| Neural Network Toolbox | |
learnh
The Hebb weight learning function increases weights in proportion to the product, the weights input, and the neuron's output. This allows neurons to learn associations between their inputs and outputs.
learnhd
The Hebb-with-decay learning function is similar to the Hebb function, but adds a term that decreases weights each time step exponentially. This weight decay allows neurons to forget associations that are not reinforced regularly, and solves the problem that the Hebb function has with weights growing without bounds.
learnis
The instar weight learning function moves a neuron's weight vector towards the neuron's input vector with steps proportional to the neuron's output. This function allows neurons to learn association between input vectors and their outputs.
learnos
The outstar weight learning function acts in the opposite way as the instar learning rule. The outstar rule moves the weight vector coming from an input toward the output vector of a layer of neurons with step sizes proportional to the input value. This allows inputs to learn to recall vectors when stimulated.
| | Transfer Functions | Custom Functions | |
© 1994-2005 The MathWorks, Inc.