自淫系列

A childhood passion for math, philosophy, electromagnetic waves, and problem-solving led Soosan Beheshti to become an electrical engineer. Her early research and studies on communication systems design piqued her curiosity and led her to consider a range of questions on data modelling for the purpose of prediction and control. This area would become the foundation of Beheshti鈥檚 research, and her models can be adapted to a broad number of machine learning applications, from medical imaging to data clustering. 

For Beheshti, simpler is better. 鈥淭hat鈥檚 modelling,鈥� she says. In her research on statistical signal and data processing, Beheshti harnesses data for parametric modelling. To get to the underlying structure of a set of observed data, she turns to Occam鈥檚 razor law of parsimony philosophy, a 14th-century problem-solving principle that argues, 鈥淓ntities should not be multiplied without necessity.鈥� 

鈥淗ow much we trust the data dictates the complexity of the model structure that we consider,鈥� says Beheshti. As more data is gathered that requires a faster modelling process, her research will continue to focus on meeting the challenges related to model complexity, validity, and reliability.