IITM BS Introduction to DL and GenAI (BSDA2001): Syllabus and Tips

Introduction to DL and GenAI (BSDA2001) is a 4 credit diploma level course in the IITM BS Data Science programme. It has no prerequisite, but Machine Learning Practice (BSCS2008) is a corequisite. That means you take MLP before it or in the same term. Together with its 2 credit project, it forms Option 2 of the Diploma in Data Science.

The course is taught by Prof. Balaji Srinivasan and Prof. Ganapathy Krishnamurthi of the Wadhwani School of AI, IIT Madras. They also run the matching project course, BSDA2001P. The course page and the handbook agree on the credits and the MLP corequisite.

What you learn in Introduction to DL and GenAI

Most topics get one theory week followed by one hands-on week. The 12 weeks fall into four blocks.

• Weeks 1 and 2: neural network basics. What an artificial neuron is, how layers stack up, and how activation and loss functions work. Then you build small networks in TensorFlow and Keras and try different activations and optimisers.
• Weeks 3 to 6: images and sequences. Convolution and pooling layers, then image classification on datasets such as MNIST and CIFAR-10. After that, recurrent networks and LSTMs, used for sentiment analysis, simple text generation and time series prediction.
• Weeks 7 to 9: generating images. Autoencoders, variational autoencoders and GANs. Then diffusion models, how they are trained and used, and a practice week generating images with GANs and pretrained diffusion models.
• Weeks 10 to 12: large language models. Word embeddings, tokenisation and attention. Then encoder models like BERT, decoder models for text generation, machine translation and fine-tuning methods such as LoRA. The last week is on prompting techniques.

How the course is assessed

The course page points to the standard course structure on the programme's Academics page. It does not list a separate assessment plan for this course.

Where it counts: Option 1 or Option 2

The Diploma in Data Science has 21 fixed credits. For the other 6 credits you choose one option, and you must finish both parts of it.

The choice has a knock-on effect. If you finish Option 1, you cannot take the Option 2 courses later at the degree level. If you finish Option 2, you can still take the Option 1 courses later as degree level electives. This rule applies to current diploma students and to batches from September 2025. See the Diploma in Data Science.

This course is not part of any minor. The degree level minors in AI use separate courses.

Who finds it hard and how to prepare

The pace is high. You meet CNNs, LSTMs, GANs, diffusion models and transformers in 12 weeks. Students who fall behind in one block find the next one harder, since each practice week assumes the theory week before it.

• Get comfortable with Python, NumPy and basic model training before week 1. Machine Learning Practice helps here.
• Plan for GPU access early. A free cloud notebook with a GPU can handle small datasets like MNIST and CIFAR-10.
• In each practice week, change one thing at a time (layer size, learning rate, activation) and note what happens. This builds real understanding faster than rerunning given code.
• Keep one notebook per architecture with your own comments. They become revision notes for the end term.

What to take before and after

Take Machine Learning Practice (BSCS2008) before or with this course. The DL and GenAI project (BSDA2001P) lists this course as a corequisite, so plan both for the same term or take the project later. At the degree level, Deep Learning (BSCS3004) is a mandatory core course for everyone, whichever option you chose.

Common questions

Does this course replace Deep Learning at the BSc level?

No. Deep Learning (BSCS3004) is one of the mandatory courses when you enter the BSc level. This diploma course is a separate introductory course.

Can I take this course without the project?

Not if you want it to complete the diploma. The handbook says you must finish one full option, both the theory course and its project. The project also lists this course as a corequisite.

Which framework does the course use?

The week 2 practice session names TensorFlow and Keras. The project course page mentions PyTorch and Hugging Face, so if you take the project too, expect to use both.

Is Option 2 better than Option 1?

Neither is better for everyone. Option 2 gives you an early start on deep learning and keeps more doors open. If you pick it, Business Analytics can still be a degree level elective later. If you pick Option 1, the DL and GenAI courses are closed to you later.