Module 1 - What is AI? - Machine Learning

Machine Learning Tom Mitchell Def'n:

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

Within Unsupervised Learning type of ML, there exists:

Clustering and Dimensionality Reduction

Main Types of Machine Learning:

Unsupervised Learning, Supervised Learning, Reinforcement Learning

Clustering:

grouping subsets of data in a dataset based off a similar attributes

Classifying Images:

provide machine with set of images of cats / dogs and use that model to identify other cats & dogs in other images

Machine Learning is one way to achieve ___

AI, hence the subset relationship

Within Supervised Learning type of ML, there exists:

Regression and Classification

Machine Learning Def'n:

Subset of AI technique which uses statistical methods to enable machines to improve with experience -- note that AI only Mimics the behaviour, whereas ML Improves with experience

Machine Learning AI4ALL Def'n:

Subset of AI, consists of techniques that enable computers to LEARN from the data and deliver AI applications

ML Ex: SPAM Filters

Task: Classify emails as spam or not as spam Experience: Analyzing emails & watching us label spam Performance: Percentage of emails correctly classified

Traditional Programming VS ML

Traditional Programming: Hardcode Rules & Decisions into the program using conditional structures ML: Provide dataset & desired outcomes, and ML creates a model to process other data

Linear Regression model types:

Underfit - under-represents the trend of the data Optimal - optimally represents the trends of the data Overfit - over-represents the trend of the data