Introduction to Neural Network Session 1 Course: T0293 – NEURO COMPUTING Year: 2013
Neuro Computing Reference Book : Haykin, S. (2009). Neural Networks and Learning Machines, 3rd ed. Pearson. ISBN: Lecturer: Dr. Widodo Budiharto HP: quiz TM in group 1 final projects T Neuro Computing2
Learning Outcomes At the end of this session, student will be able to: Explain and discuss the importance of Neural Network (LO1) 3T Neuro Computing
Lecture Outline What is a Neural Network? Neural Network Development Benefits of Neural Network The Human Brain Models of A Neuron Activation Function The future of Neural Network Artificial Intellect with Neural Network T Neuro Computing4
What is a Neural Network? A type of artificial intelligence that attempts to imitate the way a human brain works. Neural network performs useful computations through a process of learning. In other words, a neural network can be viewed as an adaptive machine. T Neuro Computing5
Machine Learning Involves adaptive mechanisms that enable computer to learn from experience. T Neuro Computing6
Brain T Neuro Computing7
System syaraf Sistem saraf tersusun atas miliaran sel yang sangat khusus yang disebut sel saraf (neuron). Dendrit berfungsi mengirimkan impuls ke badan sel saraf, sedangkan akson berfungsi mengirimkan impuls dari badan sel ke jaringan lain. Akson biasanya sangat panjang. Sebaliknya, dendrit pendek. Sebuah sinapsis menyediakan koneksi antara neuron yang memungkinkan informasi sensorik mengalir di antara mereka. T Neuro Computing8
Neural Network Development The 1940s: The Beginning of Neural Nets –McCulloch-Pitts Neurons Warren McCulloch & Walter Pitts (1943) The Neuron computes the weighted sum of the input signals and compares the result with a threshold value, Ө. Where X is the net weighted input to the neuron, Xi is the value of input I, wi is the weight of input i, n is the number of neuron inputs and Y is the output of neuron. T Neuro Computing9
Diagram of a neuron T Neuro Computing10
The 1950s & 1960s: The First Golden Age of Neural Networks –Perceptrons Group of researchers (Black, 1962; Minsky & Papert, 1969; Rosenblatt, 1958, 59, 62) Frank Rosenblatt introduced a training algorithm that provided the first procedure for training a simple ANN: a perceptron. T Neuro Computing11
Exercise Create a program to calculate the output Neuron Y With 2 data input {0.3,0.7} weight=0.2 and theta =0.6 T Neuro Computing12
answer double x[2]={0.3,0.7}; double w = 0.6; double X=0; double tetha= 0.5; for(int i=0; i<2; i++) { X+=(x[i]*w); } if(X>=tetha){ printf("output +1"); }else { printf("output -1"); } T Neuro Computing13
Multi layer neural network T Neuro Computing14
Application : Character Recognition T Neuro Computing15
Implementation Pattern Recognition Prediction Intelligent Robotics Classification T Neuro Computing16
The Human Brain Processing The characteristic of a human brain are: –Highly Complex –Non-Linear –Parallel Computer (information-processing system) To able to imitate the human brain, neural network system needs “neurons” or “processing units”. The neural network resembles the brain in two respects: 1.Knowledge is acquired by the network from its environment through a learning process. 2.Interneuron connection strengths, known as synaptic weights, are used to store the acquired knowledge. T Neuro Computing17
Models of A Neuron A neural network system contains many processing elements called neurons. There are three basic elements of a neural model: 1.A set of synapses, or connecting links, each of which is characterized by a weight or strength of its own. 2.An adder for summing the input signals, weighted by the respective synaptic strengths of the neuron. 3.An activation function for limiting the amplitude of the output of a neuron. T Neuro Computing18 Figure 1.1 Nonlinear model of a neuron.
Case study : Robotics (Obstacles Avoidance) T Neuro Computing19
Activation Function Activation function defines the output of a neuron. Generally, there are two basic types of activation functions: 1.Threshold Function In engineering, this form of a threshold function is commonly referred to as a Heaviside function. T Neuro Computing20 Figure 1.2 Threshold function. Figure 1.3 Graph of the threshold function.
2.Sigmoid Function The most common form of activation function used in the construction of neural networks. Strictly increasing function that exhibits a graceful balance between linear and nonlinear behavior. T Neuro Computing21 Figure 1.5 Graph of the sigmoid function. Figure 1.4 Sigmoid function for varying slope parameter a.
The Future of Neural Network –Neural Networks Model Multilayer Perceptrons (Backpropagation) Principal-Component Analysis (PCA) Self-Organizing Network Model (SOM) etc. –Application Pattern Recognition Image Compression Optimization etc. T Neuro Computing22
Artificial Intelligent with Neural Networks Advance Robotics Intelligent Control Technical Diagnostics Intelligent Data Analysis and Signal Processing Intelligent Expert Systems Intelligent Security Systems Intelligent Medicine Devices Image and Pattern Recognition Machine Vision AI with Neural Network
Summary A neural network can be viewed as an adaptive machine. To able to imitate the human brain, neural network system needs “neurons” or “processing units”. A neural network system contains many processing elements called neurons. Activation function defines the output of a neuron. T Neuro Computing24
Application Neural network in Computer Vision for Manipulator T Neuro Computing25
Vision based manipulator T Neuro Computing26
Create a program using C++/C# for Neural Network Presentation 10 minutes in session 13. Example projects : 1.Prediction /Classification of system 2.Face Recognition 3.Car Plate Recognition 4.Object Detection and Tracking 5.Robotics obstacle avoidance, etc T Neuro Computing27
References Textbook Haykin, S. (2009). Neural Networks and Learning Machines, 3rd ed. Pearson. ISBN: Web T Neuro Computing28
END T Neuro Computing29