Rappel: sequence models

Fixed input size

Lack of memory

Order invariance

Parallel processing capability

Designed to handle sequences of variable length

Does not have memory

Captures temporal dependencies

Used only for image processing

Learning method for RNN

Dimensionality reduction technique

Approach for image processing

Classification algorithm

Vanishing gradients

Exploding gradients

Both of the above

None of the above

Detection of past information in the cell state via the forget gate

Selection of relevant long-term information through the input gate

Adding the selected information to the cell state

Detection of important short-term information in the cell state

Generation of the new hidden state through the output gate

Decide which new information will be stored.

Eliminate irrelevant information.

Regulate the network with the tanh function.

Produce short-term memory.

It decides which information should be retained.

It multiplies the long-term memory by a forget vector.

It uses the sigmoid function.

It produces a new version of the long-term memory.

Produce short-term memory.

Take the current input and the previous short-term memory.

Pass long-term memory through an activation function.

Decide which information should be retained.

A variant of the RNN architecture.

A type of long-term memory.

A mechanism for controlling the flow of information.

A method for network regulation.