For more than a century, scientists have been using probability and statistics to measure the natural world. They want to make sense of data and find meaningful signals in the noise. But in the last few years, classical statistics have started to seem a little threadbare. Researchers now have access to large datasets, which are driving new insights in disciplines ranging from biology to ecology to economics. It's as true in biology, with the advent of genome sequencing, as it is in astronomy, with telescope surveys charting the entire sky.
The data have changed. Maybe it's time our data analysis tools did, too.
The course runs June 11 through September 3, 2018. Register online through Complexity Explorer.
The class will help students use concepts from the field of algorithmic complexity to search for solutions to fundamental questions. Scientists have long observed connections in natural systems, but finding evidence of causality — that is, why this set of circumstances leads to that outcome — is a thorny problem outside the scope of classical statistics. The class will introduce algorithms that can be used as tools that move beyond traditional mathematical approaches and harness the ideas of complexity to better illuminate causality.
Hector Zenil and Narsis Kiani, who lead the Algorithmic Dynamics Laboratory at the Karolinska Institute in Stockholm, Sweden, teach the course. The first part will introduce and explain the preliminary concepts needed to understand the second part, which is research-driven. Participants will analyze their own data through the new algorithmic tools.
“Our idea is to ask students to perform experiments on their own data,” says Zenil.
The approach introduced in the class can be applied to any discipline, says Zenil, from biological evolution to finance to physics to psychology. And while they may not provide answers to every question about fundamental causes in nature — “in the natural world, there are open questions that cannot ever be solved,” notes Zenil — these tools give researchers sophisticated ways to deal with big data.
The cost of the course is $50 and includes a textbook; financial aid is available.