IITM BS Machine Learning Practice (BSCS2008): Syllabus and Tips

Machine Learning Practice (BSCS2008), or MLP, is a 4 credit diploma level course in the IITM BS in Data Science. It is a fixed course in the Diploma in Data Science. You must complete both Machine Learning Foundations (BSCS2004) and Machine Learning Techniques (BSCS2007) before you register. MLP is the hands-on partner of MLT: you solve real problems with scikit-learn, the popular Python ML library.

The course page names Ashish Tendulkar as the instructor. He is listed as a Research Software Engineer at Google AI.

What you learn

• Weeks 1 and 2: the full workflow. Week 1 walks through an end-to-end ML project in scikit-learn. The course page lists week 2 as "Graph Theory (VOL 3)", which does not match the rest of the plan. Check the course portal for what week 2 really covers.
• Weeks 3 to 5: regression and its tools. Linear regression, batch and stochastic gradient descent, polynomial regression, regularised models, and logistic regression.
• Weeks 6 to 8: classification. Binary and multiclass classifiers in scikit-learn, then support vector machines.
• Weeks 9 to 12: bigger models. Two weeks on decision trees, ensemble learning and random forests, then neural network models in scikit-learn, and finally unsupervised learning.

The course page also lists wider goals: understand the life cycle of an ML project and the tools for each step, use libraries such as scikit-learn and tensorflow, tune models with regularisation and feature selection, and decide next steps from how a model is evaluated.

How it is assessed

The course page lists the standard pattern: 12 weeks of coursework, weekly online assignments, 2 in-person invigilated quizzes and 1 in-person invigilated end term exam. Check your grading document for the weights and for any programming exam.

The MLP project (BSCS2008P) is a separate 2 credit course with MLP as its corequisite. Its page says it runs on Kaggle for 12 to 15 weeks in milestones, and ends with a viva by an industry expert. See the MLP project guide.

Where it counts

• Diploma in Data Science. MLP (4 credits) and the MLP project (2 credits) are both fixed parts. See the Diploma in Data Science.
• Deep Learning and GenAI option. The handbook lists MLP as a corequisite of Introduction to Deep Learning and Generative AI (BSDA2001), one of the two diploma options.
• Re-entry. The handbook lists MLT and MLP as final exams for a Diploma in Data Science holder who rejoins at the Diploma in Programming level.
• Electronic Systems students. The ES handbook lists MLP (CS2008, 4 credits) as an open elective and in the ES Minor in Machine Learning.

Mismatch between sources. The DS course page lists MLF and MLT as prerequisites of MLP. The ES handbook's table lists them as corequisites instead. DS students should follow the course page. ES students should confirm with support.

Who finds it hard

MLP is less about new theory and more about getting things to work. Expect to spend time on cleaning data, choosing the right metric, and deciding what to try next when a score will not improve. If MLT felt shaky, you may run models without knowing why one fails. Go back to your MLT notes whenever a method comes up.

How to prepare

1. Get comfortable with pandas and NumPy. Loading data, handling missing values and selecting columns take a lot of time in practice.
2. Build one clean template. Split data into training and test sets, set up a pipeline, run cross-validation, and report a score. Reuse and improve this template every week.
3. Never tune on the test set. Use cross-validation to compare models. Look at the test score only at the end.
4. Question accuracy. If 95 out of 100 rows belong to one class, a model that always predicts that class scores 95% accuracy and is useless. Check precision, recall or a confusion matrix too.
5. Keep a results log. For each experiment, note what you changed and the score. This makes the project viva much easier.

What to take before or after

Finish Machine Learning Techniques (BSCS2007) and MLF first. Run the MLP project with MLP or right after it. If you pick Option 2 in the diploma, Introduction to Deep Learning and GenAI (BSDA2001) can run alongside MLP. For how all three ML courses link, read MLF, MLT and MLP explained.

Common questions

Can I take MLP and the MLP project in the same term?

Yes. The project lists MLP as a corequisite, so you can take them together or do MLP first. Doing them together lets you apply each week's methods in the Kaggle milestones.

Does MLP teach deep learning?

Only a little. Week 11 covers neural network models in scikit-learn, and the course page mentions tensorflow among the libraries. Deep learning proper comes later, in the diploma option or at the degree level.

Is MLP offered every term?

Yes. The DS handbook says all diploma level courses are offered every term. The project runs for 12 to 15 weeks, so check its dates when you register.