Understanding Artificial Neural Networks for Computational Applications

1. Introduction

2. What is an ANN? • It is a computational System Inspired by Structure, Processing method, Learning ability of Biological Brain • The term “neural network” is also applied to models of the brain

3. Why ANN? • Some tasks can be done easily (effortlessly) by humans but are hard by conventional paradigms on Von Neumann machine with algorithmic approach • Computers can perform many operations considerably faster than a human being

4. Why ANN? • Massive Parallelism • Distributed representation • Learning ability • Generalization ablity • Fault tolerance

5. Characteristics of ANN • A large number of very simple processing neuron-like processing elements • A large number of weighted connections between the elements • Distributed representation of knowledge over the connections • Knowledge is acquired by network through a learning process

6. Biological Neuron • Dendrit , bertugasmenerimainformasi • Soma, tempatpengolahaninformasi • Axon, mengiriminpuls-inpulskeselsyaraflainya • Synapse, penghubungantara 2 neuron

7. Biological vs Artificial

8. Artificial Neuron w1 SUM Activation Function Σ p2 F(y) n=Σpi.wi w2 a=f(n) . . . wi Weight

9. Neuron vs Node

10. Topology

11. Learning • Learn the connection weights from a set of training examples • Different network architectures required different learning algorithms

12. Supervised Learing • The network is provided with a correct answer (output) for every input pattern • Weights are determined to allow the network to produce answers as close as possible to the known correct answers • The back-propagation algorithm belongs into this category

13. Unsupervised Learning • Does not require a correct answer associated with each input pattern in the training set • Explores the underlying structure in the data, or correlations between patterns in the data, and organizes patterns into categories from these correlations • The Kohonen algorithm belongs into this category

14. Applications • Pattern Classification • Clustering/Categorization • Function approximation • Prediction/Forecasting • Optimization • Content-addressable Memory • Control

15. Sejarah • Model JST formal pertamadiperkenalkanoleh McCulloch dan Pitts (1943) • 1949, HebbmengusulkanjaringanHebb • 1958, Rosenblatt mengembangkanperceptronuntukklasifikasipola • 1960, Widrowdan Hoff mengembangkan ADALINE denganaturanpembelajaran Least Mean Square (LMS) • 1974, Werbosmemperkenalkanalgoritmabackpropagationuntukperceptronbanyaklapisan

16. Sejarah • 1975, Kunihiko Fukushima mengembangkan JST khususpengenalankarakter, disebutcognitron, namungagalmengenaliposisiataurotasikarakter yang terdistorsi • 1982, Kohonenmengembangkan learning unsupervised untukpemetaan • 1982, Grossbergdan Carpenter mengembangkan Adaptive Resonance Theory (ART, ART2, ART3) • 1982, Hopfield mengembangkanjaringan Hopfield untukoptimasi

17. Sejarah • 1983, perbaikancognitron (1975) denganneocognitron • 1985, Algoritma Boltzmann untukjaringansyarafprobabilistik • 1987, dikembangkan BAM (Bidirectional Associative Memory) • 1988, dikembangkan Radial Basis Function

18. Thank’s Any Questions ?