Depending on the initial condition, gradient descent may end up at different local optima.


Doubly stochastic normalization

STOCHASTIC: Analyzed statistically but not precisely predicted

Spectrum analysis of affinity matrix : Pairwise affinity matrix to measure similarity, then NORMALIZE,

Logistic regression: Classification into discrete groups.
Set a threshold

Prediction is given by sigmoid function g(
Cost: need one minimum

Comments

Popular posts from this blog

ADVERTISING STYLES