All

Question 1

How do you define machine learning? (Formal)

Answer 1

A computer program that is said to learn from experience E, with respect to some class of tasks T and performance measure P, if its performance in tasks in T, as measured by P, improves with experience E.

ETP, the experience of the task of ET’s performance

Question 2

How do you define Machine Learning (Informal)

Answer 2

A field of study that gives computers ability to learn without being explicitly programmed.

Question 3

ETP mean?

Answer 3

Experience, Task, Performance

Question 4

What does E mean in ETP?

Answer 4

The experience of a task that the machine learns from.

Question 5

What does T in ETP mean?

Answer 5

The task the machine is doing.

Question 6

What does P in ETP mean?

Answer 6

The performance measured by the task T.

Question 7

What is an example of ETP?

Answer 7

Learning how to use a radio, playing chess or checkers.

Question 8

What are the two common types of machine learning?

Answer 8

Supervised Learning and Unsupervised Learning

Question 9

What does supervised learning mean?

Answer 9

Supervised Learning is where the machine gets the correct answers and learns from it.

Question 10

What does unsupervised learning mean?

Answer 10

Unsupervised learning is where the machine makes structure by clustering the data based on the variables in the data without any knowledge what the results look like.

Question 11

Is there feedback based on the prediction of the results in unsupervised learning?

Answer 11

No.

Question 12

What is clustering?

Answer 12

Grouping things onto groups based on different variables.

Question 13

What is non-clustering?

Answer 13

Finding a structure in a chaotic environment.

Question 14

What does the notation y_i mean?

Answer 14

Output or target features

Question 15

What does the notation x_i mean?

Answer 15

Input features

Question 16

What does the notation (x_i, y_i) mean?

Answer 16

The Training example

Question 17

What does the notation m mean?

Answer 17

The Training set

Question 18

What does the notation Y mean?

Answer 18

The space of the output values.

Question 19

What does h mean?

Answer 19

The hypothesis function.

Question 20

What is a cost function?

Answer 20

Used to measure the accuracy of the hypothesis function. It also takes the average difference of all the results of the hypothesis.

Question 21

What is the idea of the gradient descent?

Answer 21

To minimize the thetas so that the hypothesis function’s results can be closer to y for a training example.

Question 22

What is the formula of a mean squared error cost function?

Answer 22

.

Question 23

What is the gradient descent?

Answer 23

An algorithm used to simultaneously update the thetas, on which is used to improve the hypothesis.

Question 24

What is the algorithm of gradient descent?

Answer 24

T means Theta in this example.
a means alpha

repeat until convergance{
T_j:=T_j-a*d(cost)/dT_j (for j=0 and j=1)
}

Question 25

What is alpha?

Answer 25

The alpha is the learning rate of the for gradient descent.

Question 26

What happens when the alpha is too high?

Answer 26

The gradient descent can overshoot the minimum and may fail to converge, or even diverge.

Question 27

What happens if the alpha is too low?

Answer 27

The gradient descent will be incredibly slow.

Question 28

How do you correctly simultaneously update the thetas?

Answer 28

Assign the new values of the thetas into temporary variables first then re-assign the back using those temporary variables in the end.

Question 29

What are derivatives?

Answer 29

A description of how a function changes with small or multiple variables.

Question 30

What is the idea of the hypothesis function?

Answer 30

To give the predicted output from the given input feature.

Question 31

Cost must go down?

Answer 31

Yes

Question 32

What is a batch?

Answer 32

Each step of gradient decent uses all the training examples

Question 33

Are there different types of gradient descent?

Answer 33

Yes

Question 34

Can gradient descent converge if alpha is fixed? (As long as alpha is not too high or low)

Answer 34

Yes

Question 35

Do you have to decrease alpha overtime?

Answer 35

No.

Question 36

Can gradient descent find the global minimum of any function?

Answer 36

No.

Question 37

Can any linear regression function find the local optima?

Answer 37

No, it can only find the global optima.

Question 38

What does optima mean?

Answer 38

The relative best solutions.

Question 39

What is regression?

Answer 39

Where the machine predicts a continuous output, Like predicting a house price based on multiple variables.