L01 - MACHINE LEARNING BASICS

What is Machine Learning?

Machine Learning is the science “concerned with the question of how to construct computer programs that automatically improve with experience”

What is the goal at Machine Learning

Getting computers to “program” themselves
Automating automation
Let the data do the work

What is the schematic for traditional programming and machine learning?

What is the Big Data?

Big Data describess data that has a big volume, variety, velocity.

It is typically too big to gain a manual overview of the data and the information that might be contained.

What is Data Mining?

Data Mining is the study of collecting, cleaning, processing, analyzing, and gaining useful insights from data.

What is the Smart Data?

Smart Data is data that is extracted from larger datasets and in particular represents useful information that can be used to solve a specific task.

What are the Machine Learning Paradigms

Supervised Learning

Unsupervised Learning (Semi-Supervised Learning)

Reinforcement Learning

What is the supervised learning?
What is the goal of supervised learning?

The available data consists of labelled examples, each data point containing features and an associated labe.

The goal of supervised learning algorithms is learning a function that maps feature vectors (inputs) to label (output), based on example input-output pairs.

What is unsupervised learning?
What is the goal of unsupervised learning?

The available data consists of unlabeled examples, meaning that each data point contains features only, without an associated label.

The goal of unsupervised learning is learning useful patterns or structural properties of the data.

What is reinforcement learning?

Reinforcement learning is an area of machine learning concerned with how intelligent agents take actions in an environment in order to maximize a cumulative reward.

What are The Task categories and High level tasks?

What are the three Interpretability of Different Approaches?

White-box:

• It is possible to examine every detail inside the box.

• The mathematical functions and parameters of the white-box model full rely on domain knowledge

• It is possible to follow the process from data input to output in every detail

Grey-box

• The content is partially visible

• The mathematical functions rely on self-learning algorithms but are optimized by additional domain knowledge

• It is partially possible to follow the process from data input to output

Black-box

• The content inside is unkown.

• The mathematical functions and parameters of a black-box model rely on self-learning algorithms

• It is not possible to follow the process from data input to output

What are suitable approaches to apply ML to solve a problem?

1. CRISP-DM

2. DMME

3. CRISP-ML (Q)

What is the data mining life cycle for CRISP-DM?

Cross-Industry Standart Process for Data Mining

1. Define project goals and business objectives

2. Understand the available data and their quality

3. Filter and select useful and relevant data

4. Create data-models that might meet the defined goals

5. Evaluate models’ performance related to the goals

6. Set the best model into operation