Sunday, January 24, 2010
Neural Network Task Two
The second task I have performed was to compare the networks that were tested with different resolution of data. The data set used was the same as the previous, which was from A to J. In the previous task, the data file used is of low resolution, that is, a total of 11 scans per character. For this task, I was using medium resolution data (24 scans) and high resolution data (47 scans) to run the network. It was found out from the previous task that the RBF network with 50 prototype neurons performs slightly better than the MLP network and therefore RBF was used here. The setting for the RBF network was exactly the same as the previous one with 50 prototype neurons except that for the number of input neurons have to be changed to 24 neurons for medium resolution and 47 neurons for high resolution. The network was trained for 20,000 iteration for each resolution before the RMS error converges. The scoring method used was also same as the previous and all results are recorded for both medium and high resolution data.
Neural Network Task One
In the first task of neural network, I had done an investigation on training two different kinds of networks using small data sets containing only character A to J. The networks used were MLP and RBF network. The data file used for training consists of the first five example of each of the ten character A to J and the data file used for testing consists of the last five example. All data files used in this task are low resolution data.
First, using the software NeuralWorks, create a MLP network with 11 input neurons, 10 hidden neurons and 10 output neurons. For low resolution data, there were only 6 horizontal scans and 5 vertical scans, a total of 11 scans. Therefore I have used 11 input neurons. Since there were only 10 classes (A to J), I think it was reasonable to use 10 to 20 hidden neurons for classification. The maximum and minimum table was selected, as the input values were not between 0 and 1. The network was trained for 20,000 iterations before the RMS error converges to a stable value. I have run the networks for three times using three different numbers of hidden neuron, 10, 15 and 20 hidden neurons. All the three results of RMS error is recorded. Next, I have evaluated the performance of the network more formally using the scoring system. A recall is made and output values are recorded in the ‘nnr’ file. Since most of the output values are below 0.5, I have decided to use the ‘winner takes all’ strategy’, that is, the classification was determined by the highest output values. Score 1 for outputs that were correctly classified and score 0 for mis-classification.
After training the MLP network, the next step was to train a RBF network and the performance was compared between these two networks. The same amount of input and output neurons were used, that is, 11 and 10 neurons respectively, since the same data files were used for training and testing. As for the prototype neurons, I have run the network for five times using five different numbers of neurons (10, 30, 50 and 70 neurons). Although there are only 10 classes, I have found that the result gets better if the number of prototype neurons were increased. The network also take about 20,000 iteration of training before its RMS error converges. The same scoring method was also used to evaluate the performance of the network and all results are recorded.
First, using the software NeuralWorks, create a MLP network with 11 input neurons, 10 hidden neurons and 10 output neurons. For low resolution data, there were only 6 horizontal scans and 5 vertical scans, a total of 11 scans. Therefore I have used 11 input neurons. Since there were only 10 classes (A to J), I think it was reasonable to use 10 to 20 hidden neurons for classification. The maximum and minimum table was selected, as the input values were not between 0 and 1. The network was trained for 20,000 iterations before the RMS error converges to a stable value. I have run the networks for three times using three different numbers of hidden neuron, 10, 15 and 20 hidden neurons. All the three results of RMS error is recorded. Next, I have evaluated the performance of the network more formally using the scoring system. A recall is made and output values are recorded in the ‘nnr’ file. Since most of the output values are below 0.5, I have decided to use the ‘winner takes all’ strategy’, that is, the classification was determined by the highest output values. Score 1 for outputs that were correctly classified and score 0 for mis-classification.
After training the MLP network, the next step was to train a RBF network and the performance was compared between these two networks. The same amount of input and output neurons were used, that is, 11 and 10 neurons respectively, since the same data files were used for training and testing. As for the prototype neurons, I have run the network for five times using five different numbers of neurons (10, 30, 50 and 70 neurons). Although there are only 10 classes, I have found that the result gets better if the number of prototype neurons were increased. The network also take about 20,000 iteration of training before its RMS error converges. The same scoring method was also used to evaluate the performance of the network and all results are recorded.
Friday, January 22, 2010
Objective of Using Neural Network
Below are what I have targeted to perform:
- Compare the performance of a MLP network and a RBF network for recognizing handwritten character.
- Determine the effects of using different resolution data for training and testing on the RBF network.
- Determine the effects of using different data sets for training and testing on the RBF network.
- Demonstrate the suitability of the Hopfeild network to recognize character.
- Write a rule-based program to validate the vehicle registration marks.
- Write a rule-based program to interpret the local memory tags and age identifiers of the vehicle registration marks.
- Write a rule-based program to suggestive alternative for misidentify characters.
Introduction of Project in Neural Network
In this project, an investigation was carried out using neural networks and knowledge-based systems to recognize, validate and interpret handwritten characters.
Neural network is one of the techniques used in artificial intelligent. It consists of many nonlinear computational elements that form the network neurons, linked by weighted interconnections.
The neurons are designed similar to the neurological system of animals. The networks are most effective in performing tasks like classification and error correction.
