Stardust@home Suggestion Thread
Moderators: Stardust@home Team, DustMods
suggestions
long time lurker, first time poster.
A couple of suggestions.
I would rather know the "figure of merit" value (the specificity for x sensitivity against or which ever it is) for a possible track than the "for/against" count. I guess the total count is useful too, especially if there is a set number a movie has to be viewed by.
The possibility of CMs for which the correct answer is 'bad focus'. I wonder how often something is marked 'no track' when in reality we didn't have enough depth to really make the distinction, or the focus was completely above the surface except for one small part.
Additional 'bad focus' buttons have been suggested before and I would tend to agree with the idea, especially as it might speed up the process for the stardust team.
Kindda off topic, but I am a scientist (chemistry) and just want to remind you guys every publication is a good one. How about writing up the process of getting this artificial-artificial-intelligence project going, and functioning. There must be interest in doing this for other science projects. Or the social science aspects of it. Get in someone from the field in, and throw data at them.
Thats all. Finally I want to remind everyone the aim isn't to get a high score, but to accurately determine tracks and to thank y'all for the chance to participate.
A couple of suggestions.
I would rather know the "figure of merit" value (the specificity for x sensitivity against or which ever it is) for a possible track than the "for/against" count. I guess the total count is useful too, especially if there is a set number a movie has to be viewed by.
The possibility of CMs for which the correct answer is 'bad focus'. I wonder how often something is marked 'no track' when in reality we didn't have enough depth to really make the distinction, or the focus was completely above the surface except for one small part.
Additional 'bad focus' buttons have been suggested before and I would tend to agree with the idea, especially as it might speed up the process for the stardust team.
Kindda off topic, but I am a scientist (chemistry) and just want to remind you guys every publication is a good one. How about writing up the process of getting this artificial-artificial-intelligence project going, and functioning. There must be interest in doing this for other science projects. Or the social science aspects of it. Get in someone from the field in, and throw data at them.
Thats all. Finally I want to remind everyone the aim isn't to get a high score, but to accurately determine tracks and to thank y'all for the chance to participate.
Smaller header
Thanks to whomever has recently responded to an old plea in this forum by reducing the height of the VM header, thus sparing the need for repeated scrolling to view the whole movie frame.
-
- Stardust@home Team
- Posts: 530
- Joined: Wed Apr 19, 2006 11:28 am
- Location: UC Berkeley Space Sciences Lab
- Contact:
Re: Smaller header
The thanks go to Robert Letteri and Will Marchant, two of our team with all the technical know-how who had heard your pleas from long ago and finally had a free moment to make the change.jsmaje wrote:Thanks to whomever has recently responded to an old plea in this forum by reducing the height of the VM header, thus sparing the need for repeated scrolling to view the whole movie frame.
-Bryan
"I am made from the dust of the stars, and the oceans flow in my veins"
- RUSH
- RUSH
Alpha List
I suggest that the My Events list contain an Alpha indication in the Official Status field. Further, when the Red Team completes its evaluation, that the Official Status be changed to "Passed by Red Team" or "Rejected by Red Team" (or something to that effect).
Bruce
-
- Posts: 4
- Joined: Tue Jun 16, 2009 5:03 am
- Location: South Australia
Additional Controls for Microscope
would be nice to have an up & down arrow to single click through frames. i do an autoscan, then scan manually, then i go to spots of interest and take a better look and it is here that being able to skip up and down a couple of frames would make life so much easier ;-} you might also get an increase in possible hits through user sensitivity
Re: Additional Controls for Microscope
Interesting idea, I'll pass the suggestion along. Thanks!wickedKlown wrote:would be nice to have an up & down arrow to single click through frames. i do an autoscan, then scan manually, then i go to spots of interest and take a better look and it is here that being able to skip up and down a couple of frames would make life so much easier ;-} you might also get an increase in possible hits through user sensitivity
Dan
-
- Posts: 1
- Joined: Sat Jan 15, 2011 12:40 pm
Re: Stardust@home Suggestion Thread
When I get a calibration movie wrong, I think that it would be useful if a message pops up and the movie reappears with an arrow pointing to the feature I should be seeing. That way I'll know what I'm doing wrong and be able to learn. As it is now I don't notice which calibration movies I've gotten wrong sometimes and have no way of improving my searching methods....
Re: Stardust@home Suggestion Thread
This is a suggestion for Phase 4. I agree with T.R. Hayward and others who have offered suggestions on new scoring systems, etc. Also that finding the most obscure and tiniest tracks should garner more points than finding the large obvious one will be welcomed.
