Dr azimifar pattern recognition lect1
0 likes40 views
The document provides an overview of pattern recognition, defining key concepts such as objects, features, and patterns, and outlining various applications in fields like bioinformatics and industrial automation. It details the importance of data preprocessing, normalization, and feature extraction in enhancing the effectiveness of supervised learning techniques. Additionally, it discusses the training and testing phases in recognizing patterns through learned parameters.
![Introduction
Definition of Pattern Recognition
Pattern Recognition Applications
Definition of Object, Feature, and Pattern
Definition of Feature Space
Data Preprocessing and Normalization
Feature Extraction and Selection
Train and Test
Different Learning Techniques
Supervised Learning
Lecture Summary
Data Preprocessing and Normalization
The preprocessing module segments the pattern of interest from
background, removes noise, and normalizes the pattern.
A feature vector includes various features characterizing the pattern
from different views (e.g., weight, color, height).
Feature entities appear in different measurement ranges (e.g.,
[−1, 1], [−1000, 1000], [0.500, 0.505]), with different impacts.
Normalization rescales all values of a feature vector to a predefined
range (e.g., [-1,1]). This process leads to an even and fair impact of
all features in the decision stage.
Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 11 / 18](https://image.slidesharecdn.com/drazimifarpatternrecognitionlect1-210228132021/85/Dr-azimifar-pattern-recognition-lect1-11-320.jpg)
Dr azimifar pattern recognition lect1
- 1. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Statistical Pattern Recognition Lecture1 Dr Zohreh Azimifar School of Electrical and Computer Engineering Shiraz University September 20, 2014 Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 1 / 18
- 2. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Table of contents 1 Introduction 2 Definition of Pattern Recognition 3 Pattern Recognition Applications 4 Definition of Object, Feature, and Pattern 5 Definition of Feature Space 6 Data Preprocessing and Normalization 7 Feature Extraction and Selection 8 Train and Test 9 Different Learning Techniques 10 Supervised Learning 11 Lecture Summary Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 2 / 18
- 3. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Introduction Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 3 / 18
- 4. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Definition of Pattern Recognition Pattern Recognition is a subset of artificial intelligence, in which intelligent softwares and hardwares are designed to mimic human thinking and behaviour. Such systems are capable of recognizing, describing, classifying and predicting unseen datum. Pattern Recognition is the best possible venue to make an automatic decision by using available sensors, processors and knowledge. Well-known concepts from statistical decision theory are unitized to compute decision boundaries between pattern classes. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 4 / 18
- 5. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Pattern Recognition Applications Table: Pattern Recognition Applications Problem Domain Application Input Pattern Pattern Class Bioinformatics Sequence Analysis DNA/Protein Known types of sequence genes/patterns Searching for Points in Compact and Data Mining meaningful patterns multidimensional well-separated space clusters Document Internest Search Text document Semantic categories Classification (e.g., sport) Industrial Printed circuit Intensity or Defective / Automation board inspection range image non-defective nature of product Biometric Personal identification Video clip Action, posture Recognition Speech Telephone dictionary Speech waveform Spoken words Recognition Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 5 / 18
- 6. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Definition of Object, Feature, and Pattern Object is defined by a set of characters and measures. Feature is defined as a measurement (numeric or nominal) obtained from the object and is viewed in a d-dimensional space. Observation is an example of feature vector obtained from a given object. Pattern is defined as opposite of chaos; an identity the can be given a name (Watanabe:1985). Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 6 / 18
- 7. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Definition of Object, Feature, and Pattern Figure: Each fruit is an object, and every object is characterized by a feature vector, including color and weight. Feature vector is called a pattern, which can be labelled. Here labels are Banana and Orange. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 7 / 18
- 8. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Definition of Feature Space Patterns are represented by feature vectors. Each feature element is considered as a dimension is a space. Feature space is a multidimensional space whose size is defined by the number of feature elements in the pattern’s feature vector. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 8 / 18
- 9. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Definition of Feature Space Table: Features and Labels associated with object and feature vector definition Color Weight Label 170 230 Banana 110 190 Orange 160 220 Banana 210 180 Banana 125 200 Orange 115 185 Orange 180 240 Banana 130 190 Orange Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 9 / 18
- 10. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Definition of Feature Space Figure: 2D feature space wherein each point represents an individual pattern. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 10 / 18
- 11. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Data Preprocessing and Normalization The preprocessing module segments the pattern of interest from background, removes noise, and normalizes the pattern. A feature vector includes various features characterizing the pattern from different views (e.g., weight, color, height). Feature entities appear in different measurement ranges (e.g., [−1, 1], [−1000, 1000], [0.500, 0.505]), with different impacts. Normalization rescales all values of a feature vector to a predefined range (e.g., [-1,1]). This process leads to an even and fair impact of all features in the decision stage. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 11 / 18
- 12. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Feature Extraction and Selection Object is medium which is sensed. Observation is the measurement gathered by our sensory elements. Feature is a preprocessed observation ready for modelling the medium. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 12 / 18
- 13. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Train and Test The recognition system is operated in two modes: training (learning) and testing Training includes: feature extraction/selection for representing input patterns and a classifier to partition the feature space. Testing assigns/predict a category to an unseen pattern based on a vector of feature extracted from it. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 13 / 18
- 14. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Train and Test Figure: Train: system learns fruits label. Test: system uses its learned parameters to predict label for unseen patterns. Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 14 / 18
- 15. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Train and Test Figure: Train and Test Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 15 / 18
- 16. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Different Learning Techniques Supervised Learning Classifier Regression Unsupervised Learning clustering Density Estimation Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 16 / 18
- 17. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Supervised Learning Figure: General Diagram of Supervised Learning Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 17 / 18
- 18. Introduction Definition of Pattern Recognition Pattern Recognition Applications Definition of Object, Feature, and Pattern Definition of Feature Space Data Preprocessing and Normalization Feature Extraction and Selection Train and Test Different Learning Techniques Supervised Learning Lecture Summary Lecture Summary 1 2 3 4 5 Dr Zohreh Azimifar, 2014 Statistical Pattern Recognition Lecture1 18 / 18