Machine Learning QUIZ

Question: How are supervised learning models’ algorithms described?

Objective

Regression

Independent

Classification

Subjective

Clustering

Ans:-

Regression

Classification

Question: What are some of the common unsupervised learning techniques?

K-means clustering

Logistic regression

Hierarchical clustering

Linear regression

Principal component analysis (PCA)

Decision tree classification

Ans:-

K-means clustering

Hierarchical clustering

Principal component analysis (PCA)

Question: Select the term that describes the offset between the line of best fit and a sample point.

Taxi-cab distance

Altitude

Residual distance

Euclidean distance

Ans- Residual distance

Question: Select the statement which describes a predictor.

Qualitative predictors imply a discrete model

Qualitative predictors imply a continuous model

Predictors are also called independent variables

Quantitative predictors imply a continuous model

Predictors are also called dependent variables

Quantitative predictors imply a discrete model

Ans:-

Qualitative predictors imply a discrete model

Predictors are also called independent variables

Quantitative predictors imply a continuous model

Question: Select the answers that describe some characteristics of K-means clustering.

Center of a cluster is called a circloid

Supervised learning method

Unsupervised learning method

Center of a cluster is called a centroid

The ‘k’ in k-means represents the number of elements in the data set

Ans:-

Unsupervised learning method

Center of a cluster is called a centroid

Question: Given a homogeneous data set, how do we best determine if it is clusterable?

Visually inspect the data plot for patterns

Perform k-means and measure its completeness score

Perform k-means to see if it succeeds in clustering the data

Perform support vector machine method before clustering

Ans:- Perform k-means and measure its completeness score

Question: We often gain performance and insight into significant data features or base elements when carrying out PCA. What might we lose when carrying out PCA?

Predictive features

Minimal variability

Unbiasness

Significant digits

Information complexity

Dimensionality

Ans:-

Predictive features

Information complexity

Dimensionality

Question: Match the cause of error with its effect on a model.

Answer Options:
A:Too few features
B:Too many features
C:Model too simple
D:Model too complex

Underfitting

A

B

C

D

Ans:- A,C

Overfitting

A

B

C

D

Ans:- B, D

Question: In the context of Neural Networks, which of these statements correctly describe a fully connected layer?

A. Every neuron in this layer takes its input from all the neurons in the previous layer

Output of every neuron in this layer is fed as input to exactly one neuron in the next layer

Output of every neuron in this layer is fed as input to a specific neuron in the next layer

Every neuron in this layer takes its input from exactly one neuron in the previous layer

Ans:- A. Every neuron in this layer takes its input from all the neurons in the previous layer

Question: What would be the activation function of a Neural Network that is made to perform linear regression on input data?

identity

logit

ReLU

tanh

Ans:- identity

Question: Match following statements about neurons in the context of machine learning with their correct Boolean values.


Answer Options:
A:A neuron can output multiple different values
B:A neuron can consist of only one function, a linear one
C:Every connection between two neurons has a weight W associated with it
D:A neuron is a mathematical function that can take multiple inputs and outputs a single value

False

A

B

C

D
Ans:- A,B
True

A

B

C

D
Ans:- C,D

Question: Match the following Neural Networks terms with their corresponding definition

Answer Options:
A:Gradient Descent Optimization
B:Epoch
C:Learning Rate
D:Batch Size

The iterative process of adjusting the model parameters to minimize the loss

A

B

C

D
Ans:- A

Number of data points fed into the Neural Network in a single iteration

A

B

C

D
Ans:-D

A single pass by the Neural Network through the entire dataset

A

B

C

D
Ans:- B

The size of the steps taken by the algorithm towards the optimal point during each pass

A

B

C

D
Ans:- C

Question: Consider the line y = x and the points [1,2], [ 2, 1], [ 3,4]. What would be the sum of the squared errors of this line with respect to these 3 points?

3

1

1.7

0

Ans:- 3

Question: In a dataset containing a feature x and effect Y, the goal of simple linear regressions is to find which of the following?

A. The values of W and b of the line Y = Wx + b whose distance from the input data points is minimum

The values of W and b of the line Y = Wx + b whose distance from the input data points is maximum

The general equation of the curve that passes through all of the input data points

The number of groups the input points can be clustered into based on some underlying logical similarity

Ans:- A. The values of W and b of the line Y = Wx + b whose distance from the input data points is minimum

Question: Match the following options correctly with the corresponding regression concept.

Answer Options:
A:The revenue of a hedge fund firm increased because the entire market went up
B:The revenue of a hedge fund firm increased because they began to use better techniques to make the right investments
C:The amount of money earned by a tennis player in a year increased due to the increase in prize money for all tennis tournaments
D:The crop yield of a farmer increased this year because she used better fertilizers and specialized techniques

Beta

A

B

C

D
Ans:- A,C

Alpha

A

B

C

D
Ans:- B,D

Question: Among the following, which future value CANNOT be predicted using linear regression on past data?

Number of iPhones that will be sold

Next number that will be output by a random number generator

A. Netflix’s stock price

Average temperature of the Earth

Ans:- Next number that will be output by a random number generator

Question: Which of the following statements about classification algorithms used to label input datasets are true?

