SUPERVISED LEARNING

SUPERVISED LEARNING

1. PERCEPTRON

   • TASK : For classification of linearly separable patterns.
   • Goals: To classify each pattern to a particular class
   • LIMITATION OF PERCEPTRON:
     • Only capable of solving linearly separable problems
     • only have 2 layers
     • only 1 processing steps
     • Assumptions made are weak

2. MLP

   • -a feedforward ANN that generates a set of outputs froma set of inputs.
   • 1) what is BP in training method? -Backpropagation net,a multilayer,feedforward net trained by backpropagation algorithms
   • 2)DID BP is MLP? Yes,it is also referred to as MLP
   • 3. Architecture
     • a) Structure
     • b) Weight Matrices
       • refer the picture above: A-stores connection weights between input & hidden layer.
       • B-stores connection weight between hidden & output layers.
       • c-stores connection weights between hidden layer 1 & hidden layer 2
     • c)Signals
       • Red arrow : Function signal: is calculated as function of the input &weights of units.
         • Blue arrow ( Error signal) : Its computation involves an error-dependent function,

Bilder und Dateien der Mindmap