Here, we vary the noise by adding a nr like this:
noisy_images[i] = np.random.poisson(images[c]*number)
So, multiplying with 1 should give just the images without any deeper level of noise. Then, the farther away from 1 that you go, and the closer to 0 you are, the more noise it is and the harder it should be for the MDS method to separate the images. And about the numbers > 1, should they also result in more noise or is the "contrast" increased so that the images end up on the same point in the scatter point? With noise=*100 the images end up on almost the same place.
This is n=50 and noise= *0.001. Now, things start to mix
https://github.com/Annanilsson-code/ProjectAppliedMolBiophys/blob/18ada5ef01430992db589106215e3d68db14b45b/cc_calculation.py#L60
Here, we vary the noise by adding a nr like this: noisy_images[i] = np.random.poisson(images[c]*number)
So, multiplying with 1 should give just the images without any deeper level of noise. Then, the farther away from 1 that you go, and the closer to 0 you are, the more noise it is and the harder it should be for the MDS method to separate the images. And about the numbers > 1, should they also result in more noise or is the "contrast" increased so that the images end up on the same point in the scatter point? With noise=*100 the images end up on almost the same place.
This is n=50 and noise= *0.001. Now, things start to mix