The algorithm works by predicting exactly the label associated with an input feature with 100% confidence

The algorithm works by assigning probabilities to every label and the label with highest probability is the predicted value

A. Classification algorithms can be used to predict if your incoming mail is spam or ham

Classification algorithms can be used to predict the future values of stock prices

Ans:-

The algorithm works by assigning probabilities to every label and the label with highest probability is the predicted value

A. Classification algorithms can be used to predict if your incoming mail is spam or ham

Question: Match the following stages of the functioning of a Neural Network with the corresponding pass during which it happens.

Answer Options:
A:Training data fed as input to the first layer
B:Gradients are calculated to update model parameters
C:Current model parameters are used to make a prediction
D:Loss is calculated by comparing prediction with the actual labels

Forward Pass

A

B

C

D
Ans:- A,C,D

Backward Pass

A

B

C

D
Ans:- B

Question: Consider the line y = 3x and the points [1,0], [ 2, 1], [ 3,2]. What would the vector representation of the residuals of this line with respect to these three points?

[1, 1, 1]

[ 1, 2, 3]

[ 3, 5, 7]

[-3, -5, -7]

Ans:- [ 3, 5, 7]

Question: Match the description with its effect.

Answer Options:
A:Low bias, high variance
B:Low variance, high bias

Underfitting

A

B

Ans:- B

Overfitting

A

B

Ans:-A

Question: Model errors are often a result of overfitting or underfitting. Select the methods used to minimize fitting errors.

Decision Tree Classifier

Support Vector Machine (SVM)

K-means

Principal Component Analysis (PCA)

Cross Validation

Ans:-

Principal Component Analysis (PCA)

Cross Validation

Question: Once you have instantiated your Neural Network Model, we use the compile function to tie our Keras API to a specific backend. Which of these parameters is NOT specified in the compile function?

The metrics that would like to obtain when using this model for prediction

The number of neurons that will be used in the instantiated layer

The optimizer that we shall be using in the instantiated model

The loss function that we shall be using for the instantiated model

Ans:- The number of neurons that will be used in the instantiated layer

Question: Let’s say you’ve created a pandas dataframe called “data” and you have called the pandas .sample(frac=1) function on this dataframe. What does this dataframe return?

It returns a dataframe that contains the first half of the records of “data” in the same order

It returns a dataframe that contains the first half of the records of “data” in shuffled order

It returns a dataframe containing a random record from “data”

It returns a dataframe that contains all the records of “data” but in shuffled order

Ans:- It returns a dataframe that contains all the records of “data” but in shuffled order

Question: Match the following methods/attributes of the Scikit-learn linear_model.LinearRegression class correctly with what they return.

Answer Options:
A:predict()
B:coef_()
C:intercept_()

Captures the y-intercept of the linear model

A

B

C
Ans:- C

Captures the slope of the linear regression model

A

B

C
Ans:- B

Captures all the predictions made by the linear regression model

A

B

C
Ans:- A

Question: Match the layers of an artificial neural network with the task it performs.

Answer Options:
A:Input layer
B:Hidden layer
C:Output layer

Weight adjusting

A

B

C

Ans:- A

Error analysis and back-propagation

A

B

C

Ans:- C

Question: Define the architecture for a Keras sequential model and initialize it.

It is used to specify the number of features of the input and output of the current layer

It is used to specify what kind of layer we want to add to our Neural Network i.e. dense, activation etc.

It is used to specify the initial values of the weights and biases in the current layer

It is used specify what kind of non-linear component you want to apply to your neuron

Ans:- It is used specify what kind of non-linear component you want to apply to your neuron

Question: Identify which of the following are metrics available to evaluate models in Scikit Learn.

mean_squared_error

linear_model

r2_score

datasets

Ans:-

mean_squared_error

Activation or logistic function

A

B

C

Ans:- B

Question: What is a support vector?

Two-dimensional dot product describing magnitude

Vectors that support back-propagation

Vectors that describe a plane separating features

Vectors that support decision trees

Ans:- Vectors that describe a plane separating features

Question: What type of data does a logistic regression allow us to work with?

Unstructured Data

Discrete Data

Semi-structured Data

Continuous Data

Ans:- Discrete Data

Question: Select examples of data that can be enumerated as dummy variables.

Income Amount

Red, Green, Blue, Brown

Female, Male

True, False

Temperature

Ans:-

Red, Green, Blue, Brown

Female, Male

True, False

Question: Select the answers that describe the Naive Bayes Classifier.

Works only in unsupervised learning

Works well with supervised learning

Sensitive to non-correlating independent variables

Works only with discrete data

Assumes predictors contribute independently

Works only with large training sets

Ans:-

Works well with supervised learning

Assumes predictors contribute independently

Question: Select the answers that describe some characteristics of decision tree classifiers.

Constructed by hand using expert systems

Can be used for discrete categories and real values

Can only be used to solve regression problems

Data is partitioned to minimize entropy

Susceptible to overfitting with large numbers of features

Can only be used to solve classification problems

Can only be used in unsupervised clustering methods

Ans:-

Can be used for discrete categories and real values

Data is partitioned to minimize entropy

Susceptible to overfitting with large numbers of features