ethankoch4
commented
6 years ago 1 - Done. Working on plotting this. 2 - I have incorporated the 'undetectable' line. Overlap I have not calculated, but it's almost the same as agreement. 3 - The above simulations were done by increasing the out_of_class_probability and using graphs with different sizes of N to simulate it approaching infinity. It also used Gibbs Sampling, Monte Carlo Markov Chains, and Belief Propagation for membership assignment. 4 - Remembering! 5 - Figure 4 is calculated by doing 20 trials with n=300 at each point and calculating the proportion of times the method succeeded at that value of alpha & beta. 6 - The failure point has been calculated in 2 ways so far using the above papers (see sbm.sbm.py critical_point & detectable_point). Still working on what we think the failure point is. 7 - Plotted!
1 - Riemann sum calculation for area under curve 2 - Figure out overlap? and how they calculate the undetectable line: (https://arxiv.org/pdf/1109.3041.pdf) 3 - Read about the above simulations 4 - Remember: (https://arxiv.org/pdf/1011.4644.pdf) 5 - Figure out what simulation is Figure 4: (https://arxiv.org/pdf/1405.3267.pdf) 6 - Read survey paper . . . where is the failure point? How do we calculate it? 7 - Plot on the graphs the theoretic thresholds.
@idc9 read the paper