* M24 erroneously claims that the EHTC methodology does not account for the point-spread-function (PSF) of the observation. In fact, the EHTC hybrid imaging approach explicitly corrects for the PSF through the CLEAN algorithm. Other algorithms implicitly incorporate PSF effects into the reconstruction algorithms. Furthermore, non-imaging (e.g., model fitting) methods confirm the ring-like structure is the most likely simple model that fits the data. Note that the EHTC methods recover non-ring structures in both synthetic data tests as well as real data that has the same PSF, for example when imaging the non-ring-like polarization of the black holes M87* and Sgr A* (10).
* M24 erroneously claims that the EHTC did not release uncalibrated raw data. In fact, all of 2017 EHTC raw and calibrated data have been publicly available since May 2022 (11).
"In our view, the ring-like image found by the EHTC is not the intrinsic structure of Sgr A* but arises from the sparse u-v coverage of the EHT array in 2017; that is, from the corresponding [...] structure in the PSF."
I.e. authors assert the image processing path inserted rather than revealed the famous correlation with Interstellar black hole imagery.
I hesitate to comment on something I know so little about, but that has never stopped me before.
The EHTC approach would appear impossible to calibrate correctly, My understanding(the thousand foot view really) is they calibrated a generative model on what they expect to see based on their expected inputs, fed in the actual inputs and got what they trained their model on. great success, good work everyone.
I guess my unease with this approach is it feels there are too many assumptions baked in.
It's not so biased as that: the process makes some assumptions (mainly, that the image structure is relatively simple) but it is agnostic to whether it's looking at a black hole or some other image (it's not like an image generation AI where they fed it in a bunch of existing images of what they thought black holes might look like). The techniques they used are used in other areas of astronomy.
The criticism in this paper seems to be focused on how one of those assumptions could lead to the ring structure, not because they biased it with presupposed images, but due to the mathematics of how the techniques attempt to determine the image from the raw data. It makes some sense to me that this could happen, but I'm not really equipped to evaluate who seems to have the better points.
Way way beyond my comprehension but I will marvel at how peer-review works in science and wonder how much further the world would have progressed if everything worked like that.
It's a bit dense for sure, lost me a few paragraphs in but this sentence got the point across well enough:
> It is not based on consistency with the observed data, but rather on the highest rate of appearance of the image morphology from a wide imaging parameter space.
I think the claim here is the original image effectively optimized for "cool shape" instead of accuracy.
I also had to jump halfway down the page to find the images, it was driving me crazy to read about their differences without any idea what they looked like.
I think their main point is that the EHTC are wrong, and they don't care much about how they prove it, which is why when the serious people at the EHTC replied it reads like this:
* M24 does not address the intrinsic variability of Sgr A* in their imaging process.
* M24 presents one single image as the end result of their analysis, disregarding the full ensemble of suitable images.
* M24 biases their results towards a point-source image through self-calibration.
* M24 misstates the biases in their own methodology as demonstrations of biases in the EHTC methodologies.
Yeah to be clear I only meant to summarize their position as I read it from the paper, I don't know enough to make a judgement on who is right. Appreciate the summary of counterpoints though, very helpful for my understanding!
my favourite abstract was one where the title asked "did cern measure a neutrino travelling faster than the speed of light?" and the abstract was "no".
for some reason I can't seem to find the paper anymore, it was on arxiv.
The authors have feuded with the EHT collaboration for quite a while. Here is a previous reply by the collaboration: https://eventhorizontelescope.org/blog/imaging-reanalyses-eh...
This time they put more effort into it: https://eventhorizontelescope.org/blog/response-independent-...
Some highlights:
* M24 erroneously claims that the EHTC methodology does not account for the point-spread-function (PSF) of the observation. In fact, the EHTC hybrid imaging approach explicitly corrects for the PSF through the CLEAN algorithm. Other algorithms implicitly incorporate PSF effects into the reconstruction algorithms. Furthermore, non-imaging (e.g., model fitting) methods confirm the ring-like structure is the most likely simple model that fits the data. Note that the EHTC methods recover non-ring structures in both synthetic data tests as well as real data that has the same PSF, for example when imaging the non-ring-like polarization of the black holes M87* and Sgr A* (10).
* M24 erroneously claims that the EHTC did not release uncalibrated raw data. In fact, all of 2017 EHTC raw and calibrated data have been publicly available since May 2022 (11).
An independent hybrid imaging of Sgr A* from the data in EHT 2017 observations
"In our view, the ring-like image found by the EHTC is not the intrinsic structure of Sgr A* but arises from the sparse u-v coverage of the EHT array in 2017; that is, from the corresponding [...] structure in the PSF."
I.e. authors assert the image processing path inserted rather than revealed the famous correlation with Interstellar black hole imagery.
I hesitate to comment on something I know so little about, but that has never stopped me before.
The EHTC approach would appear impossible to calibrate correctly, My understanding(the thousand foot view really) is they calibrated a generative model on what they expect to see based on their expected inputs, fed in the actual inputs and got what they trained their model on. great success, good work everyone.
I guess my unease with this approach is it feels there are too many assumptions baked in.
It's not so biased as that: the process makes some assumptions (mainly, that the image structure is relatively simple) but it is agnostic to whether it's looking at a black hole or some other image (it's not like an image generation AI where they fed it in a bunch of existing images of what they thought black holes might look like). The techniques they used are used in other areas of astronomy.
The criticism in this paper seems to be focused on how one of those assumptions could lead to the ring structure, not because they biased it with presupposed images, but due to the mathematics of how the techniques attempt to determine the image from the raw data. It makes some sense to me that this could happen, but I'm not really equipped to evaluate who seems to have the better points.
Way way beyond my comprehension but I will marvel at how peer-review works in science and wonder how much further the world would have progressed if everything worked like that.
It's a bit dense for sure, lost me a few paragraphs in but this sentence got the point across well enough:
> It is not based on consistency with the observed data, but rather on the highest rate of appearance of the image morphology from a wide imaging parameter space.
I think the claim here is the original image effectively optimized for "cool shape" instead of accuracy.
I also had to jump halfway down the page to find the images, it was driving me crazy to read about their differences without any idea what they looked like.
I think their main point is that the EHTC are wrong, and they don't care much about how they prove it, which is why when the serious people at the EHTC replied it reads like this:
* M24 does not address the intrinsic variability of Sgr A* in their imaging process.
* M24 presents one single image as the end result of their analysis, disregarding the full ensemble of suitable images.
* M24 biases their results towards a point-source image through self-calibration.
* M24 misstates the biases in their own methodology as demonstrations of biases in the EHTC methodologies.
https://eventhorizontelescope.org/blog/response-independent-...
Yeah to be clear I only meant to summarize their position as I read it from the paper, I don't know enough to make a judgement on who is right. Appreciate the summary of counterpoints though, very helpful for my understanding!
my favourite abstract was one where the title asked "did cern measure a neutrino travelling faster than the speed of light?" and the abstract was "no".
for some reason I can't seem to find the paper anymore, it was on arxiv.
I like this one.
The unsuccessful self-treatment of a case of “writer's block”: https://pmc.ncbi.nlm.nih.gov/articles/PMC1311997/
It's probably this paper: https://doi.org/10.1088/1751-8113/44/49/492001
oh that might be it, two words, they went all out in the final version!
edit: no, I must have embellished it in my mind: https://arxiv.org/abs/1110.2832