04 Training Models

Question 1

In an equation y = b1x1 + b2x2 +c what does ML calculate

Answer 1

In an equation x1, x2 values are there in the data(columns) and b1 and b2 are calculated by the model

Question 2

name order of equations in linear regression

Answer 2

Normal Equation followed by linear regression model

Question 3

Why is normal equation used

Answer 3

to calculate cost function

Question 4

What is the main component in linear regression model

Answer 4

theta is the main component. we need to find the value of theta where RMSE value is less.

Question 5

Drawback of normal equation

Answer 5

It is very slow on large dataset with many features.

Question 6

Methods used to run Linear regression model with large dataset

Answer 6

1. Gradient Descent
2. Batch Gradient Descent
3. Stochastic Gradient Descent
4. Mini - Batch Gradient Descent

Question 7

What is gradient descent

Answer 7

It is an algorithm to find optimal solution to a complex problems.
it measures local gradient of error function to the theta.

Question 8

what is the relation of learning rate and theta

Answer 8

if the learning rate is too low than it will take more time to reach to the theta value and if the learning rate is too high than it will cross the optimal theta value.

Question 9

Difference between Gradient Descent and Batch Gradient Descent method

Answer 9

in the Gradient Descent the change in the cost function and parameter is calculated at each step and in the batch Gradient Descent all the change is calculated using the entire training data and in single step.

Question 10

What is Stochastic Gradient Descent Method

Answer 10

it uses an instance (randomly selected) from the data to calculate the optimal theta value.

Question 11

Advantage of using Stochastic Gradient Descent Method

Answer 11

1.Very quick on large dataset.
2. effective when the data has multiple local minima’s.

Question 12

what each iteration in linear regression called

Answer 12

epoch

Question 13

what is learning schedule

Answer 13

function to determine learning rate.

Question 14

What is mini batch stochastic Gradient Descent

Answer 14

it is the mixture of stochastic and batch Gradient Descent

Question 15

Do we require scaling for any of the Gradient Descent method

Answer 15

yes

Question 16

Performance of Gradient Descent models on large datasets

Answer 16

Normal eq - fast;
BGD - Slow;
SGD - Fast
Mini-batch GD - Fast

Question 17

Performance of Gradient Descent models with many features

Answer 17

Normal eq - slow;
BGD - Fast;
SGD - Fast
Mini-batch GD - Fast

Question 18

which logistic regression model have 0 hyper parameter

Answer 18

normal equation(LinearRegression)

Question 19

How can a normal equation be used to solve a polynomial equation?

Answer 19

the feature with x^2 will be added as a new feature.

Question 20

what is overfitting

Answer 20

the model is performing well training data and is not performing well on the validation set.

Question 21

what is underfitting?

Answer 21

the model is not performing well on both training and validation set.

Question 22

how to deal with an underfitting model?

Answer 22

we need to add more features or choose a complex model.

Question 23

how to deal with an overfitting model?

Answer 23

we need to add more training date.

Question 24

name 3 types of model errors

Answer 24

1.Bias;
2.Variance;
3.Irreducible error

Question 25

what is Bias model error and how can we recognize it?

Answer 25

It is due to wrong assumptions, i.e. we think it is normal equation while it is quadratic equation. In this case the model underperformance.

Question 26

what is Variance model error and how can we recognize it?

Answer 26

It is because the model will be sensitive to even slight change.
In this case the model will overfit.

Question 27

What is regularization?

Answer 27

It is constraining the model.
it is a way to control overfitting in a model.
It is achieved by applying weights.

Question 28

what are different types of regularization?

Answer 28

1.Ridge Regression
2.Lasso Regression
3.Elastic Net

Question 29

What is Ridge Regression?

Answer 29

a regularization term is added to the algorithm forcing the algorithm to fit the data and keep the weights as small as possible.

Question 30

Is scaling necessary for Ridge Regression?

Answer 30

Yes standardization is necessary.

Question 31

which hyperparameter in regularization model controls the regularization?

Answer 31

alpha

Question 32

which method to use for ridge regression?

Answer 32

Ridge()

Question 33

which method to use for ridge regression in sgd?

Answer 33

SGDRegressor(penalty = “l2”)

Question 34

what is full form of Lasso Regression

Answer 34

Least Absolute Shrinkage and Selection Operator.

Question 35

how does lasso regression work?

Answer 35

it adds weights to feature, it adds weight 0 for the least important feature.

Question 36

how is ridge regression different from Lasso regression

Answer 36

ridge regression it adds weight & Lasso regression it adds weight 0 to the least important features.

Question 37

which method to use for Lasso regression?

Answer 37

Lasso()

Question 38

which method to use for Lasso regression in sgd?

Answer 38

SGDRegressor(penalty = “l1”)

Question 39

What is Elastic Net?

Answer 39

It is a middle ground between Lasso and Ridge regression

Question 40

which hyperparameter is used to control the ratio of lasso and ridge regression

Answer 40

l1_ratio - where if it is close to 0 it will be ridge regression and close to 1 will be lasso regression.

Question 41

Which model among linear regression, ridge & lasso is better

Answer 41

ridge & lasso are better as they have regularization.

Question 42

if I have a data with high correlated dataset which model to use between Lasso and Ridge.

Answer 42

Ridge is better to start with high correlated dataset.

Question 43

What is Early Stopping?

Answer 43

Another way to regularize a model is to use early stopping. where the model stops training as soon as its validation error reaches its minimum.

Question 44

Can Logistic Regression Model be used for classification and regression?

Answer 44

Yes

Question 45

What is Logistic Regression?

Answer 45

Like Linear Regression it calculates probability for each instance and based on the probability it classifies into 0’s or 1’s. on this Linear Regression a sigmoid function which gives result in 0’s and 1’s.

Question 46

What is Decision Boundaries?

Answer 46

it is a boundary between class 0 and 1 which will allow to differentiate between both classes.