This spring, Lehigh University students will have the opportunity to take a new course on machine learning that focuses on the theory of the statistical analysis of the performance of machine learning algorithms. The course, Machine Learning and Statistical Decision Making, is being taught by Rick Blum, I-CPIE faculty and Robert W. Wieseman Professor of Electrical & Computer Engineering, who has been working in the area of statistical performance analysis of data-driven decision making algorithms for over 30 years. As the director of Lehigh’s Signal Processing and Communication Research Lab, Blum is making contributions to the foundational theories of machine learning, statistical decision making, distributed sensor data fusion and processing, and cybersecurity. Machine learning is often considered a form of artificial intelligence in which the computer learns on its own.
Machine learning, as well as deep learning (a form of machine learning in which specific types of neural networks are trained), is an area of increasing interest as researchers explore how to automate historically “human” tasks (e.g., self-driving cars). Blum decided to teach this course with an explicit focus on performance analysis to provide students with the tools to understand the performance of machine learning algorithms and contemplate yet-unanswered questions about the capabilities of machine learning in real applications. One of those questions—which Blum considers to be one of the most important unanswered questions—is “when will deep learning work well and when will it not?” After taking the course, students will understand the existing theory and will have the required background to pursue further investigations to extend theory in this area. He hopes some will ultimately consider these important unanswered questions, which most current researchers believe to be difficult to address.
Though Blum has offered courses on the statistical performance analysis of data-driven decision making algorithms to Electrical & Computer Engineering (ECE) students in the past, given the broad interest in this topic, Blum designed the course to appeal to both ECE and non-ECE students. He has found that bringing together students with different backgrounds usually results in a more interesting class. Graduate students or outstanding undergraduate students in any major with an interest in machine learning and strong mathematics and probability training are encouraged to enroll. In Blum’s experience, Lehigh students give “excellent class presentations” that enhance the learning experience. With that in mind, students who enroll in the course can expect to contribute to the class through individual presentations and discussions.
For more information on this course, see Machine Learning and Statistical Decision Making.