Deep Learning

Syllabus

Module No.	Detailed Content	Hrs (45)	CO
1	Introduction to Deep LearningRelationship between AI, ML, and DL, neural networks: deep vs. shallow networks, basic terminologies of deep learning.	05	CO1
2	Regularization TechniquesBias variance tradeoff, L1, L2 regularization, early stopping, dataset augmentation, parameter sharing, injecting noise into input, ensemble methods, dropout, greedy layer wise pre-training, advanced activation functions, batch normalization.	09	CO2
3	Convolution Neural NetworksIntroduction, architecture-motivation, layers, kernels, operations (convolution, padding, stride, pooling), non-linear layer, stacking layers, popular CNN architectures: LeNet, AlexNet, VGGNet.	07	CO3
4	Recurrent Neural NetworksFundamentals of RNN, bidirectional RNNs, encoder-decoder architectures, Gated Recurrent Unit (GRU), Recursive Neural Networks, Long Short Term Memory Networks (LSTM).	09	CO4
5	Autoencoders, Autoencoder architecture, regularized autoencoder, denoising autoencoders, representational power: effect on layer size, and depth, stochastic encoders and decoders, contractive encoders.	09	CO5
6	Applications of Deep Learning, ImageNet Detection and classification, Audio generation using WaveNet, Natural Language Processing with Word2Vec, Bio-Informatics Face Recognition, Scene Understanding and sematic segmentation, Automated Image Captioning.	06	CO6