Font size
WorksheetsLesson 3 (Model Training)
Total questions: 20
Worksheet time: 10mins
Data wrangling is a one-time process
True
False
Ways to handle missing data
Drop missing columns
Replace with zero
Replace with mean
All of the above
Select the activity that is not a part of data wrangling
Restructuring
Grouping
Model Training
Normalization
________ helps to package data into consumable formats
Datastore
Datasets
Datasets create an in-memeory copy of the underlying data
False
True
Select the linear dimensionality reduction technique based on linear mathematical calculations
T-SNE
PCA
Feature embedding
All of the above
________ provides alert about data drift
Datastore
Dataset
Data monitor
Hyperparameters are not learnt from the data
True
False
Select the hyperparameters
Number of training epochs
Weights/coefficients
Number of hidden layers
Learning rate
Select the evaluation metric not used for classification
RMSE
Precision
F1 Score
Squared error
Algorithm which outputs real continuous outputs
Regression
Classification
Evaluation metrics for regression
Mean Squared Error(MSE)
Accuracy
Mean-Absolute-Error (MAE).
All of the above
The phenomenon in which an ML algorithm is not capable of coping with very large numbers of features
Features
Curse of dimensionality
Feature engineering
Feature selection
Potential benefits of feature engineering
Faster model training time
Smaller trained model size
More appropriate features for some algorithms
Improved model accuracy
Deep learning depends on feature engineering much more than classical machine learning
True
False
Example of covariate shift
A change in customer behavior over time.
Two features that used to be correlated are no longer correlated.
Upstream process change
A sensor breaks and starts providing inaccurate readings.
Validation data is
Data used to learn the values of the parameters.
Data used to tune the values of the hyperparameters.
Data used to check the performance of the final, fully trained model.
Classify music as belonging to multiple groups refers to
Multi-class multi-label classification
Multi-class single-label classification
In confusion matrices,the sum FP and TP tells us the number of actual positive cases in dataset
True
False
Which of the following helps reduce bias for models?
Stacking
Boosting
Bagging
