Oh wow, that’s interesting Zeewolf, and useful to learn. So it’s not so important that the boxes are tight, but that they are capture everything that qualifies (inclusion of all associated cells is more important that exclusion of empty areas).
Even still, many times I’ll see cells that are so scattered away from the center of gravity that I don’t include them, but I can’t help myself, it’s like “these just don’t seem to be part of that cluster”.
The other ones I have a hard time with are the ones where there’s a few tight clusters on the screen and then a perfectly even distribution of cells among/between all the clusters. It’s like, are those really part of any of the clusters? Are they a cluster of their own (even though there’s really no obvious center of gravity)? Or are they just “background radiation”, metaphorically speaking, representing cells that really do not represent a cluster across the two particular dimensions of the slide…how do I say this…as if their relationship between dimension A is orthogonal to dimension B. Do I gate them in at the risk of creating a false signal, or do not include them in any gates at the risk of creating a false negative?
Like 90% of these aren’t even ‘verified’ so I can just draw a box anywhere and click Submit.
The other 10%…I just draw a boxes in ‘this area’ and…tad-da 80% accurate.
BTW: I’m now at 98.2% Accuracy Rating, and still:
a) Have no clue what I’m doing
b) Wonder why this isn’t just scripted by a bot-maker.
((Like seriously, I’m using the exact same ‘logic’ on every single one that Passes; see two ‘clumps’ draw two boxes so that ~50% of the slide is in one box, and ~50% is in the other; 51% or better accuracy…you PASSED!))
It’s just tedious because I literally see no point in a human doing this. A program could easily replicate what I’m doing. And could process millions of slides in the span it takes me to do one; about 10 seconds (when I actually try). - This is suppose to be to save lives after all.
Yup, to the point that Project Discovery were forced to Neocom bar in the last patch. Bring the exoplanets! Hack the planets! Hack the planets! Hack the planets!
dude its hard to a computer to know if your box is “like” the gold standard
for that you will need a neural network ai
the game probably just run a simple area math to compare and give you the grade , thats why the big boxes that contain all the standard polygon is getting 100%
btw project discovery was much more relevant a few years ago wen machine learning was not advanced as nowadays , there is things that are very simple to a human and veeery hard to a machine
the classic example was is this a tea pot?
but now machines are learning to do this difficult things
if we are really doing something , my theory is, , we are narrowing the field to them pick our results , pass them to a neural network machine , them pass to human scientists
us < machine < real scientist
So, realistically, this isn’t doing a bit of good.
CCP should have just offered the rewards on the NEStore. With a portion of the proceeds going to the WHO, since this whole thing is about finding out how our immune system and pathogens interact.
I mean…for a game that preaches ‘The Butterfly Effect’. They sure aren’t making use of it…
“Gaming company, CCP, raises over $1 million USD to help combat infectious diseases.”
((Gaming companies: Holy crap! Look at the response; the free publicity that generated!"
“Several gaming companies have raised a combined $million.”
etc, etc.
Instead they opt to do this, where making a bigger box gets you better accuracy, on the verified slides. Which means, we’re just doing ‘big boxes’ on the unverified ones; we’re gaming the system with out even realizing it, because chances are my boxes look a lot like your boxes on the unverified.
((Spoilers: They don’t, because I can , with a 90%+ accuracy rate, guess which ones aren’t going to be verified. So I just draw a single box and continue. Which just goes to show how pointless this was, and should have been scraped for the above alternative of just buying the items and giving a portion to international organizations that deal with studying infectious diseases.))
The main reason that keeps people away from this one is that this PD is harder than the last one. And not because the problem itself is harder - it’s actually the opposite. But the UI requires us to do lot of unnecessary work so we spend most of our time in doing furious mouse clicking.
I am trying to complete these but I get errors saying Submission Sample Mishap. Every single time I try. I’ve log out and in, I have an other account that is alpha that can do it fine. But this one, which is new it wont work.
Where can I find help for this? I can’t find anything about it online.
"We cannot produce more than five submissions a minute, nor should we. Make sure you’re taking time to carefully analyze each sample. A prolific researcher is a good researcher, but only if they produce quality work. "
Remove this. I have a 99% average accuracy and never score below 80% accuracy on the ‘verified’ slides.
I capped out at lvl 75 today. My SP is 3.5m. The limitation looks like 25 lvls per 1m SP.
It’s disappointing CCP have used SP as such a drastic limitation. I have an Omega account like everyone else, this should have more weighting to progress further IMO.
I hear CCP peeps on twitch saying things like making the wider game available for all etc etc. But then they nerf. someone who wants pursue this avenue.
I get nerfing Alpha accounts, but the grinding should be the natural nurf to higher lvls for Omega accounts not SP. Once again CCP reinforces stronger players getting more, which results in even stronger, while weaker player aren’t rewarded for the grind. This is completely the opposite of what was mentioned on twitch last night.