Sie sind auf Seite 1von 9

ARTIFICIAL INTELLIGENCE (BEB 45103) Abstract: LRT is one of the intelligence machines nowadays.

In this assignment, we have decided to adjust the speed of the LRT. We want to adjust the speed until the error is less than 0.001%. Introduction: This assignment is about LRT System recognition application by using neural network. These images will be the training images. The training LRT System could be speed. Each have one input. Then, the multilayer neural network should be build using Joone. The neural network will consist of one input layer, one or more hidden layer and an output layer. This multilayer neural network is shown as in parameter. The input layer should have number of input corresponding to the number of pixels of the LRT that setup earlier. The number of neurons in the hidden layer is arbitrary and the optimum number could be determined by trial and error. Neural network can be defined as a model of reasoning based on the human. The brain consists of a densely interconnected set of nerve cells, or basic information-processing units, called neurons. The neurons are connected by links, and each link has a numerical weight associated with it. Weights are the basic means of long-term memory in ANNs. A neural network learns through repeated adjustments of these weights. It will adjust the weights until it get the new weights. Objectives: 1. To minimize the error. 2. To learn about neural network. Variables: Input Output Speed ,Y = 1 to 10 (range) Angle = - 60 to 60 (range) Distance = 2 to30 (range)

ARTIFICIAL NEURAL NETWORK

Page 1

ARTIFICIAL INTELLIGENCE (BEB 45103) Network: Multilayer perceptron (MLP). Parameters:

ARTIFICIAL NEURAL NETWORK

Page 2

ARTIFICIAL INTELLIGENCE (BEB 45103) w13 = 1.25825 w23 = -10.00516 theta 3 = 9.07664 w14 = -3.47707 w24 = 12.80490 theta 4 = 0.32142 w35 = -11.40323 w45 = 12.78040 theta 5 = -1.44624

Activation function : Sigmoid function Learning algorithm : Feedforward. Procedures: 1. Enter and save the data in TD file ( LRT.td)

ARTIFICIAL NEURAL NETWORK

Page 3

ARTIFICIAL INTELLIGENCE (BEB 45103) 2. Open the ANN.vbp. 3. Add form and select ViewNeuron.frm. Must have two forms ViewNeuron.frm and ANN.frm. 4. Start and begin the training. 5. Train the data until it get the minimize errors. Max input

Min input

ARTIFICIAL NEURAL NETWORK

Page 4

ARTIFICIAL INTELLIGENCE (BEB 45103) 6. After train get the parameters.

At Visual Basic (VB). 1. The design of the system.

ARTIFICIAL NEURAL NETWORK

Page 5

ARTIFICIAL INTELLIGENCE (BEB 45103) 2. The program

Private Sub Command1_Click() Distance = Txtdistance Angle = Txtangle 'Set max and min minDistance = 2 maxDistance = 30 minAngle = -60 maxAngle = 60 'Normalize inputs Distance = (Distance - minDistance) / (maxDistance - minDistance) Angle = (Angle - minAngle) / (maxAngle - minAngle) 'Initialize weight w13 = 1.25825 w23 = -10.00516 theta3 = 9.07664 w14 = -3.47707 w24 = 12.8049 theta4 = 0.32142 w35 = -11.40323 w45 = 12.7804 theta5 = -1.44624 'Feedforward y3 = Distance * w13 + Angle * w23 + theta3 y3 = 1 / (1 + Exp(-y3))

These parameters get from the ANN.vbp after training the data.

ARTIFICIAL NEURAL NETWORK

Page 6

ARTIFICIAL INTELLIGENCE (BEB 45103) y4 = Distance * w14 + Angle * w24 + theta4 y4 = 1 / (1 + Exp(-y4)) y = y3 * w35 + y4 * w45 + theta5 y = 1 / (1 + Exp(-y)) Txtspeed = y End Sub

3. The output

maxDistance=30 maxAngle=60

minDistance=2 minAngle=-60 ARTIFICIAL NEURAL NETWORK Page 7

ARTIFICIAL INTELLIGENCE (BEB 45103) Analysis:

From this assignment, first we have to define the variable input and output. In this LRT System the input are distance and angle. The output is speed. We use MLP network from this assignment because we have two input and one output and two hidden layer. We have five numbers of neurons. Analysis of perceptron pointed out a number of critical weaknesses of perceptron, and for a period of time, interest in perceptron. In this assignment, we use sigmoid function. Based on the program, to get the Y output, we have used the formula of sigmoid: y3= sigmoid (x1w13+x2w23-theta3), y4=sigmoid (x1w14+x2w24theta4) and y5 = sigmoid (y3w35+y4w45-theta5). We didnt get the SOSE (Sum of Square Error) less than 0.0001 because the numbers of data are not accurate. The SOSE (Sum of Square Error) that we get is 0.0012. When we run the program at the Visual Basic (VB), the output from the (VB) is approximately to the data that we have trained. The maximum distance is 30 and the minimum distance is 2. The maximum angle is 60 and the minimum angle is -60. At maximum input, the output is 0.999670088777728. It is approximately to 1 and at minimum input, the output is 0.004229559. It is approximately to 0. The activeX must be installed to the Visual Basic (VB).

ARTIFICIAL NEURAL NETWORK

Page 8

ARTIFICIAL INTELLIGENCE (BEB 45103)

Conclusion:

The computing world has a lot to gain from neural networks. Their ability to learn by example makes them very flexible and powerful. Furthermore there is no need to devise an algorithm in order to perform a specific task; i.e. there is no need to understand the internal mechanisms of that task. Perhaps the most exciting aspect of neural networks is the possibility that some day 'conscious' networks might be produced. There is a number of scientists arguing that consciousness is a 'mechanical' property and that 'conscious' neural networks are a realistic possibility. Finally, I would like to state that even though neural networks have a huge potential we will only get the best of them when they are integrated with computing, AI, fuzzy logic and related subjects. Neural networks are suitable for predicting time series mainly because of learning only from examples, without any need to add additional information that can bring more confusion than prediction effect. On the other hand, it is generally not possible to determine exactly what a neural network learned and it is also hard to estimate possible prediction error. However, neural networks were often successfully used for predicting time series. A neural network learns through repeated adjustments of these weights. It will adjust the weights until it get the new weights.

ARTIFICIAL NEURAL NETWORK

Page 9

Das könnte Ihnen auch gefallen