Training

Question 1

Training

What is Labeled data?

Answer 1

Input data to train a model with a reliable description of what it is (truth).

Question 2

Training

Highest clasifications of ML algorithms?

Answer 2

Supervised learning, unsupervised learning, and reinforcement learning

Question 3

Training

How does Supervised Learning work?

Answer 3

Aalgorithms are trained on labeled data

Question 4

Training

What is the goal of supervised learning?

Answer 4

Learn a mapping function that can predict the output for new, unseen input data

Question 5

Training

How does unsupervised learning work?

Answer 5

Trained on unlabeled data

Question 6

Training

Goal of unsupervised learning?

Answer 6

Discover inherent patterns, structures, or relationships within the input data

Question 7

Reinforcement Learning

What is reinforcement learning?

Answer 7

Semi-supervised learning. Algorithm given rewards or penalties for its actions, and the machine learns from this feedback to improve its decision-making over time.

Question 8

Reinforcement Learning

What data is used for Semi-Supervised learning?

Answer 8

Mostly labeled, some unlabeled

Question 9

Reinforcement Learning

What is Reinforcement Learning usually used for?

Answer 9

Teach AI to play games, robotics to navigate and manipulate objects

Question 10

Reinforcement Learning

Example of Reinforcement Learning in healthcare?

Answer 10

Optimize treatment plans

Question 11

Reinforcement Learning

Example of Reinforcement Learning in finance?

Answer 11

Trading strategies

Question 12

Reinforcement Learning

Are images and videos structured or unstructured data?

Answer 12

Unstructured

Question 13

RLHF

What is RLHF?

Answer 13

Reinforcement Learning from Human Feedback

Question 14

RLHF

Why bother with RLHF?

Answer 14

Better align with human values

Question 15

RLHF

How do you do RLHF?

Answer 15

Separate reward model: ask humans which of two generated answers sound more human. Use reward model to tune the real model

Question 16

Regression and Classification

What is the goal of a regression?

Answer 16

Predict an output value given input value

Question 17

Regression and Classification

What happens in Regression Training?

Answer 17

Fit a curve so you can predict new values

Question 18

Regression and Classification

What is the goal of classification?

Answer 18

Understand how things group together

Question 19

Regression and Classification

What happens in Classification training?

Answer 19

Look for clumps of values that are similar

Question 20

Regression and Classification

Example of regression training and inference?

Answer 20

Input tons of humans height and weight. Curve-fit the data. Answer how much someone weighs from their height

Question 21

Regression and Classification

Example of a Classification training and inference?

Answer 21

Input tons of height and weight of cats, dogs, and giraffes, ask animal given height and/or weight

Question 22

Regression and Classification

Classification or Regression for a continuous set of inputs?

Answer 22

Regression (numeric)

Question 23

Regression and Classification

Classification or Regression for discrete set of inputs?

Answer 23

Classification

Question 24

Regression and Classification

Is k-NN used with Classification or Regression?

Answer 24

Classification

Question 25

# *Regression and Classification* What is Training vs. Validation vs. Test Set for input data?

Answer 25

80% to train the model, 10% to validate the model, 10% to test the model

Question 26

# *Regression and Classification* What happens in Validation and Testing above?

Answer 26

Validation is for tuning hyperparameters (tune the model), test is to gauge accuracy

Question 27

# *Feature Engineering* What is Feature Engineering?

Answer 27

Massage input data so it's meaningful for training

Question 28

# *Feature Engineering* Do you do Feature Engineering on structured or unstructured data?

Answer 28

Both!

Question 29

# *Feature Engineering* Example of Feature Engineering?

Answer 29

Change a date-of-birth column of data to Age in years

Question 30

# *Feature Engineering* Example of Feature Engineering for raw data from Zillow?

Answer 30

Create "price per square foot" as a new field: it's useful for an LLM and typical questions.

Question 31

# *Feature Engineering* How do you do Feature Engineering on unstructured text?

Answer 31

Maybe label it with sentiment

Question 32

# *Feature Engineering* How do you do Feature Engineering on unstructured images?

Answer 32

Find edges or textures