clump finding density threshold problem
Hi all, I am trying to find clumps in a few wind-tunnel runs of a galaxy with ram pressure stripping, so that I can measure the mass, volume, etc. I want to make sure, then, that I know how clump finder is selecting its contours. Unfortunately, I am not getting the minimum densities that I am expecting--although not all the time! I have used IsolatedGalaxy and run the same clump finder routine, but with different maximum densities for the contours. To reproduce my error I had to make the step size smaller--here I am using step = 1.1. Basically, I am not always stepping up my minimum density contour, even when the maximum density in the clump makes it clear I *should* be. I am attaching the script where I have cmax = 9.0e-24, which shows the error. You can see that in the figure I attach, looking at the green triangles. The y-axis is the maximum density in a clump, the x-axis is the minimum density. To guide your eye I have drawn lines at each step. Basically, no point should be above a black line and to the left of the connected vertical line (because then the max value is above the next possible minimum in the steps) *unless* the cmax value was reached. To make this really clear if you look at the upper-left green triangle, the max density is ~7e-24 while the minimum density is ~3.6e-24--there are a lot of steps between those values--why is the minimum density not 6.68e-24? Strangely enough, this problem does *not* show up when I have cmax = 6.7e-24 (cyan circles). I write out the max and min values in a text file at the end of the script, so I also attach the little script I use to make the figure. Any help would be *much* appreciated! Thank you all, Stephanie -- Dr. Stephanie Tonnesen Associate Research Scientist CCA, Flatiron Institute New York, NY stonnes@gmail.com
Hi Stephanie, Thanks for this detailed report. I am thinking through how to understand this and how to process the potential issue. On Tue, Feb 26, 2019 at 11:36 PM Stephanie Tonnesen <stonnes@gmail.com> wrote:
Hi all,
I am trying to find clumps in a few wind-tunnel runs of a galaxy with ram pressure stripping, so that I can measure the mass, volume, etc. I want to make sure, then, that I know how clump finder is selecting its contours. Unfortunately, I am not getting the minimum densities that I am expecting--although not all the time!
I have used IsolatedGalaxy and run the same clump finder routine, but with different maximum densities for the contours. To reproduce my error I had to make the step size smaller--here I am using step = 1.1.
Basically, I am not always stepping up my minimum density contour, even when the maximum density in the clump makes it clear I *should* be.
So I'm not 100% sure that I follow -- I was able to run the script you provided but I can't yet know how to diagnose because I don't fully understand the issue (my fault, not yours). But, I will note a few things. The value for c_max is actually used in the contour finding; when the contour tree is iterating over values in a given collection of cells, it simply passes over any that are not included in the range. What this may mean is that if you have a field that has a lot of variation across a small spread of values, you might see some contours existing and others not based on minor changes in the max value supplied. For instance, two contours that are connected by a value of c_max will not be connected if that c_max is slightly lowered, and then based on the validators chosen this can change the hierarchy of clumps at lower densities (including what constitutes a leaf). I'm not *sure* if that can account for the behavior, but I think I need to understand the problem better before I can provide additional diagnosis. Is it possible that you could illustrate this with, say, a 2D image that you draw or illustrate somehow?
I am attaching the script where I have cmax = 9.0e-24, which shows the error. You can see that in the figure I attach, looking at the green triangles. The y-axis is the maximum density in a clump, the x-axis is the minimum density. To guide your eye I have drawn lines at each step. Basically, no point should be above a black line and to the left of the connected vertical line (because then the max value is above the next possible minimum in the steps) *unless* the cmax value was reached. To make this really clear if you look at the upper-left green triangle, the max density is ~7e-24 while the minimum density is ~3.6e-24--there are a lot of steps between those values--why is the minimum density not 6.68e-24?
Strangely enough, this problem does *not* show up when I have cmax = 6.7e-24 (cyan circles).
My
I write out the max and min values in a text file at the end of the script, so I also attach the little script I use to make the figure.
Any help would be *much* appreciated!
