We explain Poisson learning on graph-based semi-supervised learning to see if it could avoid the problem of global information loss problem as Laplace-based learning methods on large graphs. From our analysis, Poisson learning is simply Laplace regularization with thresholding, cannot overcome the problem.
Paper
References (12)
07Laplace regularization with label centralization becomes nearest class mean classifier
08What is the meaning of the offset c?
09What is the meaning of the offset c ? It depends on how the weights are constructed. One way to interpret is that P i sign ( u i ) d i = 0 would be giving the two class an equal volume
10What is Poisson learning in the RKHS? It becomes nearest class mean classifier with an offset (or Laplace regularization with an offset)
11What is the advantage of Poisson learning? Offset c , which acts as a threshold for classificationCan it improves AUC scores on Laplace regularization?
12Can Poisson learning avoid the global information loss problem on large graph? The answer is NO due to L † representation