CSC Neural Networks. Perceptron Learning Rule

Similar documents
Ch4: Perceptron Learning Rule

EEE 241: Linear Systems

Artificial Neural Networks. Historical description

Revision: Neural Network

Single layer NN. Neuron Model

Perceptron. (c) Marcin Sydow. Summary. Perceptron

Solved Problems. (a) (b) (c) Figure P4.1 Simple Classification Problems First we draw a line between each set of dark and light data points.

Lecture 4: Feed Forward Neural Networks

Νεςπο-Ασαυήρ Υπολογιστική Neuro-Fuzzy Computing

Artificial Neural Networks The Introduction

Simple Neural Nets For Pattern Classification

Artificial Neural Networks. Part 2

Learning From Data Lecture 2 The Perceptron

Computational Intelligence Lecture 3: Simple Neural Networks for Pattern Classification

Vote. Vote on timing for night section: Option 1 (what we have now) Option 2. Lecture, 6:10-7:50 25 minute dinner break Tutorial, 8:15-9

Artificial Neural Network

CSC321 Lecture 4 The Perceptron Algorithm

Networks of McCulloch-Pitts Neurons

In the Name of God. Lecture 11: Single Layer Perceptrons

ADALINE for Pattern Classification

CS 4700: Foundations of Artificial Intelligence

COMP9444 Neural Networks and Deep Learning 2. Perceptrons. COMP9444 c Alan Blair, 2017

ARTIFICIAL NEURAL NETWORK PART I HANIEH BORHANAZAD

Course 10. Kernel methods. Classical and deep neural networks.

Lecture 7 Artificial neural networks: Supervised learning

PMR5406 Redes Neurais e Lógica Fuzzy Aula 3 Single Layer Percetron

(Feed-Forward) Neural Networks Dr. Hajira Jabeen, Prof. Jens Lehmann

Linear models: the perceptron and closest centroid algorithms. D = {(x i,y i )} n i=1. x i 2 R d 9/3/13. Preliminaries. Chapter 1, 7.

ARTIFICIAL INTELLIGENCE. Artificial Neural Networks

Linear models and the perceptron algorithm

Neural Networks Based on Competition

The Perceptron. Volker Tresp Summer 2016

Introduction To Artificial Neural Networks

The perceptron learning algorithm is one of the first procedures proposed for learning in neural network models and is mostly credited to Rosenblatt.

Artificial Neural Networks

ECE521 Lectures 9 Fully Connected Neural Networks

Neural Networks. Fundamentals Framework for distributed processing Network topologies Training of ANN s Notation Perceptron Back Propagation

Data Mining Part 5. Prediction

Chapter 9: The Perceptron

Artificial Intelligence

Reification of Boolean Logic

3.4 Linear Least-Squares Filter

ARTIFICIAL NEURAL NETWORKS گروه مطالعاتي 17 بهار 92

Preliminaries. Definition: The Euclidean dot product between two vectors is the expression. i=1

An artificial neural networks (ANNs) model is a functional abstraction of the

Chapter ML:VI. VI. Neural Networks. Perceptron Learning Gradient Descent Multilayer Perceptron Radial Basis Functions

Artificial neural networks

COGS Q250 Fall Homework 7: Learning in Neural Networks Due: 9:00am, Friday 2nd November.

Inf2b Learning and Data

Part 8: Neural Networks

Supervised Learning. George Konidaris

Neural networks. Chapter 19, Sections 1 5 1

Introduction and Perceptron Learning

COMS 4771 Introduction to Machine Learning. Nakul Verma

Machine Learning. Neural Networks. (slides from Domingos, Pardo, others)

Machine Learning. Neural Networks. (slides from Domingos, Pardo, others)

Neural Networks Introduction CIS 32

Learning and Memory in Neural Networks

Introduction to Artificial Neural Networks

The Perceptron. Volker Tresp Summer 2014

Lecture 5 Neural models for NLP

A MULTILAYER EXTENSION OF THE SIMILARITY NEURAL NETWORK

Neural Networks for Machine Learning. Lecture 2a An overview of the main types of neural network architecture

Neural Networks Lecture 2:Single Layer Classifiers

Last updated: Oct 22, 2012 LINEAR CLASSIFIERS. J. Elder CSE 4404/5327 Introduction to Machine Learning and Pattern Recognition

CS534 Machine Learning - Spring Final Exam

The Perceptron. Volker Tresp Summer 2018

Apprentissage, réseaux de neurones et modèles graphiques (RCP209) Neural Networks and Deep Learning

Supervised learning in single-stage feedforward networks

Intelligent Systems Discriminative Learning, Neural Networks

Instruction Sheet for SOFT COMPUTING LABORATORY (EE 753/1)

Final Examination CS 540-2: Introduction to Artificial Intelligence

Neural networks. Chapter 20. Chapter 20 1

Learning with multiple models. Boosting.

