Download Machine Learning Basics: 1. General Introduction

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Existential risk from artificial general intelligence wikipedia, lookup

Quantum machine learning wikipedia, lookup

Person of Interest (TV series) wikipedia, lookup

Philosophy of artificial intelligence wikipedia, lookup

Barbaric Machine Clan Gaiark wikipedia, lookup

History of artificial intelligence wikipedia, lookup

Concept learning wikipedia, lookup

Pattern recognition wikipedia, lookup

Machine learning wikipedia, lookup

Transcript
Machine Learning Basics
1. General Introduction
Compiled For
Ph.D. course Work
APSU, Rewa, MP, India
Outline





Artificial Intelligence
Machine Learning: Modern
Approaches to Artificial Intelligence
Machine Learning Problems
Machine Learning Resources
Our Course
Machine Learning Basics: 1. General Introduction
Intelligence

Intelligence


Ability to solve problems
Examples of Intelligent Behaviors or
Tasks



Classification of texts based on content
Heart disease diagnosis
Chess playing
Machine Learning Basics: 1. General Introduction
Example 1: Text Classification (1)
Huge oil platforms dot the Gulf like
beacons -- usually lit up like Christmas
trees at night.
One of them, sitting astride the
Rostam offshore oilfield, was all but
blown out of the water by U.S.
Warships on Monday.
Human
Judgment
The Iranian platform, an unsightly
mass of steel and concrete, was a
three-tier structure rising 200 feet
(60 metres) above the warm waters of
the Gulf until four U.S. Destroyers
pumped some …
Machine Learning Basics: 1. General Introduction
Crude
Ship
Example 1: Text Classification (2)
The Federal Reserve is expected to
enter the government securities
market to supply reserves to the
banking system via system repurchase
agreements, economists said.
Human
Judgment
Most economists said the Fed would
execute three-day system
repurchases to meet a substantial
need to add reserves in the current
maintenance period, although some
said a more …
Machine Learning Basics: 1. General Introduction
Money-fx
Example 2: Disease Diagnosis (1)
Patient 1’s data
Age: 67
Sex: male
Chest pain type: asymptomatic
Doctor
Diagnosis
Resting blood pressure: 160mm Hg
Serum cholestoral: 286mg/dl
Fasting blood sugar: < 120mg/dl
…
Machine Learning Basics: 1. General Introduction
Presence
Example 2: Disease Diagnosis (2)
Patient 2‘s data
Age: 63
Sex: male
Chest pain type: typical angina
Doctor
Diagnosis
Resting blood pressure: 145mm Hg
Serum cholestoral: 233mg/dl
Fasting blood sugar: > 120mg/dl
…
Machine Learning Basics: 1. General Introduction
Absence
Example 3: Chess Playing

Chess Game


Two players playing one-by-one under
the restriction of a certain rule
Characteristics


To achieve a goal: win the game
Interactive
Machine Learning Basics: 1. General Introduction
Artificial Intelligence

Artificial Intelligence


Ability of machines in conducting
intelligent tasks
Intelligent Programs

Programs conducting specific intelligent
tasks
Intelligent
Processing
Input
Output
Machine Learning Basics: 1. General Introduction
Example 1: Text Classifier (1)
…
fiber = 0
Text File:
Huge oil
platforms dot
the Gulf like
beacons -usually lit up …
…
Preprocessing
…
huge = 1
Crude = 1
Classification …
…
oil = 1
platforms = 1
…
Machine Learning Basics: 1. General Introduction
Money-fx = 0
…
Ship = 1
…
Example 1: Text Classifier (2)
…
enter = 1
Text File:
The Federal
Reserve is
expected to
enter the
government …
…
Preprocessing
expected = 1
…
Crude = 0
Classification …
federal = 1
…
oil = 0
…
Machine Learning Basics: 1. General Introduction
Money-fx = 1
…
Ship = 0
…
Example 2: Disease Classifier (1)
Preprocessed data of patient 1
Age = 67
Sex = 1
Classification
Chest pain type = 4
Resting blood pressure = 160
Serum cholestoral = 286
Fasting blood sugar = 0
…
Machine Learning Basics: 1. General Introduction
Presence = 1
Example 2: Disease Classifier (2)
Preprocessed data of patient 2
Age = 63
Sex = 1
Classification
Chest pain type = 1
Resting blood pressure = 145
Serum cholestoral = 233
Fasting blood sugar = 1
…
Machine Learning Basics: 1. General Introduction
Presence = 0
Example 3: Chess Program
Searching and
evaluating
Matrix representing
the current board
Best move New matrix
Opponent’s
playing his move
Machine Learning Basics: 1. General Introduction
AI Approach

