Machine Learning Systems is a semester long course offered to masters degree
students of Ramakrishnamission & Vivekananda
Education and Research Institute (RKMVERI), from February to July, 2021.
The course is specifically designed to make the students industry ready.
With equal emphasis on theory and practical aspects of bulding a system,
the couse has been broadly divided in three main topics, namely,
- Study of large scale systems with object-oriented Python
- Hands-on machine learning and deep learning
- Deeper discussions of
the theory of machine learning
I. Large scale systems with object-oriented Python
The course starts with hands-on implementation of toy examples where
theoretical connections are deliberately forsaken while trusing the
intuition of the audience. This permits the students to gather a fast
overview of the vast areas of machine learning, and more importantly,
the students initially focus on how to put together various components
of a machine learning system in order to build a reliable and robust
functional entity.
II. Hands-on machine learning and deep learning
This part of the course focuses on hands-on demonstration with scikit-learn ecosystem for exploring various machine
learning models. Next, we introduce the concepts of deep learining and start playing with various fundamental architectures
using the PyTorch deep learning library. Commonly used machine learning methodologies like multi-layer perceptron, convolutional
layers, recurrent neural net, fully convolutional neural net (U-Net for image segmentation) are thoroughly covred in this
section.
III. Deeper dive into the theory of machine learning
Once we equip the students with the big picture we gradually move into the deeper and conceptually
harder theoretical descriptions of machine learning. The theory of machine learning covers statistical fundamentals
like non-parametric methods, convergence with perceptron, optimization with kernel methods and so on.
The cousre had 40% weightage on projects, 40% on written exams, and 20% on class participation. The studnets and the
instructor heavily relied on several edtech applications to make the course successful in the middle of a raging
pandemic when remote collaboration was the only option possible. Despite all odds the course happened to be a resounding
success and students’ feedback remained very encouraging for the instructor.
