{
  "id": 6784,
  "title": "Thoughts on mergers?",
  "url": "/competitions/flight2-main/discussion/6784",
  "author_name": "",
  "post_date": "2014-01-09T06:59:09.267Z",
  "votes": null,
  "comment_count": 12,
  "views": 6216,
  "content": "<p>There are just a couple of days to go until the merger deadline. Haven't seen much activity along those lines, other than Taki and Chris who have combined to form a Hungarian powerhouse.</p>\n<p>I wonder if there is a way to figure out potential synergy before merging with another team. Possibly trade landing and takeoff files for Sep 10th? Maybe even entire routes for Sep 10th to see if they can be improved upon by the other team's algorithm? I am thinking of posting mine to the forum.</p>",
  "messages": [
    {
      "id": "37153",
      "postDate": "01/09/2014 06:59:09",
      "content": "<p>There are just a couple of days to go until the merger deadline. Haven't seen much activity along those lines, other than Taki and Chris who have combined to form a Hungarian powerhouse.</p>\n<p>I wonder if there is a way to figure out potential synergy before merging with another team. Possibly trade landing and takeoff files for Sep 10th? Maybe even entire routes for Sep 10th to see if they can be improved upon by the other team's algorithm? I am thinking of posting mine to the forum.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37174",
      "postDate": "01/09/2014 19:24:45",
      "content": "<p>I think this is very good idea.</p>\n<p>my suggestion:</p>\n<p>I believe i have a good route optimization , and my drawback is with the files prediction. if someone think s/he has a good files prediction and need help with optimization we may have a good match. in order to test the match, s/he can send me the files (not enough to expose the algorithm) and i will run the optimization and submit the file. if we will see that the submissiona get a good score we will have a good reason to have a team together.</p>\n<p>anyone ?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37187",
      "postDate": "01/10/2014 01:10:18",
      "content": "<p>I have added a few files to a <a href=\"https://github.com/anlthms/fq2\">github respository</a> due to popular demand (okay, there was only one request, but here it is anyway).</p>\n<p>The file <a href=\"https://github.com/anlthms/fq2/blob/master/sep10waypoints.csv\">sep10waypoints.csv </a>contains optimized waypoints for Sep 10. Here's a visual of the paths taken by the flights when this file is fed to the simulator:</p>\n<p><img src=\"https://raw.github.com/anlthms/fq2/master/sep10routes.png\" alt=\"Flight paths\" width=\"930\" height=\"545\"></p>\n<p>The mean cost for the above routes is $11241.71.</p>\n<p>@sparrow, prediction is my weak spot as well, so not sure if I can help you there. If you would like to try anyway, please get <a href=\"https://github.com/anlthms/fq2/blob/master/predictedLandings_20130910_1803.csv\">predictedLandings.csv</a> and <a href=\"https://github.com/anlthms/fq2/blob/master/predictedTakeoffs_20130910_1803.csv\">predictedTakeoffs.csv</a> and run your optimizer for Sep 10th using those files for actualLandingFiles and actualTakeoffFiles respectively. CAUTION: I saved the counts as floating point numbers, so if you want to use those files with the simulator, you will need to hack the code first (fairly simple changes in Serialization.fs and Arrival.fs).</p>\n<p>I haven't done any work towards wind velocity prediction. The file predictedweather_20130910_1803.txt.gz was generated from the NOAA GRIB files shared by TecS.</p>\n<p>@admins, I tried to access the rules to make sure that sharing these files is in fact allowed, but the file is currently offline. From what I remember of the rules, this doesn't run afoul in any way. Let me know otherwise.</p>\n<p>EDIT: Comments welcome... If anyone thinks that I am making a mistake somewhere or missing out on some potentially rewarding optimization step, I would really like to know :-)</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37189",
      "postDate": "01/10/2014 01:39:16",
      "content": "<p>mean cost of the routes is $11,241.71... so if I'm running your routes through the simulator should I get that value using the provided live files? I'm asking as I'm getting $11,544.60 .. I wonder if it's just me ... :O</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37193",
      "postDate": "01/10/2014 02:03:56",
      "content": "<p>I'll double check and get back to you. It is possible that some of my local code tweaks are affecting the score.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37194",
      "postDate": "01/10/2014 02:13:33",
      "content": "<p>I was running the simulator version 2.01 - apologizes!!!</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37196",
      "postDate": "01/10/2014 03:10:44",
      "content": "<p>Not a problem. Thanks for giving it a whirl.</p>\n<p>I should add that I can get down to around $11,235 by adding more waypoints and optimizing further. Going to run it a bit longer to see when it levels out...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37253",
      "postDate": "01/10/2014 20:17:26",
      "content": "<p>I would be open to a merger. &nbsp;At the very least, it couldn't hurt to bounce ideas around.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37279",
      "postDate": "01/11/2014 11:16:08",
