Course

Syllabus

Overview of AI, ML, and DL, Multi-layer perceptrons, Activation functions (ReLU, Sigmoid, Tanh), Loss functions and cost, Performance analysis of Classifier and Regression model. Essential Data Pre-processing for Deep Learning.

Backpropagation and gradient descent, Stochastic Gradient Descent (SGD), RMSProp, Adam. Weight initialization. Overfitting and underfitting. Regularization techniques: regularization, dropout, Batch normalization. Deep Neural Networks.

Convolution layers, pooling layers, CNN architectures. 1D CNNs for signal processing, Transfer learning and fine-tuning pre-trained models, Visualization techniques: saliency maps, Grad-CAM for explaining CNN decisions, Application of CNN.

Comparison between static and sequential data, Sequence modeling, RNN architecture, Limitations of RNNs: vanishing/exploding gradients, short memory retention, Internal structure of LSTM, Gated Recurrent Unit (GRU) Networks, Comparison in terms of performance and complexity, Advanced architectures: GAN, Transformers.

Pre-requisite: Nil

Course Objectives

• To impart foundational knowledge of deep learning architectures, data preprocessing, and model optimization techniques applicable to electrical engineering problems.
• To enable students to design, implement, and evaluate CNN and RNN-based models for real-world applications.

Course Outcomes

CO1: Understand the core concepts of AI, ML, and DL along with key neural network components.

CO2: Design and evaluate deep neural networks using various optimization strategies and regularization techniques to improve training performance and generalization.

CO3: Implement convolutional neural networks (CNNs) for visual and signal-based tasks, and utilize transfer learning and model interpretability methods.

CO4: Develop sequence models using RNN, LSTM, and GRU architectures for time-series forecasting and event detection in real life problems.

CO5: Explore and apply advanced architectures such as GANs and Transformers for complex generative and attention-based modeling tasks.

CO-PO Mapping

1. Goodfellow, Y, Bengio, A. Courville, “Deep Learning”, MIT Press, 2016.
2. S. Haykin, “Neural Networks and Learning Machines”, 3rd Edition, Pearson, 2008
3. Aditi Majumder, M. Gopi, Introduction to Visual Computing: Core Concepts in Computer Vision, Graphics, and Image Processing, CRC Press; 1 edition, 2018
4. Francois Chollet, “Deep Learning with Python”, 2nd Edition, Manning Publications, 2021.