Can neural networks be useful for business?

Growing since the day that Alan Turing introduced the world to the idea of the artificial neural network in his 1948 paper "Intelligent Machinery", neural networks have evolved as a useful and dynamic tool which offer an efficient and effective business solutions to common problems for modern businesses; problems such as fitness approximation, data processing, and robotics .They are particularly effective in situations where complexity of the data or task makes providing a solution impractical by hand.

Real life applications:

• In 2005 Sirakaya, Delen and Choi published a paper which investigated community support for commercial gaming in the United States. They utilised an artificial Neural Network to conduct the research and used variables such as geographic space, proximity to population centres and church membership to make their assessment.

• Amir F. Atiya in 2001 publised his article entitled “Bankruptcy Prediction for Credit Risk Using Neural Networks: A Survey and New Results”. This paper was written for the corporation he was working for namely IEEE. Using the neural network which he developed he was able to provide a corporate bankruptcy prediction model with an accuracy from 81.46% to 85.5%.

Neural Networks provide an alternative to conventional techniques which are often limited by strict rules of normality, linearity, variable independence etc. Because a neural network can capture many kinds of relationships it allows the user to easily show or explain phenomena which otherwise may have been very difficult or impossible to explain.

Workings of a Genetic Algorithm

Facts about the genetic algorithm:

1. Inspired by Darwins theory of Evolution or the notion of ‘survival of the fittest’. We will see below that when a new population is being created the two fittest chromosomes from the previous population are selected for the creation of the new population.


2. Each feasible solution to the problem can be plotted as point on a graph which represents the search space or fitness landscape. Thus the search space contains all feasible solutions.
   
3. For any problem we are always looking for the best possible solution available, as the genetic algorithm utilises a formula this will usually be either a maximum or a minimum value. On the graph, the optimum point or the highest point always represents the fittest solution.
   
   How it works:
   1. A random population of chromosomes (possible solutions to the problem) is generated.
   2. The fitness of each chromosome (solution) is evaluated.
   3. A new population is then created by: 1) selecting two chromosomes in the population according to their fitness, 2) performing crossover probability on the parent chromosomes to create two new chromosomes (offspring), 3)then using mutation probability randomly change the gene values of the offspring 4)and finally placing the offspring in the new population.
   4. The new population is used to run the algorithm again.
   5. Check if the fittest possible solution has been found (highest point on the graph) and if not steps 2, 3 and 4 are repeated until it is.

History of Artificial Intelligence

Since Karel Capek's play in 1921, ‘Rossums Universal Robots’ coined the term ‘Robots’ in the English language as a term to describe the young inventor Rossum’s artificial version of humananity, the idea of robots and indeed artificial intelligence have been a fascination of mankind. The ancestry of modern artificial intelligence can traced far into ancient times, with the notion of artificial beings blessed ‘with conciseness and being’. It was after the invention of the programmable computer in the 1940’s that the idea of building a computer to mimic the workings of the human brain became a plausible theory to scientists.

In 1950 the father of modern computer science Alan Turing proposed the Turing Test in his paper ‘Computing Machinery and Intelligence’. This test worked on the basis that if a person could not tell the difference between communicating with a machine and communicating with human in another room, then we can call that machine an intelligent machine.

In 1956 John McCarthy coined the term ‘Artificial Intelligence’ at the ‘Dartmoth Conference’ – the first conference dedicated to the idea.

In 1957 The General Problem Solver was demonstrated by Newell, Shaw & Simon. This was a machine which offered logical solutions to problems such as theorems and algebra.

In 1962 the first industrial robot company, Unimation, was founded. It produced a materials handling robot used mainly in the manufacturing industry.

In 1969 Shakey was built by SRI International. It became the first robot to apply reasoning to its own actions, and respond to commands.

In 1979 the Stanford Cart. This was the first computer controlled vehicle and required room sized computers for operation. The Cart’s main function was the ability to follow a white line drawn across a floor using its TV camera and other methods.

In more modern times robots have begun to capture the imagination of society more and more. The was most evident when in 1997 ‘The Deep Blue’ chess program took on and beat the world chess champion, Garry Kasparov, in a widely followed match. This brought real life robotics to the masses.

Since then people have become familiar with many commercially available robotic toys which utilise the wonders of artificial intelligence. Examples include ‘Furby’ and the phenomenon which was the ‘Tomagotchi ‘.