Learning from data

Understanding Artificial Intelligence

Maarten Van den Broeck

Senior Content Developer at DataCamp

AI functions and areas involved

AI areas and AI functions

AI functions and areas involved

AI areas and AI functions

Enter Machine Learning (ML)

Machine Learning: learn from data and identify patterns

Areas of Machine Learning

Enter Machine Learning (ML)

Machine Learning: learn from data and identify patterns

Areas of Machine Learning

Enter Machine Learning (ML)

Machine Learning: learn from data and identify patterns

Areas of Machine Learning

Enter Machine Learning (ML)

Machine Learning: learn from data and identify patterns

Areas of Machine Learning

Enter Machine Learning (ML)

Machine Learning: learn from data and identify patterns

Areas of Machine Learning

Supervised Learning: classification

Classification: assign each data observation the category (class) it may belong to

Binary classification: two classes, e.g. positive/negative, male/female, etc.

Binary classification

Supervised Learning: classification

Classification: assign each data observation the category (class) it may belong to

Binary classification: two classes, e.g. positive/negative, male/female, etc.
Multi-class classification: several mutually exclusive classes, e.g. multiple species

Supervised learning: Data annotation (getting labelled observations with known class a priori) needed to learn/train a model capable of making inference

Multi-class classification

Supervised Learning: regression and forecasting

Regression: assign each data observation a numerical output or label based on its inputs

Regression to estimate a house price

Time series forecasting: predict future values of variable, based on its past behavior

Time series forecasting to predict daily bus passengers

Unsupervised and reinforcement learning

Clustering: find subgroups of data with similar characteristics (e.g. k-means algorithm)

Clustering penguins data

Anomaly detection: detecting abnormal data observations e.g. unusual card transactions

Anomaly detection

Association rule discovery: find common co-occurrences of items in transaction data

Discovering products frequently bought together

Reinforcement learning: learn by experience (trial and error) to master a complex task

Reinforcement learning to navigate a maze

How about Deep Learning?

Highly sophisticated models based on deep neural networks: solve very challenging tasks where classical ML models become limited
Learn from data as a human brain would do

Need a lot of data to learn: sometimes millions of observations

Deep Learning tasks

How about Deep Learning?

Highly sophisticated models based on deep neural networks: solve very challenging tasks where classical ML models become limited

Need a lot of data to learn: sometimes millions of observations

Deep Learning tasks

Let's practice!

Understanding Artificial Intelligence