Artificial Intelligence (AI) GLOW Class quizz

In machine learning, the condition in which the model is too simple to describe the data or to learn the structure of the data is called?

• Underfitting

• Overfitting

• Unfitting

• Bad fitting

The main difference between supervised learning and unsupervised learning is that supervised learning requires little data to conduct training, while unsupervised learning requires a lot of data.

• True

• False

When the target variable that we are going to predict has a discrete value, the method in machine learning is called

• Discretisation

• Classification

• Supervision

• Regression

The key difference between Logistic Regression and Adaline rule is that (choose the right answer below):

• Logistic regression updates the weights based on a linear activation function rather than a unit step function like in the Adaline

• Logistic regression updates the weights based on a sigmoid function rather than a unit step function like in the Adaline

• Logistic regression updates the weights based on a sigmoid function rather than an identity function like in the Adaline

• Logistic regression updates the weights based on an identity function rather than an a sigmoid function like in the Adaline

One of the key ingredients of supervised machine learning algorithms is a defined objective function that will be used during learning process. What is the objective function that we want to maximize in Logistic Regression:

• Sum of squared errors

• Mean absolute errors

• The logit function

• The log-likelihood

The correct sequence of the gradient descent (GD) algorithm in terms of the amount of training data processed for each iteration, from the smallest to the largest is:

• Batch GD, Mini-Batch GD, Stochastic Mini-Batch GD

• Stochastic GD, Batch GD, Mini-Batch GD

• Stochastic GD, Mini-Batch GD, Batch GD

• Mini-Batch GD, Batch GD, Stochastic GD

The stochastic gradient descent method has a higher probability of reaching the global minimum compared to batch gradient descent.

• True

• False

Many machine learning algorithms require some sorts of feature scaling for optimal performance, such as standardization. Standardization shifts the mean of each feature so that it is centered at 1 (one) and each feature has a standard deviation of 1 (unit variance)

• True

• False

A very computationally efficient approach to compute the partial derivatives of a complex cost function in multilayer neural networks is called

• Backpropagation

• Forward propagation

• Efficient propagation

• Complex propagation

A network which has more than one hidden layer is called:

• Shallow neural networks

• Many hidden layer networks

• Deep neural networks

• Deep hidden networks