Published: 03rd March 2021
Big Data expert Devika Subramanian: I love science, but I want science to help someone else
Professor Devika Subramanian spoke about finding socially relevant problems and using data and Machine Learning to solve them
Science has to have the power to help people, said Professor Devika Subramanian, a well-known researcher in the area of Big Data and Machine Learning, and a professor at Rice University. She spoke about the challenges and opportunities that these two areas provide as part of the Thought Leadership Series organised by Great Lakes Institute of Management, Chennai.
She began her talk by saying that there is such a huge amount of data available in the world today and we can gather it at relatively low costs. A lot of people use data for various reasons. For instance, societies like China track every single data to control. In London, because of IRA, there are cameras all over. But Devika has been trying to research how we can use all this data to help answer questions that really matter to us. "I'm a problem solver. I love the sience I do, but at the end of the day, I want science to help someone else. That's how I choose my projects," she says.
Devika has been doing research in AI and ML since 1982 and has several projects that address real-world problems. She explained how Machine Learning helps in understanding complex systems. It helps when you don't have physical models of the system and yet you want future predictions of what it can do.
For instance, for Amazon, customers are the complex system. They want to observe you over a period of time and decide what they should do to make you buy. They get data from 200 million people in the US. Their learning algorithms are not sophisticated, but with their amount of data, they can build good predictive models. "But how easy is it to put together such a system? What data should be gathered? You have to decide what aspect of the system you want to measure. That is the data selection problem," Devika said.
She described how she finds socially relevant problems like predicting hospital readmission, predicting esophageal cancer from NBI images, learning conflict models and terrorist networks by mining wire news stories, weather forecasting and predicting human behaviour. She also showed how she measured attitudes toward gun control through Twitter accounts. "Data is a new source of power. By combining engineered framework with AI models learned from data, we can improved quality of predictions and build robust decision-making systems in the future," she concluded.