Be able to define the following terms and answer basic questions about them:

Artificial Neural Network and Fuzzy Logic

Linear discriminant functions

Multi-Dimensional Neural Networks: Unified Theory

Classification with Perceptrons. Reading:

Last update: October 26, Neural networks. CMSC 421: Section Dana Nau

6.036 midterm review. Wednesday, March 18, 15

Chapter 2 Single Layer Feedforward Networks

Artificial Neural Networks Lecture Notes Part 3

Sample Exam COMP 9444 NEURAL NETWORKS Solutions

Neural Networks and Deep Learning

Neural Networks and Ensemble Methods for Classification

Neural Networks. Associative memory 12/30/2015. Associative memories. Associative memories

Artificial Neural Networks Examination, June 2004

Neural networks. Chapter 20, Section 5 1

The Perceptron Algorithm

Neural Nets in PR. Pattern Recognition XII. Michal Haindl. Outline. Neural Nets in PR 2

Neural Networks. Chapter 18, Section 7. TB Artificial Intelligence. Slides from AIMA 1/ 21

Artificial Neural Networks. Introduction to Computational Neuroscience Tambet Matiisen

Neural Networks biological neuron artificial neuron 1

Intelligent Systems: Reasoning and Recognition. Artificial Neural Networks

CHALMERS, GÖTEBORGS UNIVERSITET. EXAM for ARTIFICIAL NEURAL NETWORKS. COURSE CODES: FFR 135, FIM 720 GU, PhD

Introduction to Deep Learning

Computational Intelligence Winter Term 2017/18

CSC321 Lecture 5: Multilayer Perceptrons

Engineering Part IIB: Module 4F10 Statistical Pattern Processing Lecture 5: Single Layer Perceptrons & Estimating Linear Classifiers

Transcription:

CSC 302 1.5 Neural Networks Perceptron Learning Rule 1

Objectives Determining the weight matrix and bias for perceptron networks with many inputs. Explaining what a learning rule is. Developing the perceptron learning rule. Discussing the advantages and limitations of the single layer perceptron. 2

Development Introduced a neuron model by Warren McCulloch & Walter Pitts [1943]. Main features Weighted sum of input signals is compared to a threshold to determine the output. 0 if weighted_sum < 0 1 is weighted_sum >= 0 Able to compute any logical arithmetic function. No training method was available. 3

Development Perceptron was developed by Frank Rosenblatt [1950]. Neurons were similar to those of McCulloch & Pitts. Key feature introduced a learning rule. Proved that learning rule is always converged to correct weights if weights exist for the problem. Simple and automatic. No restriction on initial weights - random 4

Learning Rules Procedure for modifying the weights and biases of a network to perform a specific task. Supervised Learning - Network is provided with a set of examples of proper network behaviour (inputs/targets) Reinforcement Learning - Network is only provided with a grade, or score, which indicates network performance. Unsupervised Learning - Only network inputs are available to the learning algorithm. Network learns to categorize (cluster) the inputs. 5

Perceptron Architecture 6

Perceptron Architecture Output of the i th neuron 7

Perceptron Architecture Therefore, if the inner product of the ith row of the weight matrix with the input vector is greater than or equal to b i the output will be 1, otherwise the output will be 0. Each neuron in the network divides the input space into two regions. 8

Single-Neuron Perceptron Decision boundary n = 1 w T p+b = w 1,1 p 1 + w 1,2 p 2 + b = 0 9

Decision Boundary Decision boundary 1w T p+b = 0 or 1 w T p = -b All points on the decision boundary have the same inner product with the weight vector. Decision boundary is orthogonal to weight vector. 1 w T p 1 = 1 w T p 2 = -b for any two points in the decision boundary. 1 w T (p 1 p 2 ) = 0 Weight vector is orthogonal to (p 1 p 2 ). 10

Direction of the Weight Vector Any vector in the shaded region will have an inner product greater than b and Vectors in the un-shaded region will have inner product less than b. Therefore the weight vector 1 w will always point toward the region where the neuron output is 1. 11

Graphical Method Design of a perceptron to implement the AND gate. Input space each input vector labeled according to the target. Dark circle output is1 Light circle output is 0 12

Graphical Method First select a decision boundary that separates dark circles and light circles. Next choose a weight vector that is orthogonal to the decision boundary. The weight vector can be any length. Infinite no of possibilities. One choice is 13

Graphical Method Finally, we need to find the bias, b. Pick a point on the decision boundary (say [1.5 0] T ) Testing 14

Multiple-Neuron Perceptron Each neuron will have its own decision boundary. i w T p + b i = 0 A single neuron can classify input vectors into two categories. A multi-neuron perceptron can classify input vectors into 2 S categories. 15

Perceptron Learning Rule Supervised training Provided a set of examples of proper network behaviour where p q input to the network and t q corresponding output As each input is supplied to the network, the network output is compared to the target. The learning rule then adjusts the weights and biases of the network in order to move the network output closer to the target. 16

Test Problem = = = = = = 0, 1 0 0, 2 1 1, 2 1 3 3 2 1 t t t p p p 2 1 Input/target pairs 17 Removed the bias for the simplicity. Decision boundary must pass the origin. Decision boundaries Weight Vectors

Starting Point Random initial weight 1.0 1w= 0.8 Present p 1 to the network: a hardlim w T = ( 1 p 1 ) = hardlim 1.0 0.8 1 2 a = hardlim( 0.6) = 0 Incorrect Classification. 18

Tentative Learning Rule We need to alter the weight vector so that it points more toward p 1, so that in the future it has a better chance of classifying p 1.. 19

Tentative Learning Rule One approach would be to set 1 w equal to p 1. This rule cannot find a solution always. If we apply the rule 1 w = p every time one of these vectors misclassified, and network weights will simply oscillate back and forth. 20

Tentative Learning Rule Another possibility would be to add p 1 to 1 w. This rule can be stated as 21

22 Second Input Vector

23 Third Input Vector

24 Unified Learning Rule

25 Unified Learning Rule

26 Multiple-Neuron Perceptron

27 Apple/Banana Example

28 Second Iteration

29 Check

30 Perceptron Rule Capability

31 Perceptron Limitations