Machine Learning and AI (Class 21)

Reward with example (Reinforcement Learning)

Feedback from the environment, can be positive or negative ie: winning or losing in checkers

What is the Turing Test?

a test for intelligence in a computer, requiring that a human being should be unable to distinguish the machine from another human being when they interact with both without knowing which is human or machine.

Arthur Samuel Definition of Machine Learning (1959)

ML is field of study that gives computers the ability to learn without being explicitly programmed.

The difference between traditional programming and machine learning

Traditional programming takes data and a program, incorporates it in the computer, and creates outputs. Machine learning takes data and an output, incorporates it in a computer, and creates programs.

Actions with example (Reinforcement Learning)

What an agent can do in each state ie: Possible checker moves in a given state

4 Broad Concepts when Machine Learning is used

- When human expertise does not exist (Roaming Mars) - When humans can't explain their expertise (Speech Recognition) - When models can't be used (Personalized medicine) - When models are based on huge amounts of data (genomics)

Example of Unsupervised learning

K-means Clustering

Unsupervised Learning (Key feature, % Used)

Key Feature: Input is given (no output) Output: Hidden structures behind the X's ie: Clustering Output is unknown and no training data set is given Used 10-20% of machine learning

Reinforcement Learning (Sequence)

Learning best actions based on rewards/punishments Given a sequence of states and actions, with (delayed) rewards, output a policy Output: Policy

Policy (Reinforcement Learning)

The mapping from states to actions that tells you what to do in a given state.

What is Artificial Intelligence

The simulation of human intelligence by machines Coined by John McCarthy in 1956

How Machine Learning helps Marketing (6)

- Customer Experience/Support - Personalized customer care - Personalized advertising - Reduce consumer churn - Price optimization - Improved demand forecasting

Examples of Supervised Learning

- Linear Regression - Logistic Regression

Three Main Types of Machine Learning

1. Supervised Learning 2. Unsupervised Learning 3. Reinforcement Learning

What is Deep Learning

Deep learning is a subfield or extension of machine learning concerned with algorithms that imitate how the human brain operates with processing data and creating patterns for decision making. Deep learning only has one hidden layer between an input and an output to recognize patterns, while Deep Neural Networks have multiple hidden layers between them.

States with example (Reinforcement Learning)

Describes the current situation. ie: Position of checker pieces on a board

Supervised Learning (Key Feature, Types, % Used)

Key feature: Input (x) and output (y) data are both given Types -Regression: Y is a number, ie: X is size of home in ft and Y is the house price in $ -Classification: Y is a category, ie: X is the size of a tumor and Y is the type of tumor benign or malignant X can be multi-dimensional (you can have multiple inputs) Used for 70% of machine learning

How is Machine Learning related to Artificial Intelligence?

Machine Learning is a subset of Artificial Intelligence. In order for a machine to mimic human abilities, ML helps train the machine to use systematic methods to improve with experience and learn from hidden insights to help AI as a whole improve its ability to mimic human abilities.

Is Machine Learning new? Why are we hearing more about it today?

Machine Learning is not that new, it was first coined by Arthur Samuel in 1959 so it has been around for 60 years. It is becoming exponentially more prevalent today as technology has advanced and we now have an incredible amount of data collected. Machine Learning has made it possible to analyze all of this data and gain meaningful insights from it.

Tom Mitchell Definition of Machine Learning (1989)

Machine Learning is the study of algorithms that improve their performance 'P' at some task 'T' with experience 'E'. <P,T,E>