Neural Networks and Deep Learning Quiz

In case of pooling scans, for an input of size NxN and filters of size MxM and stride SxS the output size will be,

• 
• 
• 
• None of the above

Mark all that True

• BERT is essentially the encoder of a transformer model

• 1. GPT is essentially the encoder of a transformer model

• 1. BERT is essentially the decoder of a transformer model

• 1. GPT is essentially the decoder of a transformer model

• 1. Queries, keys, and values are obtained by splitting the input into 3 equal segments

GNN is mostly used when there is an/a?

• unstructured data

• structured data

• sequential data

• None of the above

The _________ produces hidden representations of the input word sequence

• a) generator

• b) encoder of the transformer

• c) discriminator

• d) decoder of the transformer

To account for some jitter in the first-level patterns, it is recommended to apply______.

• Activation function

• Drop out

• Pooling

• CNN layer

Generally, in case of CNN,

• A upsampling layer is followed by a CNN layer (i.e. a layer of "regular" neurons).

• A downsampling layer is preceded by a CNN layer (i.e. a layer of "regular" neurons).

• A upsampling layer is preceded by a CNN layer (i.e. a layer of "regular" neurons).

• A downsampling is followed by a CNN layer (i.e. a layer of "regular" neurons).

The _______ is used to evaluate the importance of the input at that time, for a given output.

• a) value

• b) key

• c) query

• d) attention

What is the minimum number of neurons needed of representing XOR function?

• 4

• 3

• 2

• 1

_____________learns to detect the most significant patterns in the signals

• a) The decoder of the autoencoder

• b) The encoder of the autoencoder

• c) The Generator of the GAN

• d) VAE

The following figure illustrate a perceptron which perform a:

• X or Y

• X xor Y

• X and Y

• None of the above