Tiny embedded grains. Specks. Inclusions

Discuss your experiences with and ideas about Stardust@home here.

Moderators: Stardust@home Team, DustMods

Driven1
Posts: 20
Joined: Sat Aug 05, 2006 7:27 am

Me too...

Post by Driven1 »

Just got burned on the same thing. I don't know if it was the same movie or not though.

I had been clicking them too because there had been no official answer on the subject of inclusions. Why miss the possible discovery of something without having a real answer on what exactly an inclusion looks like? Heck, were not even 100% on what a real track looks like. We certainly have ideas of what they should look like, and that's what we've all been looking for. As for the width of a track, we don't know that either. So given resolution of the movies, and the resolution of my screen, who knows, maybe one of those specks IS a track.

So I suppose that after all of the questions aimed at the Stardust Team about the status of the "Inclusion" matter, and with no info showing up from them relating to it, this is apparently the official answer. Don't click the specks!
Elli
Posts: 2
Joined: Fri Aug 11, 2006 8:09 pm
Location: Michigan

specks

Post by Elli »

I am seeing silver and black specks at times deep in the gel. They have a reverse focus pattern from the surface dust. They do not look like the small track example shown. These specks look like surface dust imbedded in the gel. Some seem to have indentations leading to them. Should they be selected? I selected them, but my thinking is that the particles we're looking for are too small to be seen. The directions said to expect the unexpected, so I suppose the experts will decide if such movies are worth further exploration.
icebike

Re: specks

Post by icebike »

Elli wrote: my thinking is that the particles we're looking for are too small to be seen..
I don't think I ever saw that mentioned anywhere. In fact several of the training films showed particles, especially when the entered the areogel at an angle, and were not sitting too deep.

I'd mark them if you think they are sub-surface and have an entry track.
Kalman
Posts: 11
Joined: Wed Aug 09, 2006 5:18 am
Location: Budapest, Hungary

Post by Kalman »

I think that losing a few percents from specificity is not so much problem, if your sensitivity stays high. It means that you have a good chance to recognize the real thing meanwhile you produce some false positives. I have a guess that the database is full of false positives for now, so it does'not matter to have a few more.

I walked into it 3 times so far, clicking on a blank calibration movie. The first time I felt very disappointed, but for now I have realised that I've met a kind of artifect aech time and I don't click them anymore.
Don't forget, this is not about scores but learning to find something useful. If you think what you see is important, be brave and click on it. However it's worthy to note the thing that you've clicked.
--Kalman
Elli
Posts: 2
Joined: Fri Aug 11, 2006 8:09 pm
Location: Michigan

specks

Post by Elli »

Thanks Icebike. I'll continue to mark this type of movie. If you'll look at the first page of the Tutorial Session, this is what is written: "The particles themselves are too small to see. You will be using a Virtual Microscope (VM) to look for the tracks made by the particles rather than the particles themselves. The tracks are just below the surface of the aerogel." From this I was assuming that the particles in the tracks that were visible in the training movies were dust from the surface being carried down into the gel.
Driven1
Posts: 20
Joined: Sat Aug 05, 2006 7:27 am

See this thread...

Post by Driven1 »

tiburd
Posts: 24
Joined: Thu Aug 03, 2006 7:47 pm
Location: Wyoming

Type 1 vs. Type 2 error

Post by tiburd »

On the original question "Am I too trigger-happy?" -- The cost to the team of a false positive is additional time spent ruling it out. The cost of a false negative, if EVERYONE misses a legitimate track, is a lost interstellar particle -- much worse in the long run, especially if we only find a handful. In that regard, I wonder -- can the team easily set up an additional list (besides our Data Viewed list) that indicates which of the movies we each viewed and DIDN'T click on but for which at least one other observer did click on something? That would give all of us an idea of how many "items" we might be missing in the actual movies...something that the complementary accuracy counts in the calibration movies might not track well, since interesting items in the real movies may look different from those in the calibration movies.
tiburd
icebike

Re: Type 1 vs. Type 2 error

Post by icebike »

tiburd wrote:On the original question "Am I too trigger-happy?" -- The cost to the team of a false positive is additional time spent ruling it out. The cost of a false negative, if EVERYONE misses a legitimate track, is a lost interstellar particle -- much worse in the long run, especially if we only find a handful.
But the the fact that clicked movies are sent to at least 100 other viewers tends to balance out either of these two risks. The cream of the crop are then selected for review. Its a self healing sort of system. I'm not sure a false positive hurts too much. And false negatives will be caught by others.

tiburd wrote: In that regard, I wonder -- can the team easily set up an additional list (besides our Data Viewed list) that indicates which of the movies we each viewed and DIDN'T click on but for which at least one other observer did click on something? That would give all of us an idea of how many "items" we might be missing in the actual movies...something that the complementary accuracy counts in the calibration movies might not track well, since interesting items in the real movies may look different from those in the calibration movies.
Wouldn't this tend to get every body evaluating movies in the SAME way? This is not necessarily a Good thing. At least not until there is a better idea from the scientists of exactly what to look for.

QUOTATION: A foolish consistency is the hobgoblin of little minds, adored by little statesmen and philosophers and divines.
Ralph Waldo Emerson
[/i]
niejell
Posts: 25
Joined: Wed Aug 02, 2006 1:10 pm
Location: Maryland, USA

Post by niejell »

For what it's worth - I view the calibration movies as teachers. Last week I was clicking on specks and receiving "correct answer" in calibration movies. Lately I don't see these very small specks anymore (no positive reenforcement). I think if these were truely bogus though, there would be calibration movies with specks that if clicked on would be incorrect (negative reenforement).
Shakin da tree...
niejell
Posts: 25
Joined: Wed Aug 02, 2006 1:10 pm
Location: Maryland, USA

Post by niejell »

OK - maybe I need to rethink one point. Had three calibration movies in the last hour with small closed specks.
Shakin da tree...
icebike

Post by icebike »

niejell wrote:OK - maybe I need to rethink one point. Had three calibration movies in the last hour with small closed specks.
I haven't seen any with a small closed spec, and I saw over a hundred CMs in the last sitting. I am seeing more small track CMs, some no bigger than this "o" but no fully closed specs.
geckzilla
Posts: 30
Joined: Mon Aug 07, 2006 11:05 am

Post by geckzilla »

The tiniest CMs do indeed not have a hollowed center because they are so small that it's not visible in the movie. However, they are still clearly different to the trained eye. Only the largest "closed" specks even come close to rivalling the smallest CM tracks. And the smallest CM tracks are in focus in 3-5 frames while the largest so-called inclusions are in focus for only 1-2 frames. One is a tiny tunnel while the other is likely a single dust speck.

Standard disclaimer: My (relatively uninformed) opinion only. I could be completely and utterly wrong.
DiamondGirl
Posts: 77
Joined: Tue Jun 13, 2006 12:16 pm
Location: Syracuse, NY

Post by DiamondGirl »

Not that it's an answer, but don't the pre-training notes say to be sure and note anything, no matter how small? That's the directive I've been using.
geckzilla
Posts: 30
Joined: Mon Aug 07, 2006 11:05 am

Post by geckzilla »

I was following that until I found so many. I figured I've found my fair share.
niejell
Posts: 25
Joined: Wed Aug 02, 2006 1:10 pm
Location: Maryland, USA

Post by niejell »

I guess to clerify... I said closed speck - but that's how it appears. If the object is 2 pixels in diameter there is no way to display a "doughnut".
We're probably viewing movies 72 dpi, those checking the hi-rez film would see it more clearly.
Shakin da tree...
Post Reply