AI23531 Deep learning

What is the main function of an activation function in a neural network?

• To shuffle weights randomly

• To introduce non-linearity into the network

• To reduce training time

• To control learning rate

What are the three parts of a biological neuron?

• Axon, Cortex, Synapse

• Dendrites, Cell Body, Axon

• Soma, Nucleus, Synapse

• Input, Output, Activation

What type of network is described as having no cycles or loops?

• Recurrent Neural Network

• Convolutional Neural Network

• Feedforward Neural Network

• Feedback Neural Network

What is the mathematical formula inside a perceptron?

• y = w1x1 + w2x2 + ... + b

• y = w + x

• y = ReLU(w*x)

• y = input / weight

What happens during the “forward pass” in training?

• Weights are initialized

• Loss is computed

• Data flows through the network to produce output

• Gradients are propagated

Which of the following best describes a basic Perceptron?

• Multi-class classifier with hidden layers

• A sequence-processing RNN

• A binary classifier with no hidden layer

• A deep learning model for image recognition

What is the key limitation of a single-layer perceptron?

• Inability to handle linearly separable data

• Too many weights

• Cannot learn non-linear patterns

• Requires GPUs to run

Which function is typically used as an activation function in the output layer of a classifier?

• ReLu

• Tanh

• Sigmoid

• Softmax

In a feedforward neural network, how does information flow?

• Output to Input

• Randomly between layers

• Input → Hidden → Output

• Hidden → Input → Output

What is the primary goal of backpropagation?

• To make the network deeper

• To make the network deeper

• To make the network deeper

• To update weights based on error gradients