7.3 Unsupervised Learning Algorithms and Other Models

Question 1

List three unsupervised machine learning algorithms.

Answer 1

1. Principal component analysis (PCA)
2. Hierarchical clustering
3. K-means clustering

Question 2

PCA stands for what?

Answer 2

Principal component analysis.

Question 3

PCA summarizes the information in a large number of ___________ factors and into a much small set of _________ factors.

Answer 3

correlated;

uncorrelated;

Question 4

Give the term that describes the following: The uncorrelated factors in a PCA which are linear combinations of the original features.

Answer 4

Eigenvectors;

Question 5

Within a PCA, each ________ has an ________. An _________ is the proportion of total variance in the data set explained by the _________.

Answer 5

eigenvalue;

eigenvector;

Question 6

Within a PCA, explain how eigenvectors are applied.

Answer 6

Eigenvectors are essentially (my interpretation) percentage weightings that are applied to each independent variable to change what otherwise would be a set of correlated independent variables (e.g., interest rates, stock prices; etc.) into a set of uncorrelated independent variables.

Question 7

Within a PCA, the __________ with the highest __________ is the independent variable that is the most important to the model.

Answer 7

eigenvector;

eigenvalue;

Question 8

PCA is totally a ____ ___ where you are completely dependent upon the computer to make predictions. This may work fine on __-_____ data. Maybe not so great for ___-__-_____ data.

Answer 8

black box;
in-sample;
out-of-sample;

Question 9

__-____ clustering requires that you be familiar enough with the data to know how many _______ you should have.

Answer 9

K-means;

clusters;

Question 10

K-means clustering partitions observations into “__” ________ clusters.

Answer 10

k;

nonoverlapping;

Question 11

______ _______ is an ________ unsupervised algorithm used to form a _________ of clusters. Each ____________ is its own cluster.

Answer 11

Hierarchical clustering;
iterative;
hierarchy;
observation;

Question 12

In an ____________ (or ______-__) clustering algorithm, you start with one observation as its own cluster and add other similar observations to that group.

Answer 12

agglomerative;

bottom-up;

Question 13

In a _________ (or ___-____) clustering algorithm, you start with one giant cluster and then partition that cluster into smaller and smaller clusters.

Answer 13

divisive;

top-down;

Question 14

A ________ _________ consists of nodes connected by links.

Answer 14

neural network;

Question 15

List and describe the three layers of a neural network.

Answer 15

1. Input layer (contains nodes with values for the features, i.e., the independent variables);
2. Hidden layers (contains nodes to which the multiple nodes within the input layer connect to … can be multiple hidden layers);
3. Output layers;

Question 16

in a neural network, learning happens in the ______ layer nodes.

Answer 16

hidden;

Question 17

Is a neural network able to handle nonlinearities?

Answer 17

Yes.

Question 18

What does DLN stand for?

Answer 18

Deep learning network;

Question 19

DLNs are _______ ________ networks.

Answer 19

unsupervised neural;

Question 20

Are neural networks supervised or unsupervised?

Answer 20

supervised;

Question 21

In terms of the layers, what distinguishes a neural network from a deep learning network?

Answer 21

A neural network has only a few hidden layers, whereas a DLN has many hidden layers (often more than 20).

Question 22

What types of complex activities are DLNs used for?

Answer 22

Pattern, image and speech recognition.

Question 23

DLNs are the basis of the ______ ________ revolution.

Answer 23

artificial intelligence;

Question 24

A ______ ______ algorithm attempts to maximize rewards over time within the constraints of its environment.

Answer 24

reinforcement learning (RL);

Question 25

Does reinforcement learning have direct labeled data?

Answer 25

No.

Question 26

An RL algorithm learns by testing ___ ______ and reusing its ______ _________. Millions of trial and error repetitions.

Answer 26

new actions;

previous experiences;

Question 27

What is the term that describes the process that seeks to reduce noise by discarding those attributes that contain little information? Is it a supervised or an unsupervised learning algorithm?

Answer 27

Dimension reduction;

unsupervised;