Reasoning with Knowledge



Knowledge base
Reasoning
Traditional Approaches



Handcrafted knowledge base
Complex reasoning process
Disadvantages

Knowledge acquisition bottleneck
Machine Learning Basics: 1. General Introduction
Outline





Artificial Intelligence
Machine Learning: Modern
Approaches to Artificial Intelligence
Machine Learning Problems
Research and Resources
Our Course
Machine Learning Basics: 1. General Introduction
Machine Learning

Machine Learning (Mitchell 1997)



Learn from past experiences
Improve the performances of intelligent
programs
Definitions (Mitchell 1997)

A computer program is said to learn
from experience E with respect to some
class of tasks T and performance
measure P, if its performance at the
tasks improves with the experiences
Machine Learning Basics: 1. General Introduction
Example 1: Text Classification
Classified text files
Text file 1
trade
Text file 2
ship
…
…
Training
New text file
Text
classifier
Machine Learning Basics: 1. General Introduction
class
Example 2: Disease Diagnosis
Database of medical records
Patient 1’s data
Absence
Patient 2’s data
Presence
…
…
Training
New patient’s
data
Disease
classifier
Machine Learning Basics: 1. General Introduction
Presence or
absence
Example 3: Chess Playing
Games played:
Game 1’s move list
Win
Game 2’s move list
Lose
…
…
Training
New matrix
representing
the current
board
Strategy of
Searching and
Evaluating
Machine Learning Basics: 1. General Introduction
Best move
Examples

Text Classification

Task T


Performance measure P



Precision and recall of each category
Training experiences E


Assigning texts to a set of predefined
categories
A database of texts with their
corresponding categories
How about Disease Diagnosis?
How about
Chess Playing?
Machine Learning Basics: 1. General Introduction
Why Machine Learning Is Possible?

Mass Storage


More data available
Higher Performance of Computer


Larger memory in handling the data
Greater computational power for
calculating and even online learning
Machine Learning Basics: 1. General Introduction
Advantages

Alleviate Knowledge Acquisition
Bottleneck



Does not require knowledge engineers
Scalable in constructing knowledge base
Adaptive


Adaptive to the changing conditions
Easy in migrating to new domains
Machine Learning Basics: 1. General Introduction
Success of Machine Learning

Almost All the Learning Algorithms



Reinforcement Learning


Text classification (Dumais et al. 1998)
Gene or protein classification optionally
with feature engineering (Bhaskar et al.
2006)
Backgammon (Tesauro 1995)
Learning of Sequence Labeling


Speech recognition (Lee 1989)
Part-of-speech tagging (Church 1988)
Machine Learning Basics: 1. General Introduction
Outline





Artificial Intelligence
Machine Learning: Modern
Approaches to Artificial Intelligence
Machine Learning Problems
Machine Learning Resources
Our Course
Machine Learning Basics: 1. General Introduction
Choosing the Training Experience

Choosing the Training Experience

Sometimes straightforward


Sometimes not so straightforward


Text classification, disease diagnosis
Chess playing
Other Attributes


How the training experience is controlled
by the learner?
How the training experience represents
the situations in which the performance
of the program
is measured?
Machine Learning Basics: 1. General Introduction
Choosing the Target Function

Choosing the Target Function



What type of knowledge will be learned?
How it will be used by the program?
Reducing the Learning Problem


From the problem of improving
performance P at task T with experience
E
To the problem of learning some
particular target functions
Machine Learning Basics: 1. General Introduction
Solving Real World Problems

What Is the Input?


What Is the Output?


Predictions or decisions to be made
What Is the Intelligent Program?


Features representing the real world
data
Types of classifiers, value functions, etc.
How to Learn from experience?

Learning algorithms
Machine Learning Basics: 1. General Introduction
Feature Engineering

Representation of the Real World Data


Features: data’s attributes which may be useful
in prediction
Feature Transformation and Selection


Select a subset of the features
Construct new features, e.g.



