By Anantha Krishnan M
Express News Service
Express News Service
Bangalore: Want new ideas? Log on to nature! Well, this could be the in-thing Indian defence scientists are wedded to, while exploring the mesmerising possibilities of Artificial Intelligence (AI). The Centre for Artificial Intelligence and Robotics (CAIR), the focal point of this series, is developing AI network models that emulate the computing style of intelligence inherent in nature. “Nature’s creations consist of highly complex systems. The human brain’s Neural Network is an example. The brain is able to perform complex tasks by simultaneous computations, based on sophisticated signalling between neurons. Swarm Intelligence is another example of collective intelligence seen in nature, most commonly in bees and termites,” say sources.
While computer science has progressed considerably, accurately mimicking nature and making them exhibit any significant degree of perception or intelligence, still remains beyond our reach. Increase in computation power has not yet helped in intelligent cognition. Computational intelligence offers the mechanism of ‘learning’ algorithms that arrive at solutions based on ‘experience’, much like a human does.
“The powerful train-by-example paradigm allows the algorithm to learn based on encountered data, obviating the need for complex classification metrics. In this paradigm, labelled training data is provided to the neural network model, which ‘learns’ to give the correct responses, even for those inputs not previously encountered,” say sources.
Data mining toolbox: A good data mining tool must discover patterns in data. These can then be used to make predictions. However, the biggest problem with military data is that it is unstructured. The state-of-the-art toolbox for doing Pattern Recognition and Knowledge Handling developed at the CAIR is said to be effective for intelligence analysis and decision support in Net Centric Operations.
“It can lend cognitive capabilities to a software defined radio and provide speech and speaker recognition engines. It can autonomously sort the vast amount of information received, impossible for human analysts to collate and analyse. It rejects incorrect or spurious inputs, collates usable inputs and creates inferences under the relevant topic heads,” sources said.
Swarm Intelligence Algorithms are another example of AI mimicking nature. Scientists at the CAIR has developed a site selection algorithm based on the simple, distributed nest construction behaviour of termites with no central control or supervision. “The swarm behaviour has been improvised upon to create a geographical site selection tool used to select suitable sites for any kind of requirement off a digital map, typically military deployments. The algorithm examines a designated area for suitability based on the termite behaviour patterns, using pre-fed logic, and displays selected areas on a geo-spatial information system (GIS) display,” sources said.
While computer science has progressed considerably, accurately mimicking nature and making them exhibit any significant degree of perception or intelligence, still remains beyond our reach. Increase in computation power has not yet helped in intelligent cognition. Computational intelligence offers the mechanism of ‘learning’ algorithms that arrive at solutions based on ‘experience’, much like a human does.
“The powerful train-by-example paradigm allows the algorithm to learn based on encountered data, obviating the need for complex classification metrics. In this paradigm, labelled training data is provided to the neural network model, which ‘learns’ to give the correct responses, even for those inputs not previously encountered,” say sources.
Data mining toolbox: A good data mining tool must discover patterns in data. These can then be used to make predictions. However, the biggest problem with military data is that it is unstructured. The state-of-the-art toolbox for doing Pattern Recognition and Knowledge Handling developed at the CAIR is said to be effective for intelligence analysis and decision support in Net Centric Operations.
“It can lend cognitive capabilities to a software defined radio and provide speech and speaker recognition engines. It can autonomously sort the vast amount of information received, impossible for human analysts to collate and analyse. It rejects incorrect or spurious inputs, collates usable inputs and creates inferences under the relevant topic heads,” sources said.
Swarm Intelligence Algorithms are another example of AI mimicking nature. Scientists at the CAIR has developed a site selection algorithm based on the simple, distributed nest construction behaviour of termites with no central control or supervision. “The swarm behaviour has been improvised upon to create a geographical site selection tool used to select suitable sites for any kind of requirement off a digital map, typically military deployments. The algorithm examines a designated area for suitability based on the termite behaviour patterns, using pre-fed logic, and displays selected areas on a geo-spatial information system (GIS) display,” sources said.
Copyright@The New Indian Express
Part 12, tomorrow: Reasoning,
based on Artificial Intelligence