Thank you all,
Stephanie
-- Dr. Stephanie Tonnesen Associate Research Scientist CCA, Flatiron Institute New York, NY
stonnes@gmail.com _______________________________________________ yt-users mailing list -- yt-users@python.org To unsubscribe send an email to yt-users-leave@python.org
Hi Matt, Thanks for the reply! I have been poking at this problem for a while now, so it can be hard for me to not make some jumps in my thought process. Perhaps this will help (or perhaps not!): Here is a projection plot from the script I sent you, with cmax = 9.0e-24. See the regions I circled? [image: image.png] Now, I circle the same regions in the *same* script but with cmax = 6.7e-24. [image: image.png] When I have the LOWER (6.7e-24) cmax, the contours are smaller--> at a higher density! For some reason, the 9.0e-24 cmax value is not making all the steps that the 6.7e-24 cmax value manages. Okay, and then I am re-attaching my figure where each point should be a leaf clump, and I have the max vs the min density of the clump. I have circles and written some stuff on it, so perhaps it makes a *bit* more sense now! (oh, and "no vals" in the legend means no validators. everything has a cmin=1.6e-24). [image: image.png] Let me know if I have managed to clarify anything!! Thanks again, Stephanie -- Dr. Stephanie Tonnesen Associate Research Scientist CCA, Flatiron Institute New York, NY stonnes@gmail.com On Wed, Feb 27, 2019 at 8:44 AM Matthew Turk <matthewturk@gmail.com> wrote:
Hi Stephanie,
Thanks for this detailed report. I am thinking through how to understand this and how to process the potential issue.
On Tue, Feb 26, 2019 at 11:36 PM Stephanie Tonnesen <stonnes@gmail.com> wrote:
Hi all,
I am trying to find clumps in a few wind-tunnel runs of a galaxy with ram pressure stripping, so that I can measure the mass, volume, etc. I want to make sure, then, that I know how clump finder is selecting its contours. Unfortunately, I am not getting the minimum densities that I am expecting--although not all the time!
I have used IsolatedGalaxy and run the same clump finder routine, but with different maximum densities for the contours. To reproduce my error I had to make the step size smaller--here I am using step = 1.1.
Basically, I am not always stepping up my minimum density contour, even when the maximum density in the clump makes it clear I *should* be.
So I'm not 100% sure that I follow -- I was able to run the script you provided but I can't yet know how to diagnose because I don't fully understand the issue (my fault, not yours).
But, I will note a few things. The value for c_max is actually used in the contour finding; when the contour tree is iterating over values in a given collection of cells, it simply passes over any that are not included in the range. What this may mean is that if you have a field that has a lot of variation across a small spread of values, you might see some contours existing and others not based on minor changes in the max value supplied. For instance, two contours that are connected by a value of c_max will not be connected if that c_max is slightly lowered, and then based on the validators chosen this can change the hierarchy of clumps at lower densities (including what constitutes a leaf).
I'm not *sure* if that can account for the behavior, but I think I need to understand the problem better before I can provide additional diagnosis. Is it possible that you could illustrate this with, say, a 2D image that you draw or illustrate somehow?
I am attaching the script where I have cmax = 9.0e-24, which shows the error. You can see that in the figure I attach, looking at the green triangles. The y-axis is the maximum density in a clump, the x-axis is the minimum density. To guide your eye I have drawn lines at each step. Basically, no point should be above a black line and to the left of the connected vertical line (because then the max value is above the next possible minimum in the steps) *unless* the cmax value was reached. To make this really clear if you look at the upper-left green triangle, the max density is ~7e-24 while the minimum density is ~3.6e-24--there are a lot of steps between those values--why is the minimum density not 6.68e-24?
Strangely enough, this problem does *not* show up when I have cmax = 6.7e-24 (cyan circles).
My
I write out the max and min values in a text file at the end of the script, so I also attach the little script I use to make the figure.
Any help would be *much* appreciated!
Thank you all,
Stephanie
-- Dr. Stephanie Tonnesen Associate Research Scientist CCA, Flatiron Institute New York, NY
stonnes@gmail.com _______________________________________________ yt-users mailing list -- yt-users@python.org To unsubscribe send an email to yt-users-leave@python.org
_______________________________________________ yt-users mailing list -- yt-users@python.org To unsubscribe send an email to yt-users-leave@python.org
Hi folks, As an update, Britton and Stephanie worked through this and came up with: https://github.com/yt-project/yt/pull/2170 -Matt On Wed, Feb 27, 2019 at 11:15 AM Stephanie Tonnesen <stonnes@gmail.com> wrote:
Hi Matt,
Thanks for the reply! I have been poking at this problem for a while now, so it can be hard for me to not make some jumps in my thought process. Perhaps this will help (or perhaps not!): Here is a projection plot from the script I sent you, with cmax = 9.0e-24. See the regions I circled?
