AL503 (B) - Deep Learning

Explain the historical evolution of deep learning and McCulloch-Pitts neuron model. Discuss the representation power of multilayer perceptrons.

Explain backpropagation algorithm with suitable mathematical formulation. Discuss weight initialization and batch normalization.

Explain sigmoid neurons and feed-forward neural networks.

Explain PCA and SVD and their role in representation learning.

Compare gradient descent variants : GD, SGD and Momentum.

Explain autoencoders and discuss regularization techniques used in autoencoders.

Explain denoising, sparse, contractive and variational autoencoders.

Explain relationship of autoencoders with PCA and SVD and role of dataset augmentation.

Explain convolution operation, kernels, stride, padding and pooling in CNNs.

Explain architectures of VGGNet and ResNet.

Explain regularization techniques in deep networks such as dropout, weight decay and early stopping.

Explain visualization of CNNs, Deep Dream and Deep Art.

Explain BPTT, vanishing and exploding gradient problems and truncated BPTT.

Explain LSTM and GRU architectures and their role in solving vanishing gradient problem.

Explain RBMs, Gibbs sampling and deep belief networks.

Explain GANs and applications of deep learning in vision, speech and medical domains.