Advanced personalized learning is listed at the top of the fourteen Grand Challenges in the 21st century by the National Academy of Engineering. The scientific reason behind it is that learning styles, speeds, and interests all vary from individual to individual. This challenge is particularly significant for engineering – our students are trained to define a problem from a real-world observation, identify the underlying physics, simplify and solve the problem, and explain the observation using the solution – each of these actions can be performed in numerous correct ways which will depend on student preference and personality. Therefore, the core of my teaching philosophy is student-oriented instruction to realize personalized learning.
ME 5510 Scientific Machine Learning for Mechanical Engineers, graduate level, Northeastern University, 2025 Fall
ME 3455 Dynamics and Vibrations, undergraduate level, Northeastern University, 2025 Spring
ME 5374 Special Topic: Scientific Machine Learning for Mechanical Engineers, graduate level, Northeastern University, 2024 Fall
ME 3455 Dynamics and Vibrations, undergraduate level, Northeastern University, 2023 Spring
Photo credit: Wikipedia
Structure of our course
An important feature of this course is that we emphasize the underlying mathematics behind ML algorithms and data-enabled engineering. Here are some of the topics and concepts that we derived together in class on board.
Decision boundary, universal approximation theorem, and solving the XOR problem using a 2-2-1 net
Matrix calculus and backpropagation
Adjoiont method and ODE/PDE-constrained optimization
Gaussian process and uncertainty quantification
Surrogate modeling and introduction to Bayesian Optimization
Symmetry in data and order-reduction techniques