Fundamental ARTIFICIAL NEURAL NETWORK Session 1st TOPICS & ILLUSTRATION Azhari, Dr @ 2010 Computer Science, UGM

Human Neural a b

a b d c

Neuron in simple 2D image b

Modelling Computation of a Neuron b

How to model a neuron to recognize of “OR” Example 1 recognize of “OR” How to model a neuron to recognize of “OR” OR ---- FALSE FALSE --> FALSE TRUE FALSE --> TRUE FALSE TRUE --> TRUE TRUE TRUE --> TRUE

   Example 1 OR Real world Neuron model 1 OR 1 1 X1 X2 Z w1 w2 How to model a neuron to recognize of “OR” 1 OR 1 1 Real world   X1 X2 Z OR w1 w2  Neuron model w1, w2 harus dicari berdasarkan nilai w1, w2 awal yang diberikan

illustration to train of “OR” neuron

Test: artificial neuron “OR” model TRUE OR TRUE TRUE

Use others initial weights, threshold, learning rate TRAIN again Use others initial weights, threshold, learning rate Weights result of training

Domain result of weights for “OR” neuron

   Example 2 AND Real world Neuron model 1 AND 1 X1 w1 Z w2 X2 How to model a neuron to recognize of “AND” 1 AND 1 Real world   X1 X2 Z AND w1 w2  Neuron model w1, w2 harus dicari berdasarkan nilai w1, w2 awal yang diberikan

Result for stable weight Train of “AND” neuron Result for stable weight

How to model a neuron to recognize of “XOR” Example 3 recognize of “XOR” How to model a neuron to recognize of “XOR” XOR --- FALSE FALSE --> FALSE TRUE FALSE --> TRUE FALSE TRUE --> TRUE TRUE TRUE --> FALSE

“XOR” by Multilayer neural netwok In order to solve XOR, we must create a multi-layered Perceptron with two inputs, a “hidden” layer, and an output. This hidden layer amounts to having multiple perceptrons attack the same problem, i.e. if OR is true (perceptron #1) and AND is not true (perceptron #2), then fire the XOR output.

“XOR” by Multilayer neural netwok

“XOR” by Multilayer neural netwok

“XOR” by Multilayer neural netwok One solution for training a multi-layered neural network is known as backpropogation. Certainly, there are other options for training a neural network.

ANN model how to Recognize Apple

The goals of neural network computation To understand how the brain actually works Its very big and very complicated and made of yukky stuff that dies when you poke it around To understand a new style of computation Inspired by neurons and their adaptive connections Very different style from sequential computation should be good for things that brains are good at (e.g. vision) Should be bad for things that brains are bad at (e.g. 23 x 71) To solve practical problems by using novel learning algorithms Learning algorithms can be very useful even if they have nothing to do with how the brain works

Simple Programming for ANN Backpropagation ANN Topics Architecture ANN Activation Functions Learning Methods Simple Programming for ANN Backpropagation