There are a few kinds of neural networks and those that are used in this project are MLP, RBF and Hopfield networks. They are being compared to see which is the most suitable for character recognition. Data files with different resolutions and data sets are used to train and test out the network.
Knowledge-based systems are intelligent systems that are build with the flexibility of adding new knowledge to the program without affecting the whole system. A rule-based system is a knowledge-based system where the knowledge-base is represented in the form of a sets of rules where rules are an flexible means of expressing knowledge.
This project uses a rule-based system to first validate a vehicle registration mark entered by user and then interpret the marks to identify where and when the vehicle is registered. Rules are also written to suggest alternative for misidentify characters. For example, a character ‘0’ are incorrectly identify as ‘O’ and these rules will change the character ‘O’ back to ‘0’. An agent-based system was also worked in conjunction with the rule-based system to validate vehicle registration marks.
Neural network is one of the techniques used in artificial intelligent. It consists of many nonlinear computational elements that form the network neurons, linked by weighted interconnections.
The neurons are designed similar to the neurological system of animals. The networks are most effective in performing tasks like classification and error correction.
There are a few kinds of neural networks and those that are used in this project are MLP, RBF and Hopfield networks. They are being compared to see which is the most suitable for character recognition. Data files with different resolutions and data sets are used to train and test out the network.
Knowledge-based systems are intelligent systems that are build with the flexibility of adding new knowledge to the program without affecting the whole system. A rule-based system is a knowledge-based system where the knowledge-base is represented in the form of a sets of rules where rules are an flexible means of expressing knowledge.
This project uses a rule-based system to first validate a vehicle registration mark entered by user and then interpret the marks to identify where and when the vehicle is registered. Rules are also written to suggest alternative for misidentify characters. For example, a character ‘0’ are incorrectly identify as ‘O’ and these rules will change the character ‘O’ back to ‘0’. An agent-based system was also worked in conjunction with the rule-based system to validate vehicle registration marks.
Thursday, January 21, 2010
A Brief Summary of my Project
The objective of my final project is to develop intelligent systems capable of processing handwritten vehicle registration marks.
The main techniques used in this project will be on creating neural networks (NNs) to automatically recognize the individual handwritten characters that go to make up a vehicle registration marks and also by using knowledge-based systems (KBSs) to validate the intended registration mark and identify where and when the vehicle was registered.
The first part the project start off by doing a feasibility study of a MLP (multilayer perceptron) and a RBF (Radial Basis Function) networks to see which will contribute to a better performance in recognizing characters. Only the characters A to J were tested at this point of time and low resolution data was used. It was found that RBF is a better network to do the job and it was then use to perform on medium and high resolution data. The result shows that the network performs better on medium resolution data. The network was also used to train and test on a full set of characters (‘0’ to ‘9’, ‘A’ to ‘B’ and ‘space’). The last part on neural network is to use a Hopfield network to recognize characters. It was found that out of ten test patterns there were nine correct outputs.
The second part involves the investigation of a knowledge-based system. First, a rule-based program was written to validate the format of the vehicle registration marks using flex. If users enter an invalid mark, a message will be printed out in the console to tell the users that they have entered an invalid mark. The program continues to be written for interpreting the local memory tags and the age identifiers of the marks. This would identify where and when the vehicle was registered given a valid registration marks. The final portion of the program was written to suggest alternative for those characters that have misidentified during neural network processing.
Finally a agent-based system was build for discovering information about the status of the vehicles with a particular registration marks.
The main techniques used in this project will be on creating neural networks (NNs) to automatically recognize the individual handwritten characters that go to make up a vehicle registration marks and also by using knowledge-based systems (KBSs) to validate the intended registration mark and identify where and when the vehicle was registered.
The first part the project start off by doing a feasibility study of a MLP (multilayer perceptron) and a RBF (Radial Basis Function) networks to see which will contribute to a better performance in recognizing characters. Only the characters A to J were tested at this point of time and low resolution data was used. It was found that RBF is a better network to do the job and it was then use to perform on medium and high resolution data. The result shows that the network performs better on medium resolution data. The network was also used to train and test on a full set of characters (‘0’ to ‘9’, ‘A’ to ‘B’ and ‘space’). The last part on neural network is to use a Hopfield network to recognize characters. It was found that out of ten test patterns there were nine correct outputs.
The second part involves the investigation of a knowledge-based system. First, a rule-based program was written to validate the format of the vehicle registration marks using flex. If users enter an invalid mark, a message will be printed out in the console to tell the users that they have entered an invalid mark. The program continues to be written for interpreting the local memory tags and the age identifiers of the marks. This would identify where and when the vehicle was registered given a valid registration marks. The final portion of the program was written to suggest alternative for those characters that have misidentified during neural network processing.
Finally a agent-based system was build for discovering information about the status of the vehicles with a particular registration marks.
My Final Year Project
In this blog, I'll be sharing the content of my final year project which relates to intelligent systems. I am a computer science student and the final year project I have done was all about Artificial Intelligence . It's about developing intelligent systems to recognize handwritten characters on vehicle registration marks.
Subscribe to:
Posts (Atom)