Notebook 9

Question 1

What are some advantages of Decision trees?

Answer 1

Question 2

What are some disadvantages of Decision trees?

Answer 2

Question 3

How do we apply the Decision Tree classifier to our iris dataset?

Answer 3

Question 4

How to you show a Decision Tree plot? How would you change this to only create a Tree of max depth of 2?

Answer 4

clf = tree.DecisionTreeClassifier(max_depth=2)

Question 5

How do we interpret this decision tree plot?

Answer 5

Question 6

What does the Emsemble method do?

Answer 6

Question 7

What are the import statements need for Random Forest and Gradient Boosted Trees?

Answer 7

Question 8

Use a Random Forest Classifier on this dataset?

Answer 8

GO thourgh hyperparameters on Data base

Question 9

Use a Gradient Boosting on this dataset?

Answer 9

Question 10

How do you code this?

Answer 10

Question 11

How do we apply the Random Forest Classifer to this imbalanced data set?

Answer 11

If this worked ocrrectly you should find an AUC of about 0.84

The RandomForestClassifer has a parameter class_weight. The “balanced” mode uses the values of y to automatically adjust weights inversely propotional to class frquencies in the input data –> setting class_weight = “balanced” should improve the the AUC to about 0.87 for this artifical data set

Question 12

What is a Multi-layer Perceptron?

Answer 12

Question 13

What are the advatanges of a multi-layer Perceptron?

Answer 13

Question 14

What are the disadvantages of a multi-layer Perceptron?

Answer 14

Question 15

What does MLPClassifier do?

Answer 15

Question 16

How does the MLPClassifier train the algorithm?

Answer 16

Question 17

What does the MLPClassifier allow for?

Answer 17

Question 18

What does the MLPRegressor do?

Answer 18

Question 19

What parameter for both the MLPRegressor and MLPClassifer use for regularisation?

Answer 19

Question 20

What arhe three algorithms MLP uses to train?

Answer 20

Question 21

What is the time complexity of bakcwards propogation?

Answer 21

Question 22

What are some tips on using MLP in practise?

Answer 22

Question 23

What are all the import statements required for using MLPClassifer?

Answer 23

Question 24

Answer 24

Question 25

Note that the dataset has 100 samples by default. This is a good size for our purposes. How does the complexity of backpropagation depend on the sample size?

Answer 25

ADD TO THIS

Question 26

Answer 26

Question 27

Why do we not perform fit as well on test data?

Answer 27

Its because of overfitting but what a more robust answer

Question 28

What does the MinMaxScaler do?

Answer 28

Question 29

Answer 29

Why does the training data differ?

Question 30

What does QuantileTranformer do?

Answer 30

Question 31

Answer 31

Question 32

What is import to do before proceeding tuning the hyperparameters of MLP?

Answer 32

Question 33

Sanity Check for MLP using SVC?

Answer 33

Question 34

Answer 34

Question 35

Use the MLPClassifier?

Answer 35

Note: give the Classifier score for moons using the StandardScaler is 0.95 we cna conclude that is the best

Question 36

What are some of the most important Hyperparameters for the MLP?

Answer 36

Question 37

What are the different choices of activation hyperparameters for the MLPClassifier??

Answer 37

* model get worse with higher alpha, and learning_rate_init.

Question 38

What are the different solvers hyperparameters for the MLPClassifier?

Answer 38

Note: The default solver ‘adam’ works pretty well on relatively large datasets (with thousands of training samples or more) in terms of both training time and validation score. For small datasets, however, ‘lbfgs’ can converge faster and perform better.

Question 39

For the hidden_layer_sizes hyperparameter what do you need to do to screate a single layer with 10 neurons? What about 3 layers of 10 neurons?

Answer 39

Question 40

Perform hyperparameter tuning using a loop for the the number of neurons in a single hidden layer? print a graph of the results and decide on what hte best parameter is?

Answer 40

Question 41

Perform hyperparameter tuning using a loop for the the number of neurons from 2-40 for 2 hidden layers? print a graph of the results and decide on what hte best parameter is?

Answer 41

Question 42

Answer 42

Question 43

Answer 43

Question 44

How do we plot the gradient descent over time of this small samples?

Answer 44

Question 45

What is the output of this data? How do interpret it?

Answer 45

Question 46

In supervised learning, what is importance to note about the features?

Answer 46

Question 47

What is the feature_importances_property? How and when do we use it?

Answer 47

Question 48

Apply the RandomForestClassifier to the Iris dataset using feature_importances_property?

Answer 48

Question 49

What should the importances sum to?

Answer 49

Question 50

What is Permutation Importance?

Answer 50

Question 51

Apply the RandomForestClassifier to the Iris dataset using permutation_importance? What does the perm_importance sum to? How can interpret the graph?

Answer 51