neural networks introduction

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.


1. Introduction:

An Artificial Neural Network (ANN) is a data processing model. It is stimulated by biological nervous systems. For example brain is one of the biological nervous systems. The main purpose of this model is the novel formation of the data analyzing method. It is developed by many analyzing elements which are used to analyze the particular troubles. An Artificial Neural Network is a structure depends on the tasks of biological neural networks. An Artificial Neural Network is a combination of number of neurons which are connected together based on a particular network structure. The goal of Artificial Neural Network is to translate the given data input to meaningful extracted outputs. It can recognize a visual pattern or structure. It can control the movement of objects.

An input is given to the neural network. The preferred or objective response place at the output. It will be performed in the training process. An error will be generating based on the data differentiation within the preferred output and the application's output. The error data will be given to the application again to correct the system in a strategic procedure. It based on the learning procedure. This procedure will be continued until we get the desired output from the application. Here the performance is heavily based on the data information.

Learning processes is simple through many examples. Like that Neural Networks getting trained by some existing example patterns. An Artificial Neural Network is developed for a particular purpose like character recognition, structured pattern recognition and information differentiation, by some existing learning method.

1.1 Why use neural networks?

An Artificial Neural Network has the ability to extract the right meaning from difficult or complex information. It is used to derive the patterns and identify the procedures which are very complicate to differentiate by other techniques. A neural network will be trained to analyze the given complex data or information. The neural network becomes an expert after getting trained by some existing learning process. Neural networks are developed by many processing units which are connected into a network. Their calculation power based on working together on any task - this is sometimes termed parallel processing. A neural network provides many advantages in many different situations. Some of the advantages are the following,

  1. If any element in the neural network gets damaged, the total system will not fail. It will perform the task without any disturbance. Because it is parallel in nature.
  2. We no need to reprogram the network.
  3. Neural network can be applied in any kind of task.
  4. Neural network can be applied without any problem
  5. Adaptive learning: It is the capability to find out how to perform the given operation based on some existing procedure, training or experience.
  6. Self-Organization: An Artificial Neural Network can develop a new pattern or data representation by its own at the time of learning.
  7. Real Time Operation: An Artificial Neural Network can perform the parallel tasks.
  8. A neural network can perform operations that a linear program can not.

1.2 Background:

Neural network applications are present in the recent programming development. This field was established many years before. Much advancement has been made in this field because of the in efficiency of some computer simulations.

A neural network performs different approach to solve the problem comparing with the conventional computers. Normally convention computer follows an algorithmic procedure to solve any kind of problem. That is the computer has some set of procedures to solve the problem. Based on the steps it find outs the solution. It restricts the conventional computers' capability to solve the problem efficiently. We already understand that to solve the problem. But the computers should do much more what we don't know to do exactly. An Artificial Neural Network can perform the complex tasks. Neural network performs the task like the human brain does. The neural network is the collection of many analyzing elements. It may perform the parallel task to solve the problem. Artificial neural networks are like biological neural networks. It performs more than one task at a single time. It is mostly used in the field of statistics, artificial intelligence and cognitive psychology. Neural network models are structured based on the simulation of nervous system. Nowadays many number of software are implemented using artificial neural networks. It is very useful for the complexity of data, complexity of patterns and various difficult tasks.

1.3 Applications of neural networks

Artificial Neural Networks are normally used in the following models:

  1. Recognizing the speakers in communications
  2. Recovers the communication if any problem occurred in the software.
  3. Undersea mining
  4. Used to analyze the textures
  5. To recognize the three dimensional objects
  6. Hand-writing recognition
  7. Face recognition
  8. Data validation

Neural networks also used in the medicine. At present the research is on the part of human body and also disease recognition using many scans. The human cardiovascular system can be model by the Neural Networks.

Neural networks also used in the business purposes. Resource allocation and scheduling can be done using the artificial neural networks. It is mostly used in the field of data mining. It is used to differentiate and identify the different patterns present in the database.