Discretization of real value features
Combinations of existing features
Post Processing to Fit the Classifier

Does not change the nature
Machine Learning Basics: 1. General Introduction
Intelligent Programs

Value Functions



Classifiers (Most Commonly Used)



Input: features
Output: value
Input: features
Output: a single decision
Sequence Labeling


Input: sequence of features
Output: sequence of decisions
Machine Learning Basics: 1. General Introduction
Examples of Value Functions

Linear Regression



Input: feature vectors x  ( x1 , x2 ,, xn )
n
Output: f (x)  w  x  b   wi xi  b
Logistic Regression


i 1
Input: feature vectors x  ( x1 , x2 ,, xn )
1
Output: f (x) 
 w x b
1 e
Machine Learning Basics: 1. General Introduction
Examples of Classifiers

Linear Classifier



Input: feature vectors x  ( x1 , x2 ,, xn )
n
Output: y  sgn( w  x  b)  sgn(  wi xi  b)
Rule Classifier

Decision tree


i 1
A tree with nodes representing condition
testing and leaves representing classes
Decision list

If condition 1 then class 1 elseif condition 2
then class 2 elseif ….
Machine Learning Basics: 1. General Introduction
Examples of Learning Algorithms

Parametric Functions or Classifiers

Given parameters of the functions or
classifier, e.g.


Estimating the parameters, e.g.


Linear functions or classifiers: w, b
Loss function optimization
Rule Learning


Condition construction
Rules induction using divide-and-conquer
Machine Learning Basics: 1. General Introduction
Machine Learning Problems

Methodology of Machine Learning



General methods for machine learning
Investigate which method is better under
some certain conditions
Application of Machine Learning


Specific application of machine learning
methods
Investigate which feature, classifier,
method should be used to solve a certain
problem
Machine Learning Basics: 1. General Introduction
Methodology

Theoretical


Mathematical analysis of performances of
learning algorithms (usually with
assumptions)
Empirical

Demonstrate the empirical results of
learning algorithms on datasets
(benchmarks or real world applications)
Machine Learning Basics: 1. General Introduction
Application

Adaptation of Learning Algorithms


Directly apply, or tailor learning
algorithms to specific application
Generalization

Generalize the problems and methods in
the specific application to more general
cases
Machine Learning Basics: 1. General Introduction
Outline





Artificial Intelligence
Machine Learning: Modern
Approaches to Artificial Intelligence
Machine Learning Problems
Machine Learning Resources
Our Course
Machine Learning Basics: 1. General Introduction
Introduction Materials

Text Books



T. Mitchell (1997). Machine Learning,
McGraw-Hill Publishers.
N. Nilsson (1996). Introduction to
Machine Learning (drafts).
Lecture Notes


T. Mitchell’s Slides
Introduction to Machine Learning
Machine Learning Basics: 1. General Introduction
Technical Papers

Journals, e.g.



Machine Learning, Kluwer Academic
Publishers.
Journal of Machine Learning Research,
MIT Press.
Conferences, e.g.


International Conference on Machine
Learning (ICML)
Neural Information Processing Systems
(NIPS)
Machine Learning Basics: 1. General Introduction
Others

Data Sets



UCI Machine Learning Repository
Reuters data set for text classification
Related Areas





Artificial intelligence
Knowledge discovery and data mining
Statistics
Operation research
…
Machine Learning Basics: 1. General Introduction
Outline





Artificial Intelligence
Machine Learning: Modern
Approaches to Artificial Intelligence
Machine Learning Problems
Machine Learning Resources
Our Course
Machine Learning Basics: 1. General Introduction
What I will Talk about

Machine Learning Methods



Simple methods
Effective methods (state of the art)
Method Details



Ideas
Assumptions
Intuitive interpretations
Machine Learning Basics: 1. General Introduction
What I won’t Talk about

Machine Learning Methods



Classical, but complex and not effective
methods (e.g., complex neural networks)
Methods not widely used
Method Details

Theoretical justification
Machine Learning Basics: 1. General Introduction
What You will Learn

Machine Learning Basics





Methods
Data
Assumptions
Ideas
Others


Problem solving techniques
Extensive knowledge of modern
techniques
Machine Learning Basics: 1. General Introduction
References






H. Bhaskar, D. Hoyle, and S. Singh (2006). Machine
Learning: a Brief Survey and Recommendations for
Practitioners. Computers in Biology and Medicine, 36(10),
1104-1125.
K. Church (1988). A Stochastic Parts Program and Noun
Phrase Parser for Unrestricted Texts. In Proc. ANLP1988, 136-143.
S. Dumais, J. Platt, D. Heckerman and M. Sahami
(1998). Inductive Learning Algorithms and
Representations for Text Categorization. In Proc. CIKM1998, 148-155.
K. Lee (1989). Automatic Speech Recognition: The
Development of the Sphinx System, Kluwer Academic
Publishers.
T. Mitchell (1997). Machine Learning, McGraw-Hill
Publishers.
G. Tesauro (1995). Temporal Difference Learning and
TD-gammon. Communications of the ACM, 38(3), 58-68.
Machine Learning Basics: 1. General Introduction
The End