Now, I circle the same regions in the *same* script but with cmax = 6.7e-24.
When I have the LOWER (6.7e-24) cmax, the contours are smaller--> at a higher density! For some reason, the 9.0e-24 cmax value is not making all the steps that the 6.7e-24 cmax value manages.
Okay, and then I am re-attaching my figure where each point should be a leaf clump, and I have the max vs the min density of the clump. I have circles and written some stuff on it, so perhaps it makes a *bit* more sense now! (oh, and "no vals" in the legend means no validators. everything has a cmin=1.6e-24).
Let me know if I have managed to clarify anything!!
Thanks again, Stephanie
-- Dr. Stephanie Tonnesen Associate Research Scientist CCA, Flatiron Institute New York, NY
stonnes@gmail.com
On Wed, Feb 27, 2019 at 8:44 AM Matthew Turk <matthewturk@gmail.com> wrote:
Hi Stephanie,
Thanks for this detailed report. I am thinking through how to understand this and how to process the potential issue.
On Tue, Feb 26, 2019 at 11:36 PM Stephanie Tonnesen <stonnes@gmail.com> wrote:
Hi all,
I am trying to find clumps in a few wind-tunnel runs of a galaxy with ram pressure stripping, so that I can measure the mass, volume, etc. I want to make sure, then, that I know how clump finder is selecting its contours. Unfortunately, I am not getting the minimum densities that I am expecting--although not all the time!
I have used IsolatedGalaxy and run the same clump finder routine, but with different maximum densities for the contours. To reproduce my error I had to make the step size smaller--here I am using step = 1.1.
Basically, I am not always stepping up my minimum density contour, even when the maximum density in the clump makes it clear I *should* be.
So I'm not 100% sure that I follow -- I was able to run the script you provided but I can't yet know how to diagnose because I don't fully understand the issue (my fault, not yours).
But, I will note a few things. The value for c_max is actually used in the contour finding; when the contour tree is iterating over values in a given collection of cells, it simply passes over any that are not included in the range. What this may mean is that if you have a field that has a lot of variation across a small spread of values, you might see some contours existing and others not based on minor changes in the max value supplied. For instance, two contours that are connected by a value of c_max will not be connected if that c_max is slightly lowered, and then based on the validators chosen this can change the hierarchy of clumps at lower densities (including what constitutes a leaf).
I'm not *sure* if that can account for the behavior, but I think I need to understand the problem better before I can provide additional diagnosis. Is it possible that you could illustrate this with, say, a 2D image that you draw or illustrate somehow?
I am attaching the script where I have cmax = 9.0e-24, which shows the error. You can see that in the figure I attach, looking at the green triangles. The y-axis is the maximum density in a clump, the x-axis is the minimum density. To guide your eye I have drawn lines at each step. Basically, no point should be above a black line and to the left of the connected vertical line (because then the max value is above the next possible minimum in the steps) *unless* the cmax value was reached. To make this really clear if you look at the upper-left green triangle, the max density is ~7e-24 while the minimum density is ~3.6e-24--there are a lot of steps between those values--why is the minimum density not 6.68e-24?
Strangely enough, this problem does *not* show up when I have cmax = 6.7e-24 (cyan circles).
My
I write out the max and min values in a text file at the end of the script, so I also attach the little script I use to make the figure.
Any help would be *much* appreciated!
Thank you all,
Stephanie
-- Dr. Stephanie Tonnesen Associate Research Scientist CCA, Flatiron Institute New York, NY
stonnes@gmail.com _______________________________________________ yt-users mailing list -- yt-users@python.org To unsubscribe send an email to yt-users-leave@python.org
_______________________________________________ yt-users mailing list -- yt-users@python.org To unsubscribe send an email to yt-users-leave@python.org
_______________________________________________ yt-users mailing list -- yt-users@python.org To unsubscribe send an email to yt-users-leave@python.org
participants (2)
-
Matthew Turk -
Stephanie Tonnesen