A Walk Through Neural Network - By Heifer Jeffer

2
Introduction Neural network is a computer system modelled on the human brain and nervous system in order to learn the machine how to behave independently like human Example Let’s assume that S & U are two sets live in world P , Qs,As,Qu & Au are respectively the inputs of S & U. b is a bias ,W is weight of (Qs,As,Qu & Au) , L = [W1 Qj + W2 Aj + W3 b Let’s Design an algorithm to learn our machine how to isolate (classify) the two sets (S, U) of world P Definition S = As , Qs U =Au ,Bu Qj =Qs + Qu Aj = As + Au P = Aj , Qj L = [W1 Qj + W2 Aj + W3 b] 2013 Stanford University Venture Lab Heider Jeffer PhD Student Operation Research [A WALK THROUGH NEURAL NETWORK] The Paper starts immediately by Solving a problem in order to explain the meaning of Neural Network Web :http://goo.gl/FhpHP Blog:http://goo.gl/9xpqj E-Mail:[email protected]

description

A Walk Through Neural Network - By Heifer Jeffer

Transcript of A Walk Through Neural Network - By Heifer Jeffer

Page 1: A Walk Through Neural Network - By Heifer Jeffer

Introduction

Neural network is a computer system modelled on the human brain and nervous system in order to learn the machine how to behave independently like human

Example

Let’s assume that S & U are two sets live in world P , Qs,As,Qu & Au are respectively the inputs of S & U.

b is a bias ,W is weight of (Qs,As,Qu & Au) , L = [W1 Qj + W2 Aj + W3 b

Let’s Design an algorithm to learn our machine how to isolate (classify) the two sets (S, U) of world P

Definition

S = As , Qs

U =Au ,Bu

Qj =Qs + Qu

Aj = As + Au

P = Aj , Qj

L = [W1 Qj + W2 Aj + W3 b]

2013

Stanford University Venture Lab Heider Jeffer PhD Student Operation Research

[A WALK THROUGH NEURAL NETWORK] The Paper starts immediately by Solving a problem in order to explain the meaning of Neural Network

Web :http://goo.gl/FhpHP

Blog:http://goo.gl/9xpqj

E-Mail:[email protected]

Page 2: A Walk Through Neural Network - By Heifer Jeffer

Solution

[Au ,Bu ≤ o.5 &

As,Bu ≥ 0.5]

Or

[Au ,Bu ≥ o.5 &

As,Bu ≤ 0.5]

Out (P=Aj,Qj) ← sigmoid [W1 Qj + W2 Aj + W3 b]

YES

End

No

Delta Rule D W1 = dL(Aj,Qj) / d W1 D W2 = dL(Aj,Qj) / d W2 D W3 = O because Bias is a constant

W1 = D W1 + W1

W2 = D W2 + W2

W3 = W3

Start