Artificial Neural Network: Definition, Advantages , Application Scope of Neural Networks,

 Artificial Neural Network: Definition

An artificial neural network (ANN) may be defined as an infonnation·processing model that is inspired by the way biological nervous systems, such as the brain, process information. This model rriis ro replicate only the most basic functions of brain. The key element of ANN is the novel structure of irs information processing system. An ANN is composed of a large number of highly interconnected prOcessing elements (neurons) wo_rking in unison to solve specific problems. Anificial neural networks, like people, learn by example. An ANN is configured for a specific application, such as pattern recognition or data classification through a learning process. In biological systems, learning involves adjustments to the synaptic connections that exist between the neurons. ANNs undergo a similar change that occurs when the concept on which they are built leaves the academic environment and is thrown into the harsher world of users who simply to get a job done on computers accurately all the time. Many neural networks now being designed are statistically quite accurate, but they still leave their users with a bad raste as they falter when it comes to solving-problems accurately. They might be 85-90% accurate. Unfortunately, few applications tolerate that level of error.

Advantages of Neural Networks

Neural networks, with their remarkable ability to derive meaning from complicated or imprecise data, could be used to extract patterns and detect trends that are too complex·ro be noticed by either humans or other computer techniques. A trained neural network could be thought of as an "expert" in a particular category of information it has been given m an.Jyze. This expert could be used to provide projections in new situations of interest and answer "what if' questions. Other advantages of worlcing with an ANN include:

 l. Adaptive learning: An ANN is endowed with the ability m learn how to do taSks based on the data given for training or initial experience. 

2. Selforganizlltion: An ANN can create irs own organization or representation of the information it receives during learning tiine. 

3. Real-time operation: ANN computations may be carried out in parallel. Special hardware devices are being designed and manufactured to rake advantage of this capability of ANNs.

 4. Fault tolerattce via reduntMnt iufonnation coding. Partial destruction of a neural network leads to the corrcseonding degradation of performance. However, caP-@lfuies.may .be reJained even after major .---·· 

Currently, neural ne[\vorks can't function as a user interface which translates spoken words into instructions for a machine, but someday they would have rhis skilL Then VCRs, home security systems, CD players, and word processors would simply be activated by voice. Touch screen and voice editing would replace the word processors of today. Besides, spreadsheets and databases would be imparted such level of usability that would be pleasing co everyone. But for now, neural networks are only entering the marketplace in niche areas where their statistical accuracy is valuable. 

Many of these niches indeed involve applications where answers provided by the software programs are not accurate but vague. Loan approval is one such area. Financial institutions make more money if they succeed in having the lowest bad loan rate. For these instirurions, insralling systems that are "90% accurate" in selecting the genuine loan applicants might be an improvement over their current selection Indeed, some banks have proved that the failure rate on loans approved by neural networks is lower than those approved by tkir 

best traditional methods. Also, some credit card companies are using neural networks in their application screening process.

 - ' I h1s newest method of looking into the future by analyzing past experiences has generated irs own unique set of problems. One such problem is to provide a reason behind a computer·generated answer, say, as to why a particular loan application was denied. To explain how a network learned and why it recommends a particular decision has been difficult. The inner workings of neural networks are "black boxes." Some people have even called the use of neural networks "voodoo engineering." To justifY the decision·making process, several neural network tool makers have provided programs that explain which input through which node dominates the decision-making process. From this information, experts in the application may be able to infer which data plays a major role in decision· making and its imponance.

 Apart from filling the niche areas, neural nerwork's work is also progressing in orher more promising application areas. The next section of this chapter goes through some of these areas and briefly details the current work. The objective is to make the reader aware of various possibilities where neural networks might offer solutions, such as language processing, character recognition, image compression, pattern recognition, etc. Neural networks can be viewed from a multi-disciplinary poim of view as shown in Figure 1-l. /

Application Scope of Neural Networks 

The neural networks have good scope of being used in the following areas: 

I. Air traffic control could be automated with the location, altitude, direction and speed of each radar blip taken as input to the nerwork. The output would be the air traffic controller's instruction in response to each blip. 

2. Animal behavior, predator/prey relationships and population cycles may be suitable for analysis by neural networks. 

3. Appraisal and valuation of property, buildings, automobiles, machinery, etc. should be an easy task for a neural network.

4. Bet#ng on horse races, stock markets, sporting events, etc. could be based on neural network predictions.

 5. Criminal sentencing could be predicted using a large sample of crime details as input and the resulting semences as output. 

6. Compkr physical and chemical processes that may involve the interaction of numerous (possibly unknown) mathematical formulas could be ·modeled heuristically using a neural network. 

7. Data mining, cleaning and validation could be achieved by determining which records suspiciously diverge from the pattern of their peers.

 8. Direct mail advertisers could use neural network analysis of their databases to decide which customers should be targeted, and avoid wa.•iring money on unlikely targets.

 9. Echo pauerns from sonar, radar, seismic and magnetic instrumems could be used to predict meir targets. 

10. Econometric modeling based on neural networks should be more realistic than older models based on classical statistics. 

11. Employee hiring could be optimized if the neural nerworks were able to predict which job applicant would show the best job performance. 

12. Expert consultants could package their intuitive expertise imo a neural network ro automate their services.

 13. Fraud detection regarding credit cards, insurance or £aXes could be automated using a neural network analysis of past incidents. 

14. Handwriting and typewriting could be recognized by imposing a grid over the writing, then each square of the grid becomes an input to the neural necwork. This is called "Optical Character Recognition." 

15. Lake water levels could be predicted based upon precipitation patterns and river/dam flows. 

16. Machinery control could be automated by capturing me actions of experienced machine operators into a neural network.

 17. Medical diagnosis is an ideal application for neural networks.

18. Medical research relies heavily on classical statistics to analyze research data. Perhaps a neural network should be included in me researcher's tool kit. 

19. Music composition has been tried using neural networks. The network is trained to recognize patterns in the pirch and tempo of certain music, and rhen the network writes irs own music. 

20. Photos ttnd fingerprints could be recognized by imposing a fine grid over the photo. Each square of the grid becomes an input to me neural network.

 21. Rmpes ttnd chemical fonnulations could be optimized based on the predicted outcome of a formula change. 

22. Retail inventories could be optimized by predicting demand based on past pauerns. 

23. River water levels could be predicted based on upstream reports, and rime and location of each report. 

24. Scheduling of buses, airplanes and elevators could be optimized by predicting demand. 

25. Staff scheduling requiren1:ents for restaurants, retail stores, police stations, banks, etc., could be predicted based on the customer flow, day of week, paydays, holidays, weather, season, ere. 

26. Strategies for games, business and war can be captured by analyzing the expert player's response ro given stimuli. For example, a football coach must decide whether to kick, piss or ru'n on the last down. The inputs for cltis decision include score, time, field location, yards w first down, etc. 

27. Traffic flows could be predicted so rhar signal tiiTling could be optimized. The neural network could recognize "a weekday morning hour during a schOol holiday" or "a winter Sunday morning." -28. Voice recognition could be obtained by analyzing )he audio oscilloscope panern, much like a smck market graph. 1 

29. Weather prediction may be possible. Inputs would)ndude weather reports from surrounding areas. Outpur(s) would be the future weather in specific areas based on the input information. Effects such as ocean currents and jet streams could be included