Some Idea from Ng Video

by allenlu2007

 

1. Gradient descent formula for linear regression and logistic regression are identifcal

except h_theata definition is different.   Why the gradient formula are the same?  

 

2. Why go neural net?  XOR –> perception or logistic regression can not classify.  But neural net can.

So the key is to do island classification?

=> machine learning and topology relationship????

 

Very interesting topic:  neural network, topology and manifold.

http://colah.github.io/posts/2014-03-NN-Manifolds-Topology/

Advertisements