1.3.1 Pattern Recognition

The main purpose of neural networks is pattern recognition. It can be executed by training the network. At the time of training, the neural network will be trained to relate the desired outcome with the inputs. Whenever we use that particular network, it analyzes the input and produces the related output. The output pattern is based on the training provided. Pattern recognition is very often used for data classification, diagnosis, and hand-writing recognition, identifying fingerprints, character recognition and learning from examples. The feed-forward networks are mostly useful to the pattern recognition. This kind of networks has no feedback to the input. Neural networks learn from the patterns provided as input. It learns by the mistakes provided as feedback to the system. It is used to identify the mistakes and reconstruct the patterns. It is very difficult to develop like these neural networks.

1.3.2 Neural Networks in Practice

Neural Networks are broadly used in many real world business applications. Many industries are using many applications of Neural Networks. All business applications have been successfully executed in all industries. Normally Neural networks are used to find the patterns and trends. They are used to predict the trends in data.

Some of the prediction and forecasting applications are,

  1. trades forecasting
  2. manufacturing procedure control
  3. consumer investigate
  4. data verification
  5. threat managing
  6. objective marketing

Some of the specific examples are,

  1. Recognizing the speakers in communications
  2. Recovers the communication if any problem occurred in the software
  3. Undersea mining
  4. Used to analyze the textures
  5. To recognize the three dimensional objects
  6. Hand-writing recognition
  7. Face recognition
  8. Data validation

1.3.3 Neural networks in medicine

Artificial Neural Networks are the important investigation area in medical side. The research in medical field on Artificial Neural Networks is mainly for biomedical systems. It may provide an extensive application in this field. At present the investigation in mainly on human body modeling and diseases recognition through many scan techniques. Some of the scan techniques are CAT scans, cardiograms and ultrasonic scans. Neural networks are very useful in disease recognition. It will be achieved by various scans mentioned above. For these types of applications we no need to specify any algorithms. To identify the disease these algorithms will not be used. To identify these diseases we need only some example disease representation and various disease variations. Selecting the number examples in not important. The examples should be selected in very careful manner and in an efficient manner.

1.3.4 Modeling and Diagnosing the Cardiovascular System

Neural Networks are mainly used in human cardiovascular system modeling application. Diagnosis can be achieved by comparing with the individual cardiovascular system model and patient's physiological measurement. This process will be performed regularly and the risky situations will be identified in the very first stage. It helps to handle the disease easily.

The individual's cardiovascular system model is based on many physiological variables in various stages. Some of the variables are diastolic blood pressures, systolic blood pressures, heart rate and breathing rate. If is successfully implemented in an individual, then it will be used as physical condition model for the individuals. This model should be work properly on all individuals without any supervision of others. It is the process of neural network.

There is one more reason to use the Artificial Neural Networks. Artificial Neural Networks can able to provide the sensor fusion. This sensor fusion combines various values of the different sensors. Sensor fusion stimulates the Artificial Neural Networks to identify various difficult relationships between the sensor values of the individuals. Artificial Neural Networks can able to detect various complex medical situations through the Sensor fusion method.

1.3.5 Neural Networks in business

Neural Networks applications are used in several areas of business. Business has many general areas and specializations. Neural network applications are used almost in all specialization areas. There are two main fields of business applying the concept of neural networks. They are scheduling and resource allocation. Database mining is one of the main research areas of Artificial Neural Networks. In this field most of the data will be classified and differentiated with the use of Artificial Neural Networks.

1.3.6 Marketing

The Airline Marketing Tactician is one of the marketing applications of Artificial Neural Networks. The Airline Marketing Tactician is abbreviated as AMT. AMT is computer system which consists of many intelligent technologies. A feed forward neural network concept has been implemented in this Airline Marketing Tactician application. This application is specially used of monitoring and booking advice for every departure. These kind of information provides profitability and technological advantage top the system users. The Artificial Neural Networks concept is used for discovering the influence of many undefined variables interactions. These kinds of interactions were used through the Artificial Neural Networks. It is used to develop various useful conclusions.

