Different Types of K.B.S.

Different Types of K.B.S.Different Types of K.B.S.

• Expert Systems• Mimic the reasoning processes of human experts

• Example Applications include -• Diagnostic systems

• (Doctor, Technician, car mechanic etc.)

• Identification systems

• (Materials spillage, Bacterial agent identifier, etc.)

• Decision Support systems • (Planning, scheduling, design systems)

Different Types of K.B.S.Different Types of K.B.S.

• Expert System Definitions• “A computer program that uses symbolic

knowledge and inference to reach conclusions” (Dictionary of AI, D. Mercadal, 1990)

• “A computer system which can act as a human expert within one particular field of knowledge” (P.Smith, 1990)

Different Types of K.B.S.Different Types of K.B.S.• Expert System Definitions• “An expert system is regarded as the embodiment

within the computer of knowledge based component from an expert skill, in such a form that the system can offer intelligent advice or take an intelligent decision about a processing function. A desirable additional characteristic, which many would consider fundamental, is the capability of the system, on demand, to justify it’s own line of reasoning in a manner directly intelligible to an enquirer. The style adopted to attain these characteristics is rule based programming”

• (Formal definition agreed by the British Computer Society’s specialist group on Expert Systems)

Different Types of K.B.S.Different Types of K.B.S.Expert Database

Acquisition Module

Knowledge base

Inference engine

Explanatory interface

Acquires knowledge

Representation of knowledge

Reasons using knowledge

The human window

User

Core of ES

Different Types of K.B.S.Different Types of K.B.S.• Expert Systems are suitable when -• The problem is important to business• The expertise required is available and stable• The knowledge required is scarce• The problem is recurrent• The problem is the right level of difficulty• The domain is well defined and of manageable size• The solution depends on logical reasoning, not

“common sense” or general knowledge

Different Types of K.B.S.Different Types of K.B.S.Expert Database

Acquisition Module

Empty K. B.

Inference engine

Explanatory interface

Medical knowledge

User

Core of ES

Car mechanic’s knowledge

Design knowledge

Different Types of K.B.S.Different Types of K.B.S.The Brain

Pattern recognition Association Complexity Noise tolerant

The Machine

Calculation Precision Logic

Different Types of K.B.S.Different Types of K.B.S.The Von Neumann architecture uses a single processing unit

The brain

- tens of millions of operations per second - Absolute arithmetic precision

- uses many slow, unreliable processors acting in parallel

Different Types of K.B.S.Different Types of K.B.S.• Features of the Brain -• 10 Billion neurones• Average several thousand connections each• Hundreds of operations per second• Low reliability• Die frequently and are never replaced• Problems are compensated for by massive

parallelism

Different Types of K.B.S.Different Types of K.B.S.The structure of neurones

Different Types of K.B.S.Different Types of K.B.S.• The Structure of Neurones• A neurone only “fires” if it’s input signal

exceeds a threshold level within a short time period

• Synapses vary in strength– Good connections allow a large signal– Slight connections only allow a weak signal– Synapses can be either exhibitory or inhibitory

Different Types of K.B.S.Different Types of K.B.S.

Sj f (Sj) Xj

ao

a1

a2

an

+1wj0wj1

wj2

wjn

A Classic Artificial Neuron

Different Types of K.B.S.Different Types of K.B.S.

Supervised Training Unsupervised Training

Perceptron/Multi-LayerPerceptron

RadialBasis

Function

KohonenSelf-Organising

Map

ART 2

BayesianMethods

Others

Neural Network Taxonomies

Different Types of K.B.S.Different Types of K.B.S.• Multilayer Perceptron

Output Values

Input Signals (External Stimuli)

Output Layer

AdjustableWeights

Input Layer

Different Types of K.B.S.Different Types of K.B.S.• Types of Layer• The Input Layer

– Introduces input values into the network– No activation function or other processing

• The Hidden Layer(s)– Perform classification of features– Two hidden layers are sufficient to solve any problem– More layers may do better

• The output Layer– Functionally just like the hidden layers– Outputs are passed on to the outside world

Different Types of K.B.S.Different Types of K.B.S.• Back Propagation• Calculate output error for each pattern• Adjust weights into output nodes to reduce the

error• Propagate errors backwards towards input layer• Repeat iteratively until satisfied• Presenting a complete set of training data is called

an epoch

Different Types of K.B.S.Different Types of K.B.S.• Building a Network• Encode problem in a form suitable for Neural

Networks• Gather training data• Define network architecture• train network• Use the trained network on new problems

Different Types of K.B.S.Different Types of K.B.S.• Network Architecture• Number of layers and units per layer• Input and output will be defined by the problem• Hidden layers defined by the designer• Decide how many training patterns to use

Different Types of K.B.S.Different Types of K.B.S.• Overtraining

• A sufficient number of nodes can classify any training set exactly

• May have poor generalisation ability

• Cross Validation• Typically, 50% of training patterns are not used• These are used to test the network’s abilities by

determining a validation error• Training is stopped when the validation error

starts to go up

Different Types of K.B.S.Different Types of K.B.S.• Example Applications• Engine Management• Engine behaviour is influenced by a large number

of parameters such as -– temperature at various points– Fuel/air mixture– lubricant viscosity– etc. etc...

Different Types of K.B.S.Different Types of K.B.S.• Example Applications• Signature Recognition• All signatures are different• There are structural similarities which are

difficult to quantify• Neural networks can recognise features of

signatures with a high level of accuracy• They can consider the speed at which a signature

was written, as well as the shape

Different Types of K.B.S.Different Types of K.B.S.• Example Applications• Stock Market Prediction• “Technical Trading” refers to trading based solely

on known statistical parameters (I.e. previous price)

• Neural networks have been used to attempt to predict changes in prices

• The success of neural networks here is difficult to assess due to secrecy

Different Types of K.B.S.Different Types of K.B.S.• Example Applications• Mortgage Assessment• Neural networks can be used to assess lending

risks• Artificial networks have produced a 12% reduction

in errors compared with human experts

Different Types of K.B.S.Different Types of K.B.S.• Case Based Reasoning• Case Based Reasoning (CBR) provides an

automated method for storing experience and reusing it to make decisions in the future

• Example Applications• Help desk applications• Application of the Law

Different Types of K.B.S.Different Types of K.B.S.• Implementing C.B.R.• Collect the important features which define each

new case presented to the system• Retrieve past cases matching these features most

closely• Use the matching case to solve the problem• If no match found find an alternative solution and

record both problem and solution• If multiple solutions are found then resolve any

ambiguities• Multiple solutions may sometimes be acceptable

Different Types of K.B.S.Different Types of K.B.S.• Implementing C.B.R.• The process is crucially dependent on 3

things -• Appropriate methods for indexing cases using

their key attributes• Efficient mechanisms for retrieving cases given a

set of index values• Good presentation of the information to the user