Dear Hosts,
\nI have visually observed some train samples and find out some strange samples.
\nSmall clarification: If we look at snippets of certain sample of ON TARGET part, we assume to get 3 images which looks pretty similar.
\nBut I have found out some strange samples:
ON TARGET differs from second and third ON TARGETON TARGET completely differs from third ON TARGETON TARGET snippets are really different Assuming I didn't make any typos while visualizing (unfortunately, it seems that Kaggle won't allow me to make any file uploads at the moment), here's a description for others' convenience…
\ntrain/0/00973367d6481e2.npy, the first ON TARGET snippet has some more Gaussian noise than the other twotrain/0/00f129775fd7792.npy, the first three snippets (incl. the first two ON TARGET ones) have a strong but not-narrow band of signal while the others have a lot of noise everywhere (with a squiggle of a signal in the third ON TARGETtrain/0/010b8dc237f2d74.npy, there's a strong, wide-band of \"signal\" in the latter half of ON TARGET 1 and the very beginning of OFF TARGET 1, then nothing for the remainder of OFF TARGET 1, then a weak continuation of the wide-band signal in ON TARGET 2, then a strong wide-band for OFF TARGET 2and the first half ofON TARGET 3, and then no such signal in the second half ofON TARGET3 andOFF TARGET 3`.Why might these snippets be anomalous? Setting aside the possibility of some unintentional error in data generation, it could well be intended to model some of the real-world inconveniences that thwart radio astronomers, esp. SETI ones. I've read anecdotes about mundane sources of interference ranging from someone microwaving their lunch during some of the snippets to a nearby clock or wristwatch (I think the latter was responsible for the candidate signal associated with Proxima Centauri that got some media coverage lately, BLC-1. Sofia Sheikh presented on it at the Breakthrough Discuss 2021 event, but I'm not sure if any paper has been posted yet.)
", "rawMarkdown": "Assuming I didn't make any typos while visualizing (unfortunately, it seems that Kaggle won't allow me to make any file uploads at the moment), here's a description for others' convenience...\n\n1. In `train/0/00973367d6481e2.npy`, the first `ON TARGET` snippet has some more Gaussian noise than the other two\n2. In `train/0/00f129775fd7792.npy`, the first three snippets (incl. the first two `ON TARGET` ones) have a strong but not-narrow band of signal while the others have a lot of noise everywhere (with a squiggle of a signal in the third `ON TARGET`\n3. In `train/0/010b8dc237f2d74.npy`, there's a strong, wide-band of \"signal\" in the latter half of `ON TARGET` 1 and the very beginning of `OFF TARGET` 1, then nothing for the remainder of `OFF TARGET` 1, then a weak continuation of the wide-band signal in `ON TARGET` 2, then a strong wide-band for `OFF TARGET` 2` and the first half of `ON TARGET 3`, and then no such signal in the second half of `ON TARGET` 3 and `OFF TARGET 3`.\n\nWhy might these snippets be anomalous? Setting aside the possibility of some unintentional error in data generation, it could well be intended to model some of the real-world inconveniences that thwart radio astronomers, esp. SETI ones. I've read anecdotes about mundane sources of interference ranging from someone microwaving their lunch during some of the snippets to a nearby clock or wristwatch (I think the latter was responsible for the candidate signal associated with Proxima Centauri that got some media coverage lately, BLC-1. Sofia Sheikh presented on it at the Breakthrough Discuss 2021 event, but I'm not sure if any paper has been posted yet.)", "votes": 1, "replies": [ { "id": 1449841, "postDate": "2021-08-05T01:21:20.287Z", "content": " I've read anecdotes about mundane sources of interference ranging from someone microwaving their lunch during some of the snippets to a nearby clock or wristwatch (I think the latter was responsible for the candidate signal associated with Proxima Centauri that got some media coverage lately, BLC-1. Sofia Sheikh presented on it at the Breakthrough Discuss 2021 event, but I'm not sure if any paper has been posted yet.)\nEmerges from rabbit hole… A video of the conference can be found here.
\nhttps://twitter.com/brkthroughprize/status/1381691086821810179
\nSofia Sheikh's 20 minute talk starts at the 4 hour, 32 minute mark. She mentioned this Kaggle contest. I thought it was interesting.
Dear Hosts,
\nI have visually observed some train samples and find out some strange samples.
\nSmall clarification: If we look at snippets of certain sample of ON TARGET part, we assume to get 3 images which looks pretty similar.
\nBut I have found out some strange samples:
ON TARGET differs from second and third ON TARGETON TARGET completely differs from third ON TARGETON TARGET snippets are really different Assuming I didn't make any typos while visualizing (unfortunately, it seems that Kaggle won't allow me to make any file uploads at the moment), here's a description for others' convenience…
\ntrain/0/00973367d6481e2.npy, the first ON TARGET snippet has some more Gaussian noise than the other twotrain/0/00f129775fd7792.npy, the first three snippets (incl. the first two ON TARGET ones) have a strong but not-narrow band of signal while the others have a lot of noise everywhere (with a squiggle of a signal in the third ON TARGETtrain/0/010b8dc237f2d74.npy, there's a strong, wide-band of \"signal\" in the latter half of ON TARGET 1 and the very beginning of OFF TARGET 1, then nothing for the remainder of OFF TARGET 1, then a weak continuation of the wide-band signal in ON TARGET 2, then a strong wide-band for OFF TARGET 2and the first half ofON TARGET 3, and then no such signal in the second half ofON TARGET3 andOFF TARGET 3`.Why might these snippets be anomalous? Setting aside the possibility of some unintentional error in data generation, it could well be intended to model some of the real-world inconveniences that thwart radio astronomers, esp. SETI ones. I've read anecdotes about mundane sources of interference ranging from someone microwaving their lunch during some of the snippets to a nearby clock or wristwatch (I think the latter was responsible for the candidate signal associated with Proxima Centauri that got some media coverage lately, BLC-1. Sofia Sheikh presented on it at the Breakthrough Discuss 2021 event, but I'm not sure if any paper has been posted yet.)
", "votes": 1, "replies": [ { "id": 1449841, "author_name": "glazed", "author_url": "", "post_date": "2021-08-05T01:21:20.287000", "content": " I've read anecdotes about mundane sources of interference ranging from someone microwaving their lunch during some of the snippets to a nearby clock or wristwatch (I think the latter was responsible for the candidate signal associated with Proxima Centauri that got some media coverage lately, BLC-1. Sofia Sheikh presented on it at the Breakthrough Discuss 2021 event, but I'm not sure if any paper has been posted yet.)\nEmerges from rabbit hole… A video of the conference can be found here.
\nhttps://twitter.com/brkthroughprize/status/1381691086821810179
\nSofia Sheikh's 20 minute talk starts at the 4 hour, 32 minute mark. She mentioned this Kaggle contest. I thought it was interesting.