One other suggestion I have from experience. I know the 50 micron CM's are supposed to have no track, and that I advance my score by clicking "No Track." (They are a welcome reward for working hard to find tracks in real 100 micron movies.) However, we have been told over and over to "expect the unexpected." So, when I saw something "unexpected" in a recent 50 CM that looked suspiciously like a track, I clicked on it---hoping to bring it to the attention of the research team. OOPS! My Specificity lowered to 99.99 immediately! (It has stayed there) I'll NEVER do that again. Could the new scoring system be set up in such a way that we will not be discouraged from clicking on "the unexpected?" Perhaps another option in addition to the "No Track" or "Bad Focus" at the bottom of the window we now have? The points I score and my position in the rankings should not be my primary goal, but unfortunately it is.
Evelyn - ERSTRS
One other suggestion I have from experience. I know the 50 micron CM's are supposed to have no track, and that I advance my score by clicking "No Track." (They are a welcome reward for working hard to find tracks in real 100 micron movies.) However, we have been told over and over to "expect the unexpected." So, when I saw something "unexpected" in a recent 50 CM that looked suspiciously like a track, I clicked on it---hoping to bring it to the attention of the research team. OOPS! My Specificity lowered to 99.99 immediately! (It has stayed there) I'll NEVER do that again. Could the new scoring system be set up in such a way that we will not be discouraged from clicking on "the unexpected?" Perhaps another option in addition to the "No Track" or "Bad Focus" at the bottom of the window we now have? The points I score and my position in the rankings should not be my primary goal, but unfortunately it is.
Evelyn - ERSTRS
Re: Stardust@home Suggestion Thread
Hi Evelyn,
Given what you've said here, I think you'll like Phase 4. We just recently tested the first iteration and decided to go back to the drawing board for a bit more tweaking, but many of the suggestions you've made here are included.
More soon to be sure!
Dan
Given what you've said here, I think you'll like Phase 4. We just recently tested the first iteration and decided to go back to the drawing board for a bit more tweaking, but many of the suggestions you've made here are included.
More soon to be sure!
Dan
Re: Stardust@home Suggestion Thread
Hello,
I have already searched through over 1.5k real movies in the phase IV (and much, much more in previous phases) so far so I consider myself to be quite experienced duster. I'd like to comment on something.
I noticed that many of power movies are similar to each other- they are usually quite bright, clear and in many cases there is a circular shaped feature present in most of the power movies. It is very easy to recognize which one is the power movie.
I am able to 'detect' the power movies with very high accuracy (>90%-my estimation). Usually when I have a strong feeling that I'm looking at power movie I spend much more time searching for a track. Because of that the probability of finding a track in such a movie is much higher than in other movies. I've clicked on several really tiny tracks that I would't have noticed I hadn't had feeling that I'm scanning through one of power movies.I believe that that I'm not the only one who's able to detect power movies and spend some extra time on them. I would suggest you to make the power movies more diversed -less characteristic features, some movies paritially out of focus, some of them bright, some of them darker, etc., so they are not so similar to each other. I think that would let you to measure our track detection skills better and values obtained this way would reflect our true power.
those are just two examples of the power movies that look similarly
http://stardustathome.ssl.berkeley.edu/ ... 6&theY=364
http://stardustathome.ssl.berkeley.edu/ ... 1&theY=265
Another thing- lets assume that somebody really wants to get really high in ranking. Wouldn't it be a good method to skip over those huge, low rated, obvious tracks by clicking the "bad focus" button and click only the smaller ones that are relatively easy to find (rated at 15-30)? I noticed that usually when I click on large track my power goes down (what is expected once you know the formula for the power) despite I actually detected the track. Maybe the formula should be changed? Some kind of weighted mean maybe, or personal histogram or graph showing detection probability versus point value of the tracks...?
And finally- something that was mentioned many times before. There is really a need for some kind of confirmation for track/no track. It does happen (not very often, but it does) that I accidentally click somwhere on the movie and you get a false click or on "no track" button- what is also possible knowing that no track buttons are located just above and just below the focus slider. It is not so difficult to get the false results while working with touchpad with tapping. Maybe it would be nice to let people choose if they want to have the confirmation popup by ticking some option in their profile.
I don't want to complain here or lecture you guys. I am just sharing my suggestions so you can improve your project.
And above all- big thanks to you to give me the opportunity to participate in this project- I've been with you since the very beggining.