      "content": "<p>Hi Anil,</p>\n<p>My thoughts about methodology to compare results in your way (Sep10) . if we want to compare the route optimization i think we need to run the optimization and evualte it with the original files that were provided for Sep10 which are the full known data. if we want to comapre the prediction part i think that we need to run the optimization with the prediction files and evualte with the original files (with the full known data).</p>\n<p>what you did and how did you calculate your score ?</p>\n<p>i ran the route optimization and evulate with the orginal files and got score of 11,222. i ran only on one stage of my optimization so i believe i can gain at least ~10 points with more running time.</p>\n<p>by the way what about groundConditions file ?</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37280",
      "postDate": "01/11/2014 12:41:40",
      "content": "<p>Hi kevin,</p>\n<p>if what i have suggested looks reasonable to you let's talk.</p>\n<p>you can write me to: e43uhaj+tov84c@sharklasers.com</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37289",
      "postDate": "01/11/2014 18:24:49",
      "content": "<p>I'm also open to make a team with some other one! I spend a lot of time with optimization of routes. But I had no time to work on predicting &quot;actual landings&quot;, &quot;actual takeoffs&quot;, ground conditions and weather. With using the full date set for Sep 10, I end up somewhere around $11,165 average cost per route. So if some of you has a good model for predicting the input date and is open for collaboration, let me know!</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37295",
      "postDate": "01/11/2014 19:43:59",
      "content": "<p>[quote=sparrow ;37279]</p>\n<p>...if we want to comapre the prediction part i think that we need to run the optimization with the prediction files and evualte with the original files (with the full known data).</p>\n<p>[/quote]</p>\n<p>Yes, that's what I did. I should have stated that in my post. Optimizing with the original files would be misleading as it would just overfit.</p>\n<p>As you said, optimizing with the full known data could be a measure of how good the optimizer is. Of course, an optimizer that posts the lowest cost on the known data may not lead to lowest cost on the test data. The tricky part is to constrain the optimizer to reduce overfitting. I am still struggling with that.</p>\n<p>I didn't predict ground conditions and instead used the provided conditions from July 4th.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "37341",
      "postDate": "01/13/2014 10:57:11",
      "content": "<p>[quote=DerDasMachenMuss;37289]</p>\n<p>I'm also open to make a team with some other one! </p>\n<p>[/quote]</p>\n<p>Hi DerDasMachenMuss,</p>\n<p>I would be very interested in a merger as well.Unfortunately, the deadline for mergers has passed...</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 37174,
      "author_name": "drorrr",
      "author_url": "",
      "post_date": "01/09/2014 19:24:45",
      "content": "<p>I think this is very good idea.</p>\n<p>my suggestion:</p>\n<p>I believe i have a good route optimization , and my drawback is with the files prediction. if someone think s/he has a good files prediction and need help with optimization we may have a good match. in order to test the match, s/he can send me the files (not enough to expose the algorithm) and i will run the optimization and submit the file. if we will see that the submissiona get a good score we will have a good reason to have a team together.</p>\n<p>anyone ?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37187,
      "author_name": "anlthms",
      "author_url": "",
      "post_date": "01/10/2014 01:10:18",
      "content": "<p>I have added a few files to a <a href=\"https://github.com/anlthms/fq2\">github respository</a> due to popular demand (okay, there was only one request, but here it is anyway).</p>\n<p>The file <a href=\"https://github.com/anlthms/fq2/blob/master/sep10waypoints.csv\">sep10waypoints.csv </a>contains optimized waypoints for Sep 10. Here's a visual of the paths taken by the flights when this file is fed to the simulator:</p>\n<p><img src=\"https://raw.github.com/anlthms/fq2/master/sep10routes.png\" alt=\"Flight paths\" width=\"930\" height=\"545\"></p>\n<p>The mean cost for the above routes is $11241.71.</p>\n<p>@sparrow, prediction is my weak spot as well, so not sure if I can help you there. If you would like to try anyway, please get <a href=\"https://github.com/anlthms/fq2/blob/master/predictedLandings_20130910_1803.csv\">predictedLandings.csv</a> and <a href=\"https://github.com/anlthms/fq2/blob/master/predictedTakeoffs_20130910_1803.csv\">predictedTakeoffs.csv</a> and run your optimizer for Sep 10th using those files for actualLandingFiles and actualTakeoffFiles respectively. CAUTION: I saved the counts as floating point numbers, so if you want to use those files with the simulator, you will need to hack the code first (fairly simple changes in Serialization.fs and Arrival.fs).</p>\n<p>I haven't done any work towards wind velocity prediction. The file predictedweather_20130910_1803.txt.gz was generated from the NOAA GRIB files shared by TecS.</p>\n<p>@admins, I tried to access the rules to make sure that sharing these files is in fact allowed, but the file is currently offline. From what I remember of the rules, this doesn't run afoul in any way. Let me know otherwise.</p>\n<p>EDIT: Comments welcome... If anyone thinks that I am making a mistake somewhere or missing out on some potentially rewarding optimization step, I would really like to know :-)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37189,
