Scalable Learning Machines
Developing learning systems that can be customized according to the skills and abilities of students.
Classroom learning is passé. Well, almost. That has been brought about by the availability of borderless online learning systems based on big data. Over the past six years or so, scores of universities around the world have come forward with open online courses, which are collectively called Massive Open Online Courses (MOOCs). More than 8000 MOOCs are offered by 800 universities, and this is a growing trend. In the last quarter alone, 200 universities have come forward with 600 MOOCs. This burgeoning repository of Big Data presents a huge opportunity for the application of machine learning in education. However, data analysts have tended to rely overly on neural networks instead of applying modern machine learning technologies. To address this lacuna, Prof Volkan Cevher at EPFL’s Laboratory for Information and Inference Systems has initiated a study that aims to design the next generation of online education systems within a machine learning framework.
Contrary to existing approaches, the project seeks to develop learning systems that can be customized according to the skills and abilities of students. The quality of learning will be verified by using statistical theory to ensure that the delivery of educational content is individualized automatically to cater to the specific learning styles, skills, and background of the students. By applying such modern machine learning principles, the research will result in future-proof knowledge delivery systems.
An expert in optimization, Prof Cevher is applying learning analytics to optimize dozens of data points in the education system. In layman terms, the project aims to prepare the upcoming learning activity for a particular learner by looking at their performance in previous activities and the past performances of all learners in that upcoming activity. The study is also developing an appropriate (Bayesian) optimization framework for such individualized systems.
Earlier this year, the research project (“Theory and Methods for Accurate and Scalable Learning Machines”) received funding from the Swiss National Science Foundation (SNSF).