PS.I hope it is understandable- English is not my first language.
I have already searched through over 1.5k real movies in the phase IV (and much, much more in previous phases) so far so I consider myself to be quite experienced duster. I'd like to comment on something.
I noticed that many of power movies are similar to each other- they are usually quite bright, clear and in many cases there is a circular shaped feature present in most of the power movies. It is very easy to recognize which one is the power movie.
I am able to 'detect' the power movies with very high accuracy (>90%-my estimation). Usually when I have a strong feeling that I'm looking at power movie I spend much more time searching for a track. Because of that the probability of finding a track in such a movie is much higher than in other movies. I've clicked on several really tiny tracks that I would't have noticed I hadn't had feeling that I'm scanning through one of power movies.I believe that that I'm not the only one who's able to detect power movies and spend some extra time on them. I would suggest you to make the power movies more diversed -less characteristic features, some movies paritially out of focus, some of them bright, some of them darker, etc., so they are not so similar to each other. I think that would let you to measure our track detection skills better and values obtained this way would reflect our true power.
those are just two examples of the power movies that look similarly
http://stardustathome.ssl.berkeley.edu/ ... 6&theY=364
http://stardustathome.ssl.berkeley.edu/ ... 1&theY=265
Another thing- lets assume that somebody really wants to get really high in ranking. Wouldn't it be a good method to skip over those huge, low rated, obvious tracks by clicking the "bad focus" button and click only the smaller ones that are relatively easy to find (rated at 15-30)? I noticed that usually when I click on large track my power goes down (what is expected once you know the formula for the power) despite I actually detected the track. Maybe the formula should be changed? Some kind of weighted mean maybe, or personal histogram or graph showing detection probability versus point value of the tracks...?
And finally- something that was mentioned many times before. There is really a need for some kind of confirmation for track/no track. It does happen (not very often, but it does) that I accidentally click somwhere on the movie and you get a false click or on "no track" button- what is also possible knowing that no track buttons are located just above and just below the focus slider. It is not so difficult to get the false results while working with touchpad with tapping. Maybe it would be nice to let people choose if they want to have the confirmation popup by ticking some option in their profile.
I don't want to complain here or lecture you guys. I am just sharing my suggestions so you can improve your project.
And above all- big thanks to you to give me the opportunity to participate in this project- I've been with you since the very beggining.
PS.I hope it is understandable- English is not my first language.
Re: Stardust@home Suggestion Thread
Hi Klemek,
Your English is just fine! Thanks for the suggestions and info. I will discuss all of this with the team.
Question: When you say "I noticed that usually when I click on large track my power goes down (what is expected once you know the formula for the power) despite I actually detected the track." - don't you really mean Skill instead of Power (i.e., Skill Score)?
Thanks again!
Dan
Your English is just fine! Thanks for the suggestions and info. I will discuss all of this with the team.
Question: When you say "I noticed that usually when I click on large track my power goes down (what is expected once you know the formula for the power) despite I actually detected the track." - don't you really mean Skill instead of Power (i.e., Skill Score)?
Thanks again!
Dan
Re: Stardust@home Suggestion Thread
Klemek wrote:Hello,
I believe that that I'm not the only one who's able to detect power movies and spend some extra time on them.
Great observations, no you are definetly not the only one that is somewhat able to detect those power movies and as a result spend more time on it. This is something that I started noticing during phase 3 actually. Nevertheless I try to spend as much time on real movies as I do on power movies, afterall real tracks are what I am looking for. There are a few darker power movies from what I have seen but its true that most of the time, I can detect which movie is a power movie, sometimes even before using the VM focus.Klemek wrote: Another thing- lets assume that somebody really wants to get really high in ranking. Wouldn't it be a good method to skip over those huge, low rated, obvious tracks by clicking the "bad focus" button and click only the smaller ones that are relatively easy to find (rated at 15-30)? I noticed that usually when I click on large track my power goes down
I was also concerned about your second point about the skipping the lower score power movies. I also do believe that these movies could be simply skipped by dusters only because it will have a negative effect on their score. I think that this is something to look into (work on a new formula for calculating skill rating), but then again, to me the score is only an indicator of how well you are doing and the real goal is finding those real tracks.
Martin
Re: Stardust@home Suggestion Thread
I mean skill.DanZ wrote: Question: When you say "I noticed that usually when I click on large track my power goes down (what is expected once you know the formula for the power) despite I actually detected the track." - don't you really mean Skill instead of Power (i.e., Skill Score)?
Dan