Sie sind auf Seite 1von 8

“ANGEL KANCHEV” UNIVERSITY OF RUSE

INTELLIGENT SYSTEMS

Student:Parvana Yunusova

Ruse
2017

What is Neural Network?


The simplest definition of a neural network(NN) is provided by the inventor of
one of the first neurocomputers,Dr.Robert Hecht-Nielsen.He defines a neural
network as follows:
“Neural Network is a computing system made up of a number of simple,hihgly
interconnected processing elements which process information by their
dynamic state response to external inputs.”
An Artificial Neural Network (ANN), usually called Neural Network is a
mathematical model or computational model that tries to smilate the structure
and/or functional aspects of biological neural networks.In essense, creating of
ANN has been inspired from Biological Neural Network.In case of a biological
neural net, a neuron consists of a soma (cell body), axons (sends signals), and
dendrites (receives signals).

Figure 1 biological neuron


The dendrites of a neuron receive the information by special connections, the
synapses. After the cell nucleus (soma) has received a plenty of activating
signals by synapses or dendrites, the soma accumulates these signals. As soon
as the accumulated signal exceeds a certain value (called threshold value), the
cell nucleus of the neuron activates an electrical pulse which then is
transmitted to the neurons connected to the current one. The pulse is
transferred to other neurons by means of the axon.
Figure 2. . Illustration of a artificial neural network

Neural networks resemble the human brain in the following two ways.

 A neural network acquires knowledge through learning.

 A neural network’s knowledge is stored within inter-neuron connection


strengths known as synaptic weights.
Neural Network are typically organized in layers.Layers are made up of a
number of interconneted “nodes” ,also refered as neurons.Every neuran is
connected with other neuron through a connection link.Connections
represents weights. Weights are the information used by the neural network to
solve a problem.Patterns are presented to th network via the input layer ,which
communicates to one or more hidden layers where the actual processing is
done via a system of weighted connections.Most ANN’s contain some form of
‘learning rule’ which modifies the weights of the connections according to the
input patterns that it is presented with.In a sense,ANNs learn by example as do
their biological counterparts;a child learns to recognize dogs from example of
dogs.
The Artificial Neural Network receives input from the external world in the form
of pattern and image in vector form. Each input is multiplied by its
corresponding weights. The weighted inputs are all summed up inside
computing unit (artificial neuron). In case the weighted sum is zero, bias is
added to make the output not- zero or to scale up the system response. Bias has
the weight and input always equal to ‘1’.In order to limit to response to arrive
desired value,the sum is passed through activation function.

Objectives of the report


In this report we try to use ANN for regression.Regression is one of the
supervised learning problem.What is supervised learning? -In supervised
learning,problem has a labelled training set and decision boundary that
separate them as the positive label examples and negative label examples.The
goal is to find this boundary.So that according to given data set we already
know our correct output should look like.

Figure 3. .typical supervised learning problem’s data


So,in supervised learning we are taking inputs and estimating corresponding
outputs. If we are given regression problem,it means that we have pairs of
data.For example,in the following regression model we have a one dimensional
value for x and correspond to x one dimensional value of t.

Figure 4.Regression
The objective in regression problem is to learn some sort of function(in the
picture indicated by green line) and via this function we will be able to predict
output t using input x (indicated by circles in the picture). So that,we can say
regression analyzes the relationship between a dependent(target) and
independent(predictor) variables .This problem is called regression because
output is supposed to be real valued unlike classification.

The use of NN for regression


For regression poblem we use the egg data set which consist of 5 feature of an
egg and based on these feature it is categorized into 4 size(S,M,L,XL). The data
set will be seperated two group:train and test set.The features of an egg are:
Area(S),Perimeter(P),Bix axis(D),small axis(d),mass(m). The data set will be
seperated two group:train and test set.Train and test set consists of 20 ,10
data respectively. In this regression problem,the inputs are
Area(S),Perimeter(P),Bix axis(D),small axis(d) ,output is mass(m) and the
relationship will be found between them for each size.In this report matlab
neural network toolbox will be used for this regression problem.

Results
Regression will be implement for S size:
Performance(training):

After training we use network for test data and get following results.
Output : desired output:

Mean Square Error : 2.6567


Regression will be implement for M size:

Output : desired output:


Mean Square Error : 9.1068
Regression will be implement for L size:
Output : desired output:

Mean Square Error : 1.4692


Regression will be implement for XL size:
Output : desired output:

Mean Square Error : 1.7094

Das könnte Ihnen auch gefallen