ES 667: Deep Learning

IIT Gandhinagar · Semester I, 2026–27

ES 667 develops deep-learning models from their statistical and computational foundations. The course begins with likelihood, loss functions, and multilayer networks, and proceeds to convolutional models, sequence models, Transformers, representation learning, and generative models.

Current class

Generalization & Regularization

We will move from the train–validation gap to weight decay, early stopping, dropout, data augmentation, Mixup, and label smoothing.

References

There is no required textbook. Useful references include:

  • Simon J. D. Prince, Understanding Deep Learning (online edition)
  • Christopher M. Bishop and Hugh Bishop, Deep Learning: Foundations and Concepts
  • Aston Zhang, Zachary C. Lipton, Mu Li, and Alexander J. Smola, Dive into Deep Learning (online edition)