1.3.7 Credit Evaluation

Many number of neural network applications have been developed by The HNC Company. Credit scoring system is one among them. This system provided the profit up to 27% comparing with the existing application system. Mortgage screening is also one of the applications of Artificial Neural Networks which was implemented by the HNC Company. Automated mortgage insurance underwriting system is another application of Artificial Neural Networks. It was developed by the Nestor Company.

1.4 Optical character recognition

OCR, usually abbreviated as Optical character recognition. It is used to recognize the characters which are present in the images of handwritten, printed text or type written.

OCR is one of the concept of artificial intelligence, pattern recognition and computer vision. This type of system requires training about the samples of each character images to recognize the correct image efficiently. Using this concept of OCR most of the systems can able to provide the output that closely matched with the input image. Optical character recognition is becoming more and more important in the modern world. It helps humans ease their jobs and solve more complex problems. An example is handwritten character recognition.

Some handwriting recognition system allows us to input our handwriting into the system. This can be done either by controlling a mouse or using a third-party drawing tablet. The input can be converted into typed text or can be left as an "ink object" in our own handwriting.

Optical Character Recognition is one of the main applications of Artificial Neural Network. Character recognition concept is the difficult one comparing with the other applications. In the beginning it looks like simple. But applying this application in a system is very difficult in practical. Optical Character Recognition plays very important role in the various companies and various industries. Optical Character Recognition is mainly used in the fields of shipping and banking sectors. An automatic scanning method is used the U.S post offices. It is used to recognize the numerical digits of ZIP codes. The numerical characters are in the form of printed image. This printed image will be recognized by the Optical Character Recognition system. Optical Character Recognition is the procedure to recognize the character images, printed or scanned images. The system produces the approximate output based on the input given to system to recognize. Optical Character Recognition is one of the main fields of computer science. It reads the text from the image format and provides the output based on the input given. An Optical Character Recognition system can able to get a magazine or book as an input to recognize or modify the content to get the required output format. In the Optical Character Recognition system the inputs such as bitmap image or scanned image will be analyzed for the light and dark places of that image. It is used to identify each and every alphabets and numeric digits. Optical Character Recognition system is mainly used in many libraries. Optical Character Recognition system is also used for the credit card slips, arrange the mails and to process the checks in many banking sectors. A huge number of magazines, articles and letters are prepared and sorted using the Optical Character Recognition system every day to speed the process.

Every Optical Character Recognition system contains complex software to analyze and recognize the images. Some of the Optical Character Recognition system includes the combination of both specialized hardware and complex software for recognizing the alpha numeric characters, even though many low cost systems performing this process entirely by software. Some of the advanced Optical Character Recognition systems can able to recognize the text in various fonts and styles. The advanced Optical Character Recognition systems also have the complexity with handwritten characters and texts. The Optical Character Recognition systems are already being used in many industries and fields. But it required huge time to recognize on those days. But nowadays it becomes simpler, efficient and it can able to perform in a few seconds. The advanced Optical Character Recognition system is widely used in many fields of industries, because of its cost effectiveness and time consuming properties.

In the olden days the Optical Character Recognition system can be used. First the source document should be scanned through an optical scanner for reading the documents and pages as an image file. It was in the form of dots. And much complex software was used for recognizing the content present in the document. The Optical Character Recognition system software starts the process to recognize. This software can differentiate the images and alpha numeric texts. And it can identify the characters in the light and dark places.

The older Optical Character Recognition system compares this image with existing bitmap images. This process can be done based on the style and fonts present in the image. The present Optical Character Recognition system used various algorithms for the recognition process. Neural network technology algorithms are used to analyze the text characters. Each and every procedure identifies the dark and lights areas and also the character strokes. And it matches with the stored bitmap images to produce the result based on the character strokes. The advanced Optical Character Recognition software system can able to recognize a various font and texts. But it is still difficult to recognize the handwritten characters efficiently. The programmers are handling many approaches and procedures to improve the efficiency of handwriting recognition.