Not The Whole Truth

New algorithms excel where even
the fastest computer fail, says Biplab Das 

How to solve real-life problems that stump even the fastest digital computers? To address this problem, scientists have discovered new computing tools, which are collectively known as soft computing. "Unlike digital computers, soft computing mimics human-like decision-making process that deals with imprecise and vague data that abound in real life," said Prof. Sankar Pal of the Indian Statistical Institute (ISI). 

He was delivering a lecture at the Paribesh Bhavan on 'Soft Computing, Machine Intelligence and Data Mining: Prospects in Bioinformatics,' the first of the 'Bengal Science Lectures', a unique series organised by the science and technology council, West Bengal government. 

"One of the main components of soft computing is genetic algorithm," said Pal. "It is designed to simulate processes in nature that favour the survival of the fittest animal." He explained that genetic algorithms seek out the optimum solution for a given a problem from a myriad solutions. 

According to Pal, genetic algorithms are useful when traditional mathematical analysis is available. "For its ability to process data billions times larger than those used by conventional computers, genetic algorithms have huge applications in bioinformatics," he said. "They have been used for predicting protein structures and modelling various aspects of the natural immune system." 

Genetic algorithm help in the comparison of protein structures, which is crucial for the pharmaceutical industry for the development of new drugs. "By comparing 3D structures of proteins, we can uncover many unknown facets of biological activities like the molecular basis of diseases and the 'cross-talk' between cells at the molecular level," Pal said. 

The real world doesn't capture binary distinctions well, and components of soft computing make machine capable of going beyond the boundary of binary logic. "One such component is fuzzy logic," Pal said. "It has been evolved to handle concept of partial truths - between what's 'completely true' and 'completely false'." While digital reasoning recognises only 'yes' and 'no' values, fuzzy logic handles ambiguous values like 'maybe', 'nearly' and 'very'. 

Fuzzy logic gives us a way to deal with such situations. "In fuzzy systems, values are indicated by a number (called a truth value) in the range from 0 to 1, where 0.0 represents absolute falsehood and 1.0 represents absolute truth," Pal said. "While this range evokes the idea of probability, fuzzy logic and fuzzy sets operate quite differently from probability. For instance, if I tell you that my height is 5 feet 6 inches, you won't be able to figure out whether you will consider me short or not. You may reckon me short for a man but tall for a woman." 

Pal claimed that besides being very good at tackling real-life problems, soft computing can increase machine intelligence. How to analyse and choose the required information from large databases? Computer scientists have hit upon a solution called 'data mining'. "It is, in some ways, an extension of statistics, with little artificial intelligence and machine learning twists thrown in," Pal said. 

Giving an example, he said a catalogue retailer might need to decide who should receive information about a new product. "The information handled by the data mining process is contained in a historical database of previous interactions with the customers, such as age, zip code, their responses," he said. "The data mining software would use this historical information to build a model of customer behaviour that could be used to predict which customers would be likely to respond to the new product." 

By using this information a marketing manager can select only those customers who are most likely to respond. The operational business software can then feed the results of the decision to the appropriate touch point systems (call centers, direct mail, email systems, etc.) so that the right customers receive the right offers. According to Pal, data mining can help business planner anticipate customers' behaviour patterns and trends. For example, an outdoor equipment retailer can predict the likelihood of a backpack being purchased on the basis of a consumer's purchase of sleeping bags and hiking shoes.