Project Disconvery makes no sense

So I tried Project Discovery years ago and again just now. I did the tutorial. it really made little sense but it gave me credit for whatever I did, so maybe that is part of the problem. But also because the tutorial makes no sense. The tutorial gives certain instructions, but the results are not in line with those instructions.

According to the directions we are supposed to demarcate clusters of cells. The first two examples in the tutorial was clear, even if they said it was more complex. But then the last one it says, “hint, there are five groupings” when there were clearly four clusters. I thought maybe it was splitting the stretched out cluster on the bottom left into two clusters, but no, it was calling the middle diffuse area a “cluster” even though it was clearly just a convergence of the other four clusters.

Then I do another, and another, and it is the same sort of pattern each time. There will be one or two clusters, but according to the scoring, there are two or three clusters. I group the clusters in the evident patterns and then the scoring gives me a fail and displays groupings that don’t reflect the patterns presented.

So here is an example of clearly one cluster. So in white is what I submitted. The cluster is like a comet leaving a trail behind it, but the trail is clearly related to the dense cluster. It is not related to a separate comet that is maybe off-screen. So one cluster. But the game thinks it is two, but the dividing line for making it two clusters doesn’t make sense.

If one were to make a dividing line, wouldn’t one put the demarcation where there is a clear gap (red)? If one were to divide between the nucleus and the tail, then wouldn’t it make more sense to draw the line where the nucleus begins to diffuse into a tail. (blue)
Or if one were to get real specific, draw one circle around the green cells, and the others around the blue cells. Or draw one around the red, another around the yellow, another around the green, and another around the blue and make it four.

I can think of all kinds of different ways to demarcate this group based on different criteria, but the “correct” one I just can’t see the rationalization based on the instruction “Demarcate clusters” The dividing line in orange between “clusters” seems to be very arbitrary going right through where the cells are still thick enough to be considered a stream. Maybe if there were some instruction in the tutorial about identifying subclusters or something the score might make sense, Generally speaking, however, one would think that for there to be two clusters, there would need to be two points with an increasing density of cells. If the density clearly decreases according to the distance from one point, then how can any other point be considered a new cluster? Wouldn’t there need to be another zone where the density increases?

So I I assume that these large diffuse areas are considered a “cluster” then on subsequent patterns I fail again because if I designate a large quadrant as a “cluster” then the pattern will designate that part of another closter. And even though the tutorial said to group all of the cells as one could, these golden groupings are very sloppy often leaving out thick groupings of cells.

TLDR: PD gives instructions but the answers are not in accordance with those instructions, and I’m not good at guessing what these unspoken directions are supposed to be based on the sample data. If you’ve figured it out, please explain.

2 Likes

Sometimes there are control-samples in the data sets due for analysis. This is to be expected as CCP and the Scientists want to stop anyone gaming the science. It’s all about the Science so it should be.

There are sometimes strange samples that make me sit up and wonder, and I believe there is (? or if not should be) a thread somewhere on the forum that highlight anomalous samples that seem wrong.

Keep trying at it, it’s a worthy project!

(Currently I’m only level 175 but working on it!)

Alright, I’m just going to assume it’s borked and try not to sweat all the fails. Not sure that I can progress very well since I have a hard time scoring over 40