1. Suppose you are planning to train a standard recurrent neural network model on a dataset of 100,000 text documents, each exactly 50 words long. You choose to represent each word in a document using a pre-trained, 300 dimensional word embedding. Which of the following feature input array shapes would be CORRECT for input into a PyTorch recurrent module?

A. (vocab_size, 100000, 300)

B. (50, 100000, 300)

C. (100000, 50, 300)

D. (100000, 15000)

2. Which of these is a CORRECT description of how a recurrent neural network (RNN) layers output can be mapped to a single prediction output?

A. The RNN layers final memory state vector is summed to return the final prediction

B. The RNN layer returns a sequence of processed values that are then averaged to return the final prediction

C. Its not possible to map RNN output to a single prediction output

D. The RNN layers final memory state vector is used as flat input into fully connected layers that terminate in the final prediction

3. Which of the following is a CORRECT description of how transfer learning is applied in text processing recurrent neural networks (RNN)?

A. Pre-trained RNN layer weights are used for processing character sequences before custom weights are used to make final predictions

B. Pre-trained word embeddings are used as vectorized representations of words, which are passed in a sequence as input to custom trained RNN layer(s)

C. Pre-trained tf-idf embeddings are used as vectorized representations of documents, which are passed as a flat input to custom trained RNN layers(s)

D. Transfer learning cannot be applied in text processing recurrent neural networks

Please Please I need Quick answers ! Please