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

The video tutorial explores the concept of neural networks with linear activations and a single sigmoid-activated neuron in the last layer. It explains how such a structure collapses into a linear layer followed by a sigmoid, resembling logistic regression. The tutorial clarifies that logistic regression is not typically considered a neural network, despite its perceptron-like structure. It concludes that introducing non-linear activations in hidden layers transforms the model into a neural network.

What happens to a neural network with linear activations in all layers except the last one with a sigmoid activation?

• It becomes a support vector machine.

• It collapses into a single linear layer followed by a sigmoid, resembling logistic regression.

• It functions as a linear regression model.

• It becomes a deep neural network.

Why is logistic regression not typically called a neural network?

• Because it uses a different type of activation function.

• Because it lacks multiple layers with non-linear activations.

• Because it is only used for regression tasks.

• Because it is a type of decision tree.

What is the primary function of logistic regression?

• To reduce dimensionality.

• To predict continuous values.

• To classify data points.

• To perform clustering.

Under what condition does a model become a neural network?

• When it is trained using backpropagation.

• When it has a single layer with linear activation.

• When it includes non-linear activations in any of the hidden layers.

• When it uses a sigmoid activation in the output layer.

What is the role of non-linear activations in a neural network?

• They are used only in the output layer.

• They allow the network to perform linear transformations.

• They enable the network to model complex patterns.

• They simplify the network architecture.