Scribd
Upload
57 views

Slides Basics Whatisml

This document introduces machine learning, artificial intelligence, and related concepts. It defines machine learning as a field that uses algorithms to learn from data and improve performance on some task without being explicitly programmed. Deep learning is described as a subset of machine learning that uses neural networks, which were inspired by the human brain. The document also discusses the relationships between machine learning, artificial intelligence, deep learning, and statistics. It outlines the main types of machine learning problems: supervised learning, unsupervised learning, and reinforcement learning.

Uploaded by

Chaitanya Parale
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
57 views

Slides Basics Whatisml

This document introduces machine learning, artificial intelligence, and related concepts. It defines machine learning as a field that uses algorithms to learn from data and improve performance on some task without being explicitly programmed. Deep learning is described as a subset of machine learning that uses neural networks, which were inspired by the human brain. The document also discusses the relationships between machine learning, artificial intelligence, deep learning, and statistics. It outlines the main types of machine learning problems: supervised learning, unsupervised learning, and reinforcement learning.

Uploaded by

Chaitanya Parale
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

Introduction to Machine Learning

ML-Basics: What is Machine Learning?

Learning goals
Understand basic terminology of
and connections between ML,
AI, DL and statistics
Know the main directions of ML:
Supervised, Unsupervised and
Reinforcement Learning
Image via www.vpnsrus.com
MACHINE LEARNING IS CHANGING OUR WORLD

Search engines learn what you want


Recommender systems learn your taste in books, music, movies,...
Algorithms do automatic stock trading
Google Translate learns how to translate text
Siri learns to understand speech
DeepMind beats humans at Go
Cars drive themselves
Smart-watches monitor your health
Election campaigns use algorithmically targeted ads to influence
voters
Data-driven discoveries are made in physics, biology, genetics,
astronomy, chemistry, neurology,...
...


c Introduction to Machine Learning – 1 / 9
THE WORLD OF ARTIFICIAL INTELLIGENCE
... and the connections to Machine Learning and Deep Learning

Artificial Machine Deep


Intelligence Learning Learning

Many people are confused what these terms actually mean.

And what does all this have to do with statistics?


c Introduction to Machine Learning – 2 / 9
ARTIFICIAL INTELLIGENCE

AI is a general term for a very large and rapidly developing field.


There is no strict definition of AI, but it’s often used when machines
are trained to perform on tasks which until that time could only be
solved by humans or are very difficult and assumed to require
"intelligence".
AI started in the 1940s - when the computer was invented.
Scientists like Turing and John von Neumann immediately asked
the question: If we can formalize computation, can we use
computation to formalize "thinking"?
AI includes machine learning, natural language processing,
computer vision, robotics, planning, search, game playing,
intelligent agents, and much more.
Nowadays, AI is a "hype" term that many people use when they
should probably say: ML or ... basic data analysis.


c Introduction to Machine Learning – 3 / 9
MACHINE LEARNING

Mathematically well-defined and


solves reasonably narrow tasks.
ML algorithms usually construct
predictive/decision models from
data, instead of explicitly
programming them.
A computer program is said to
learn from experience E with
respect to some task T and
some performance measure P, if
its performance on T, as
measured by P, improves with
experience E.
Tom Mitchell, Carnegie Mellon
University, 1998


c Introduction to Machine Learning – 4 / 9
DEEP LEARNING

DL is a subfield of ML which studies neural networks.


Artificial neural networks (ANNs) might have been (roughly)
inspired by the human brain, but they are simply a certain model
class of ML.
ANNs have been studied for decades. DL uses more layers,
specific neurons were invented for images and tensors and many
computational improvements allow training on large data.
DL can be used on tabular data, but typical applications are
images, texts or signals.
The last 10-15 years have produced remarkable results and
imitations of human ability, where the result looked intelligent.

"Any sufficiently advanced technology is indistinguishable from magic."


Arthur C. Clarke’s 3rd law


c Introduction to Machine Learning – 5 / 9
ML VS. STATS

ML and Statistics have historically been developed in different


fields, but many methods and especially the mathematical
foundations are equivalent.
Traditionally, models from ML focused more on precise predictions
whereas models from statistics focused more on the ability to
interpret the patterns that generated the data and the ability to
derive sound inference.
Nowadays, ML and predictive modelling in statistics basically work
on the same problems with the same tools.
Unfortunately, the communities are still divided, don’t talk to each
other as much as they should and everyone is confused due to
different terminology for the same concepts.
Most parts of ML we could also call:
Nonparametric statistics plus efficient numerical optimization.


c Introduction to Machine Learning – 6 / 9
UNSUPERVISED LEARNING
Data without labels y
Search for patterns within the inputs x
Unsupervised as there is no “true” output we can optimize against

2.5 ●●●
Dimensionality reduction (PCA,
●●
●●
●●●●● ●
●● ●
Autoencoders ...);
●
●●
●● ●
2.0 ●●
●
●
●
●● ●
●● compress information in X
●
●● ●●●●●
●● Clustering: group similar
Petal.Width

●● ● ●
1.5 ●●●●●●
●
● ●●●● ● observations
●● ●
●●
●●
●
●
●●●●
● ● ●
1.0 ●●●●● Outlier detection, anomaly
● detection
0.5 ●
●●
●●● Association rules
●
●●
●●
●●●
●●
●●●
0.0
●●●
2 4 6
Petal.Length


c Introduction to Machine Learning – 7 / 9
REINFORCEMENT LEARNING
RL is a general-purpose framework for AI. At each time step an agent
interacts with environment. It: observes state; receives reward;
executes action.

Goal: Select actions to maximize future reward.


Reward signals may be sparse, noisy and delayed.


c Introduction to Machine Learning – 8 / 9
WHAT COMES NEXT

We will deal with supervised learning for regression and


classification: predicting labels y based on features x, using
patterns that we learned from labeled data.
First, we will go through fundamental concepts in supervised ML:
What kind of "data" do we learn from?
How can we formalize the goal of learning?
What is a "prediction model"?
How can we quantify "predictive performance"?
What is a "learning algorithm"
How can we operationalize learning?
We will also look at a couple of fairly simple ML models to obtain a
basic understanding.
More complex stuff comes later.


c Introduction to Machine Learning – 9 / 9

You might also like