What does the term 'hyperparameters' refer to in neural network training?

• Parameters that are learned during training

• The input values provided to the network

• Settings like the number of layers and nodes, which are set before training

• Super parameters that provide the best performance

• NN weights are hyperparameters as they are the most important in NNs

Why might one choose to use a neural network instead of linear regression for a given dataset?

• Neural networks are always faster to train

• Neural networks can model complex, non-linear relationships better than linear regression

• Neural networks require less data

• Linear regression cannot be used for regression tasks

In the context of neural networks, what does 'convergence' refer to?

• The point where the weights are randomly set

• The moment when the learning rate is maximized

• When the predicted output matches the actual output closely enough, stopping further updates

• When the number of hidden layers is increased

Which of the following is NOT a typical application of deep learning?

• Self-Driving Cars

• News Aggregation

• Visual Recognition

• Simple Arithmetic Calculations

Which of the following best describes the Gradient Descent algorithm?

• A method to increase the error between actual and predicted values

• An iterative process to update weights to minimize the error

• A technique to initialize weights randomly

• A process to normalize input data

Given a non-linear regression problem , how can we use a neural network to solve it ?

• by adding more data

• use hyperparameters to reduce overfitting

• increase number of Hidden nodes

• increase number of layers

• Use a sigmoid function to provide non-linearity

What is the primary function of a neuron in an artificial neural network?

• To store data

• To receive input and compute an output based on weighted sums

• To transmit data between layers

• To execute conditional statements

In a single-layer perceptron, what is the role of the bias term (b)?

• It multiplies the input directly

• It acts as a threshold for activation

• It adds a constant value to the weighted sum

• It adjusts the learning rate

What key concept allows deep learning models to perform tasks like language translation and image recognition effectively?

• Single-layer perceptrons

• Hand-coded feature extraction

• Multiple layers (depth) in the network

• High learning rates

What is the primary difference between Stochastic Gradient Descent (SGD) and Batch Gradient Descent (BGD)?

• SGD updates weights after each instance, while BGD updates weights after all instances in the dataset

• SGD uses multiple layers, whereas BGD uses a single layer

• SGD is used for regression tasks, while BGD is used for classification tasks

• There is no significant difference between SGD and BGD