Flashcards in Intro Material Deck (15):

1

## Describe fully vs partially observable

### Fully if at any given time you can make the optimal decision based on all the information. Partial (poker) is only partially observable. It utilizes history decisions to make more informed decision.

2

## Describe benign vs adversatiral environment

### Benign would be like the weather. Adversarial is like a chess game where the objective of the environment is to win against your ai program

3

## Describe the four sets of characteristics for an artificial intelligence environment

###
Partially vs Fully observable.

Continuous vs discrete

stochastic vs deterministic

benign vs adversarial

4

## What are three processes of deliberation that are important to AI

### reasoning, learning, memory

5

## What is Bayes Rule and its components?

###
P(A|B) = ( P(A) * P(B|A) ) / P(B), WHERE

P(A|B) is called the posterior

P(B|A) is called the likelihood

P(A) is called the prior

P(B) is called the marginal likelihood

P(B) expands to P(B) = SUM(a) ( P(B|A=a)*P(A=a)

6

## Explain Bayes Rule Graphically reasoned

###
A is not observable, B is observable.

A causes B. A -> B

Utilizes diagnostic reasoning to determine A given b, A given not be. It's an inverse of what we are able to observe with B given A, or B given not A

7

## Describe deductive reasoning

### Deductive reasoning works from the more general to the more specific. Sometimes this is informally called a "top-down" approach. We might begin with thinking up a theory about our topic of interest. We then narrow that down into more specific hypotheses that we can test. We narrow down even further when we collect observations to address the hypotheses. This ultimately leads us to be able to test the hypotheses with specific data -- a confirmation (or not) of our original theories.

8

## Describe inductive reasoning

### Inductive reasoning works the other way, moving from specific observations to broader generalizations and theories. Informally, we sometimes call this a "bottom up" approach In inductive reasoning, we begin with specific observations and measures, begin to detect patterns and regularities, formulate some tentative hypotheses that we can explore, and finally end up developing some general conclusions or theories.

9

## What's a short way to describe supervised vs unsupervised?

### Supervised is about functional approximation, unsupervised is about deriving a description of the data.

10

## What is reinforcement learning?

### Learning through delayed rewarding systems. Think of tic tac toe. You get your verification information after the game has completed. Supervised you get a label correct or not right after inputing into the function.

11

## What is the common thread between supervised, reinforcement and unsupervised?

### Optimization. Supervised - labels data well, Reinforcement - behavior scores well, Unsupervised - clusters scores well

12

##
Describe Machine Learning as an overview answering the following 5 questions:

1) What

2) What From

3) What For

4) How

5) Outputs

###
1) Parameters, structure, hidden concepts

2) Supervised, Unsupervised, Reinforcement

3) Prediction, Diagnostics, Summarization

4) Passive, Active, Online, Offline

5) Classification, Regression

13

## What is cross validation?

### Split training data into three buckets. Train (~80%), Cross Validation (~10% to find optimized parameters), Test (~10%)

14

## Describe the "Kernel Trick"

### It's mapping a feature space to a higher dimension in order to turn a non-linearly seperable space into seperable. Think of a circle of x surrounded by greater circle of - classified elements. Create new x3 = sqrt(x1^2 + x2^2)

15