• Attempt any five questions.
• All questions carry equal marks.
Discuss the GPU implementation of randomized SVD and write its applications.
What is the representation power of a multilayer network of sigmoid neurons? Explain in detail.
What is batch normalization? Explain how does it work and also write its advantages.
When should auto encoders be used instead of PCA/SVD for dimensionality reduction? Justify.
Explain in brief sparse auto-encoder and contractive auto-encoder.
Write a short note on regularization of auto encoders.
Discuss in detail about ReLU activation function.
Illustrate the CNN terms Padding and Pooling.
What is unit pruning? Discuss the need of unit pruning in deep learning.
Explain briefly about directed graphical model.
Explain briefly deep recurrent neural networks and its architecture with a neat diagram.
Discuss in detail Long Short Term Memory (LSTM) and write the advantages of LSTM.
NADE
MADE
Representation learning.
Feed forward neural network
Weight decay and also the benefits of using weight decay
Deep belief networks
