Supervised learning with neural networks

cartcletchΤεχνίτη Νοημοσύνη και Ρομποτική

19 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

57 εμφανίσεις

Introduction to

Machine Learning*

Prof. D. Spears

COSC 4010/5010, Section 1

Spring 2004

* This material is taken from the textbook,
Machine Learning, Volume I
,


Eds. Michalski, Carbonell, and Mitchell, Tioga, 1983, and from
Artificial


Intelligence

by Russell and Norvig.

Definition of Machine Learning


Informal definition
: Any computer program that
improves its performance at some task through
experience and/or data.


Formal definition
: 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 tasks in T, as measured by P,
improves with experience E.

Wow! Look

at how much

it learned!

Other Disciplines From Which
Machine Learning Draws Ideas and
Techniques

machine

learning

AI

probability

&

statistics

computational

complexity

theory

control

theory

information

theory

philosophy

psychology

neurophysiology

ethology

decision

theory

game

theory

optimization

biological

evolution

statistical

mechanics

Some Learning
Strategies/Techniques


Rote learning


Inductive inference


Stochastic/Bayesian inference


Deductive inference


Reinforcement learning


Neural network learning


Evolutionary learning


Clustering


Analogical learning


Learning from human instruction (being told)


Learning by discovery


Case
-
based reasoning


Speed
-
up learning


Multi
-
strategy learning is very popular



Examples of Types of Knowledge
Acquired Via Learning


Declarative Knowledge


Concepts


Preferred values of parameters


Grammars


Taxonomies


Procedural Knowledge


Rules


Rule strengths


Graphs/networks


Computer programs


Plans

Example strategies

for acquisition:

Inductive inference

Evolutionary learning

Clustering

Analogy

Induction

Reinforcement learning

Evolutionary learning

Stochastic learning

Example Data Structures Used

for Learned Knowledge



Decision trees


Logical expressions


Neural networks


Condition
-
action rules


Rule sets


Finite
-
state automata


Lisp code


C code

Type of knowledge:

Concepts

Behavioral rules

Plans

Computer programs

History of Machine Learning


1950’s:
Neural modeling


E.g., perceptrons (Rosenblatt, 1958)


Groundwork for this work was laid by researchers in mathematical
biophysics (Rashevsky, 1948) (McCulloch and Pitts, 1943).


Major thrust was on learning tabula rasa. Focus on self
-
organization
and neuron
-
like learning elements.


1960’s:
Pattern recognition and decision
-
theoretic learning


Acquire linear, polynomial, or related forms of a discriminant function
from a given set of training examples, e.g., (Nilsson, 1965).


Samuel’s checker’s program (Samuel, 1959, 1963). Acquired a master
level of performance.


Statistical decision theory for pattern recognition, e.g., (Watanabe,
1960) (Duda & Hart, 1973).


1969: Minsky & Papert on theoretical limitations of perceptron
learning.


1970s:
Adaptive control


Self
-
adjust parameters to maintain stability in spite of disturbances, e.g.,
(Davies, 1970) (Fu, 1971).

History of Machine Learning
(cont’d)


1960’s and 70’s:
Models of human learning


High
-
level symbolic descriptions of knowledge, e.g., logical expressions
or graphs/networks, e.g., (Karpinski & Michalski, 1966) (Simon & Lea,
1974).


META
-
DENDRAL (Buchanan, 1978), for example, acquired task
-
specific expertise (for mass spectrometry) in the context of an expert
system.


Winston’s (1975) structural learning system learned logic
-
based
structural descriptions from examples.


1970’s:
Genetic algorithms


Developed by Holland (1975)


1970’s
-

present:
Knowledge
-
intensive learning


A tabula rasa approach typically fares poorly. “To acquire new
knowledge a system must already possess a great deal of initial
knowledge.” Lenat’s CYC project is a good example.



History of Machine Learning
(cont’d)


1970’s
-

present:
Alternative modes of learning

(besides examples)


Learning from instruction, e.g., (Mostow, 1983) (Gordon & Subramanian,
1993)


Learning by analogy, e.g., (Veloso, 1990)


Learning from cases, e.g., (Aha, 1991)


Discovery (Lenat, 1977)


1991: The first of a series of workshops on
Multistrategy Learning
(Michalski)


1970’s


present:
Meta
-
learning


Heuristics for focusing attention, e.g., (Gordon & Subramanian, 1996)


Active selection of examples for learning, e.g., (Angluin, 1987), (Gasarch &
Smith, 1988), (Gordon, 1991)


Learning how to learn, e.g., (Schmidhuber, 1996)


History of Machine Learning
(cont’d)


1980


The First Machine Learning Workshop was held at Carnegie
-
Mellon
University in Pittsburgh.


1980


Three consecutive issues of the
International Journal of Policy
Analysis and Information Systems

were specially devoted to machine
learning.


1981


A special issue of SIGART Newsletter reviewed current projects in
the field of machine learning.


1983


The Second International Workshop on Machine Learning, in
Monticello at the University of Illinois.


1986


The establishment of the
Machine Learning

journal.


1987


The beginning of annual international conferences on machine
learning (ICML).


1988


The beginning of regular workshops on computational learning
theory (COLT).


1990’s


Explosive growth in the field of data mining, which involves the
application of machine learning techniques.


A general model of

learning agents

environment

critic

learning

element

problem

generator

performance

element

AGENT

feedback

learning goals

knowledge

changes

external

performance


standard

sensors

effectors

Evaluating Learners

A

C

C

U

R

A

C

Y




AMOUNT OF TRAINING DATA SEEN

on unseen data

Learning curves

Some Ideas for Projects


Multi
-
agent / swarm reinforcement learning


Concept learning using logical, stochastic, neural, or evolutionary
representations or hybrids


Learning a good representation for learning concepts (meta
-
learning)


Data mining: Discovering patterns in large data sets (medical? consumer?)


Modeling the process of scientific discovery


Evolving a simple artificial brain


Cognitive models of human learning


“Safe” learning


Learning in artificial life/worlds


Learning in soccer
-
playing agents


Unsupervised learning (clustering) to develop taxonomies


Learning to predict temporal sequences


Training a neural network to recognize objects, faces, etc.


Multi
-
agent learning to cooperate or compete


Learning to improve game playing strategies


Evolving computer programs (genetic programming)


Comparative studies of different learning methods


A variant of a study found in a machine learning conference paper


Analogical learning (e.g., applying knowledge of one case to a new case)


Learning a model of a student for intelligent tutoring