# Event Horizon Telescope Data Cleanup

I was wondering how the researchers were able to minimize or neutralize confirmation bias regarding the data collected?

As I understand it, the algorithm pruned much of the data, which would make sense, since it would likely be useless noise. See for example the video Event Horizon Telescope Press Conference. First Image Of a Black Hole. at about 38:40.

Five petabytes is a lot of data… The image you saw of course isn’t 5 petabytes in size, it’s a few hundred kilobytes, so our data analysis has to collapse this five petabytes of data into an imaget that’s more than a billion times smaller.

But apart from that, could certain features be misidentified as artifacts, and be thrown out in favour of preselecting what researchers expected to see?

The only reason I ask is that the image looks almost identical to simulated predictions from two years ago - which may raise a slight cause for alarm.

Thank you.

In the first stage process - "common features" - are they like existing stamps of astronomical features (and image comparatives) that have been already observed in our universe (in both the visible and invisible spectrum and used for comparison)? And a second question - if we have never observed the earmarks of a black-hole before, how are we to use existing stamps of knowledge of astronomical features that we have seen to assess and validate an as of yet unknown (which would seem to directly impact what artifacts are preserved and what artifacts are thrown out and discarded)?

• Please see official Event Horizon Press Conference @ link cited here @ 39 minutes, 35 seconds: youtube.com/watch?v=XlrfBiuMzfEhttps://www.youtube.com/… – DAS Apr 11 at 14:42
• Correction. It would be more informative to view above video @ 38 minutes, 40 seconds. Perhaps "rejected" is the wrong word; "pruned" may be more appropriate. – DAS Apr 11 at 14:48

This process is described in great detail in this paper. The abstract says:

To assess the reliability of these results, we implemented a two-stage imaging procedure. In the first stage, four teams, each blind to the others' work, produced images of M87 using both an established method (CLEAN) and a newer technique (regularized maximum likelihood). This stage allowed us to avoid shared human bias and to assess common features among independent reconstructions.

They then generated a large number of "plausible" images, reconstructed the radio signals they would have given and ran them through the image analysis pipeline to see if it showed any biases.

In the second stage, we reconstructed synthetic data from a large survey of imaging parameters and then compared the results with the corresponding ground truth images. This stage allowed us to select parameters objectively to use when reconstructing images of M87. Across all tests in both stages, the ring diameter and asymmetry remained stable, insensitive to the choice of imaging technique.

Incidentally, it's not really appropriate to think of the rejected data as image data in the sense you would get from a camera. It is (almost) recordings of the exact electromagnetic waveforms received by the telescopes. The image arises from subtle correlations between these datasets. So most of what is thrown away (apart from noise) is information about the exact phase of the radio signal on a nanosecond-by-nanosecond basis, which is not really interesting.

• It's just hard to deny that the image was constructed, as opposed to being a raw picture. When I found out an algorithm was used to build the image, and that some human interpretation was involved as well, I was like: Hello darkness my old friend. Astrophysicists are so invested in General Relativity, many of them sadly can't be trusted to be neutral. – White Prime Apr 11 at 16:04
• There is no such thing as a "raw image", even the human eye constructs electrical impulses from the arriving light. CCDs convert arriving light into electrical signals as well and for the EHT they are converting the amplitude and phase of the radio signals recorded at different places into an image through the well established technique of interferometry. None of this requires GR to be true or not; that only comes when you want to test the image against GR's predictions for the area around a black hole. – astrosnapper Apr 11 at 18:50
• @WhitePrime You win the prize. To some extent I'm surprised it has taken more than 24 hours for the conspiracy theorists to show up. Sadly. – Rob Jeffries Apr 11 at 21:50
• @Rob Jeffries So, because I'm skeptical I'm a conspiracy theorist? That's beneath you, Rob. I actually can't stand conspiracy stuff. – White Prime Apr 11 at 23:18
• Then stop writing things like "Astrophysicists are so invested in General Relativity, many of them sadly can't be trusted to be neutral. " @WhitePrime because that is a conspiracy theory. – Rob Jeffries Apr 11 at 23:28