Reinforcement Learning and Deep RL Python Theory and Projects - DNN Why Activation Function Is Required Exercise

The video tutorial discusses the structure of a neural network where all layers except the output layer have linear activations. The output layer contains a single neuron with a sigmoid activation function. The tutorial explores whether such a model can still be considered a neural network, given the linear activations in the hidden layers. It also covers the variability in the number of neurons across different hidden layers.

What is unique about the output layer in the described neural network?

• It has multiple neurons.

• It has only one neuron.

• It uses a linear activation function.

• It is the first layer in the network.

How do the hidden layers in the network differ from each other?

• They are all connected directly to the output layer.

• They all have the same number of neurons.

• They can have different numbers of neurons.

• They all use sigmoid activation functions.

What type of activation function is used in the hidden layers?

• Linear

• Tanh

• ReLU

• Sigmoid

What activation function is used in the output layer of the network?

• Sigmoid

• Tanh

• Linear

• ReLU

What is the main question posed about the neural network model?

• How many layers it should have.

• What type of data it can process.

• Whether it can classify images.

• If it can be considered a neural network with linear activations.