Microsoft Research in India has formed a group to work on advanced software algorithms, looking for a way to more efficiently process and analyse the large volumes of data generated by web applications and scientific research.
The challenges thrown up by the increased data volumes can't be solved by throwing more computing resources at them, said Padmanabhan Anandan, managing director of Microsoft Research India.
The focus of new algorithms has to shift to sampling data efficiently, rather than trying to analyse all of the data, even if results may not be completely accurate, he said.
“You have to do statistical sampling when mining such large quantities of data to get patterns in it, and perhaps help scientists see laws in it,” said Ravindran Kannan, a principal researcher at Microsoft Research, who heads the new group.
Algorithms are conceptual and mathematical descriptions of a method for solving a problem, which are then included in computer programs. Traditional algorithms depend on analysing data that is stored locally, and focus on achieving exact results even if it takes time. That approach is no longer suitable because of the large volumes of data being analyzed, Kannan said.
The research into principles and techniques for the new algorithms will be relevant across a wide variety of applications, he said.
Microsoft decided to work on algorithms in India because educational institutions in the country have considerable expertise in theoretical computer science, which makes it easier to find and attract talented staff, Anandan said.
The research lab in Bangalore also works on other areas including cryptography, security, mobility, and technologies for bridging the digital divide. It works in tandem with top educational institutions in the country, such as the Indian Institute of Science (IISc) and the Indian Institutes of Technology (IITs). Research that could be a differentiator for Microsoft’s own business tends to be done in-house, however, and later patented by Microsoft, Anandan said.