data_mining:neural_network:types

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
data_mining:neural_network:types [2017/08/13 11:50] – [Symmetrically connected NN] phreazerdata_mining:neural_network:types [2019/10/26 12:09] (current) – ↷ Links adapted because of a move operation phreazer
Line 1: Line 1:
 ====== Types of NNs ====== ====== Types of NNs ======
 +===== Perceptron ===== 
 +See [[data_mining:neural_network:perceptron|Perceptron]]
 ===== Feed-forward NN ===== ===== Feed-forward NN =====
 First layer is input, last layer output, hidden layers inbetween. First layer is input, last layer output, hidden layers inbetween.
Line 13: Line 14:
   * Harder to train   * Harder to train
  
-Natural for modelling sequential data:+Natural for modeling sequential data:
   * Equivalent to very deep nets with one hidden layer per time slice, except that they use same weights at every time sclice and get input at very time slice.   * Equivalent to very deep nets with one hidden layer per time slice, except that they use same weights at every time sclice and get input at very time slice.
   * Can rember info in the hidden state for a long time.   * Can rember info in the hidden state for a long time.
  
 +See [[data_mining:neural_network:sequences:sequence_learning|Sequence learning]]
 ===== Symmetrically connected NN ===== ===== Symmetrically connected NN =====
  
Line 24: Line 26:
   * Restricted: Cannot model cycles    * Restricted: Cannot model cycles 
   * "Hopfield nets" if they have no hidden layer.   * "Hopfield nets" if they have no hidden layer.
 +
 +===== Convolutional NN =====
 +See [[data_mining:neural_network:cnn:cnn|Convolutional neural network]]
  • data_mining/neural_network/types.1502617851.txt.gz
  • Last modified: 2017/08/13 11:50
  • by phreazer