected in the community throughout the course of the competition. G2Net seemed active for a month or two, which was actually better than most other hosts who disappear entirely after the first week. If I was running a comp that I paid for, I'd be egging the contestants along until the last second.", "votes": null }, { "id": "1528992", "postDate": "09/30/2021 04:38:28", "content": "@michaeljwill @bayeswolf
\nNow, when the competition is over, it would be nice if organizers release the simulation parameters & software used to generate competition data. Some of us would maybe like to continue to play with this nice problem. Doing so with unlimited train data can entirely change the game.
Thanks for the feedback @hengck23! It's great to see there's interest in another G2Net competition. This was very much a learning process for us so it's great to hear the communities thoughts.
\nI'll try to give a bit of context for why we decided not to provide the parameters or data generation code. I think there were two main reasons:
\nHope this gives you a bit of insight into our thought process.
\nIf we do run another competition, I think we'll certainly consider releasing at least the parameters, if not the complete data generation code as well. Though that would also depend on what the exact challenge was.
", "rawMarkdown": "Thanks for the feedback @hengck23! It's great to see there's interest in another G2Net competition. This was very much a learning process for us so it's great to hear the communities thoughts. \n\nI'll try to give a bit of context for why we decided not to provide the parameters or data generation code. I think there were two main reasons:\n\n* We were concerned that if we released the simulation parameters and/or the data generation code the solution for the test data could be brute-forced, thus defeating the point of the challenge. We were also worried about something similar to what happened in the SETI competition happening here.\n* Each signal has 15+ parameters associated with it and understanding what these all mean and how they affect the final signal isn't simple. We felt that all of this extra information could make the competition less accessible since understanding it requires a reasonable amount of prior knowledge.\n\nHope this gives you a bit of insight into our thought process.\n\nIf we do run another competition, I think we'll certainly consider releasing at least the parameters, if not the complete data generation code as well. Though that would also depend on what the exact challenge was.", "votes": null }, { "id": "1537290", "postDate": "10/07/2021 10:57:29", "content": "@josefslavicek Glad to hear there might be continued interest in the data after the competition. We're certainly considering releasing the simulation parameters (I personally really hope we can) and maybe the generation code. If we do, we'll make sure it's clearly linked and available to everyone.
", "rawMarkdown": "josefslavicek Glad to hear there might be continued interest in the data after the competition. We're certainly considering releasing the simulation parameters (I personally really hope we can) and maybe the generation code. If we do, we'll make sure it's clearly linked and available to everyone.", "votes": null }, { "id": "1559908", "postDate": "10/27/2021 08:08:04", "content": "Hey All,
\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.
\nThank you all,
\nRegards,
\nChris
It'd also be nice for the organizers to stay connected in the community throughout the course of the competition. G2Net seemed active for a month or two, which was actually better than most other hosts who disappear entirely after the first week. If I was running a comp that I paid for, I'd be egging the contestants along until the last second.
", "votes": null, "replies": [] }, { "id": 1528992, "author_name": "josefslavicek", "author_url": "", "post_date": "09/30/2021 04:38:28", "content": "@michaeljwill @bayeswolf
\nNow, when the competition is over, it would be nice if organizers release the simulation parameters & software used to generate competition data. Some of us would maybe like to continue to play with this nice problem. Doing so with unlimited train data can entirely change the game.
@josefslavicek Glad to hear there might be continued interest in the data after the competition. We're certainly considering releasing the simulation parameters (I personally really hope we can) and maybe the generation code. If we do, we'll make sure it's clearly linked and available to everyone.
", "votes": null, "replies": [] } ] }, { "id": 1537286, "author_name": "michaeljwill", "author_url": "", "post_date": "10/07/2021 10:53:15", "content": "Thanks for the feedback @hengck23! It's great to see there's interest in another G2Net competition. This was very much a learning process for us so it's great to hear the communities thoughts.
\nI'll try to give a bit of context for why we decided not to provide the parameters or data generation code. I think there were two main reasons:
\nHope this gives you a bit of insight into our thought process.
\nIf we do run another competition, I think we'll certainly consider releasing at least the parameters, if not the complete data generation code as well. Though that would also depend on what the exact challenge was.
", "votes": null, "replies": [] }, { "id": 1559908, "author_name": "zerafachris", "author_url": "", "post_date": "10/27/2021 08:08:04", "content": "Hey All,
\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.
\nThank you all,
\nRegards,
\nChris