      "author_name": "alzmcr",
      "author_url": "",
      "post_date": "01/10/2014 01:39:16",
      "content": "<p>mean cost of the routes is $11,241.71... so if I'm running your routes through the simulator should I get that value using the provided live files? I'm asking as I'm getting $11,544.60 .. I wonder if it's just me ... :O</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37193,
      "author_name": "anlthms",
      "author_url": "",
      "post_date": "01/10/2014 02:03:56",
      "content": "<p>I'll double check and get back to you. It is possible that some of my local code tweaks are affecting the score.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37194,
      "author_name": "alzmcr",
      "author_url": "",
      "post_date": "01/10/2014 02:13:33",
      "content": "<p>I was running the simulator version 2.01 - apologizes!!!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37196,
      "author_name": "anlthms",
      "author_url": "",
      "post_date": "01/10/2014 03:10:44",
      "content": "<p>Not a problem. Thanks for giving it a whirl.</p>\n<p>I should add that I can get down to around $11,235 by adding more waypoints and optimizing further. Going to run it a bit longer to see when it levels out...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37253,
      "author_name": "kevingray",
      "author_url": "",
      "post_date": "01/10/2014 20:17:26",
      "content": "<p>I would be open to a merger. &nbsp;At the very least, it couldn't hurt to bounce ideas around.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37279,
      "author_name": "drorrr",
      "author_url": "",
      "post_date": "01/11/2014 11:16:08",
      "content": "<p>Hi Anil,</p>\n<p>My thoughts about methodology to compare results in your way (Sep10) . if we want to compare the route optimization i think we need to run the optimization and evualte it with the original files that were provided for Sep10 which are the full known data. if we want to comapre the prediction part i think that we need to run the optimization with the prediction files and evualte with the original files (with the full known data).</p>\n<p>what you did and how did you calculate your score ?</p>\n<p>i ran the route optimization and evulate with the orginal files and got score of 11,222. i ran only on one stage of my optimization so i believe i can gain at least ~10 points with more running time.</p>\n<p>by the way what about groundConditions file ?</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37280,
      "author_name": "drorrr",
      "author_url": "",
      "post_date": "01/11/2014 12:41:40",
      "content": "<p>Hi kevin,</p>\n<p>if what i have suggested looks reasonable to you let's talk.</p>\n<p>you can write me to: e43uhaj+tov84c@sharklasers.com</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37289,
      "author_name": "derdasmachenmuss",
      "author_url": "",
      "post_date": "01/11/2014 18:24:49",
      "content": "<p>I'm also open to make a team with some other one! I spend a lot of time with optimization of routes. But I had no time to work on predicting &quot;actual landings&quot;, &quot;actual takeoffs&quot;, ground conditions and weather. With using the full date set for Sep 10, I end up somewhere around $11,165 average cost per route. So if some of you has a good model for predicting the input date and is open for collaboration, let me know!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37295,
      "author_name": "anlthms",
      "author_url": "",
      "post_date": "01/11/2014 19:43:59",
      "content": "<p>[quote=sparrow ;37279]</p>\n<p>...if we want to comapre the prediction part i think that we need to run the optimization with the prediction files and evualte with the original files (with the full known data).</p>\n<p>[/quote]</p>\n<p>Yes, that's what I did. I should have stated that in my post. Optimizing with the original files would be misleading as it would just overfit.</p>\n<p>As you said, optimizing with the full known data could be a measure of how good the optimizer is. Of course, an optimizer that posts the lowest cost on the known data may not lead to lowest cost on the test data. The tricky part is to constrain the optimizer to reduce overfitting. I am still struggling with that.</p>\n<p>I didn't predict ground conditions and instead used the provided conditions from July 4th.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 37341,
      "author_name": "julesvanligtenberg",
      "author_url": "",
      "post_date": "01/13/2014 10:57:11",
      "content": "<p>[quote=DerDasMachenMuss;37289]</p>\n<p>I'm also open to make a team with some other one! </p>\n<p>[/quote]</p>\n<p>Hi DerDasMachenMuss,</p>\n<p>I would be very interested in a merger as well.Unfortunately, the deadline for mergers has passed...</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "37153": "",
    "37174": "",
    "37187": "",
    "37189": "",
    "37193": "",
    "37194": "",
    "37196": "",
    "37253": "",
    "37279": "",
    "37280": "",
    "37289": "",
    "37295": "",
    "37341": ""
  },
  "source": "meta"
}