# Fine-tuning Stacked AEs

 Revision as of 03:22, 12 May 2011 (view source)Jngiam (Talk | contribs) (→Finetuning with Backpropagation)← Older edit Revision as of 11:04, 26 May 2011 (view source)Watsuen (Talk | contribs) Newer edit → Line 36: Line 36: Note: While one could consider the softmax classifier as an additional layer, the derivation above does not. Specifically, we consider the "last layer" of the network to be the features that goes into the softmax classifier. Therefore, the derivatives (in Step 2) are computed using $\delta^{(n_l)} = - (\nabla_{a^{n_l}}J) \bullet f'(z^{(n_l)})$, where  $\nabla J = \theta^T(I-P)$. Note: While one could consider the softmax classifier as an additional layer, the derivation above does not. Specifically, we consider the "last layer" of the network to be the features that goes into the softmax classifier. Therefore, the derivatives (in Step 2) are computed using $\delta^{(n_l)} = - (\nabla_{a^{n_l}}J) \bullet f'(z^{(n_l)})$, where  $\nabla J = \theta^T(I-P)$. }} }} + + + {{CNN}}