Download Machine Learning: Artificial Intelligence isn`t just a Science Fiction

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Machine Learning: Artificial
Intelligence isn't just a Science
Fiction topic
Raul Garreta - Tryolabs / MonkeyLearn
My Credentials
● Computer Science Engineer from Udelar,
Msc in Machine Learning + NLP
● Co-Founder, CTO & Product Manager at
Tryolabs.
● Co-Founder at MonkeyLearn.
● Professor in ML at InCo, Udelar.
● Co-authored "Learning Scikit-learn:
Machine Learning in Python"
Contents
● Brief intro to AI & Machine Learning (ML)
● ML Applications
● Cloud ML tools
What is AI?
From a behavioral point of view, is an artificial
agent that shows certain characteristics of
intelligence like:
●
●
●
●
●
Reasoning
Knowledge representation
Learning
Planning
Perception
What is AI?
Behavioral test = Turing Test
If I write an enough complex Ifthen-else structure, could it
pass the test?
Random behavior?
Different fields within AI
Artificial Intelligence
● General Artificial Intelligence
● Expert Systems
○
○
○
○
Natural Language Processing
Computer Vision
Machine Learning
...
Machine Learning
Algorithms that allow computers
to automatically learn to perform
a task from data.
Can improve their performance
over time, by adding more data.
Machine Learning Definitions
Arthur Samuel (1959): "Field of study that gives computers
the ability to learn without being explicitly programmed"
Tom Mitchell (1997): "A computer program is said to learn
if its performance at a task T, as measured by a
performance P, improves with experience E"
Machine Learning Algorithms
● Learn to associate a particular input (set of
features) to a particular output (class,
number or group of instances)
● That is the process of training a ML model.
● And use the learned model to predict the
outcome on new instances
Inputs: Instances
Usually we have instances of data that
represent objects: documents, images, users,
etc.
And can be represented by a set of features:
● A document is represented by a set of words.
● An image is represented by a set of pixels.
● A user can be represented by the age, level of
education, gender, interests, etc.
Machine Learning Problems
Classification: assign a label (class)
to a set of items.
Regression: assign a number
(evaluation) to a set of items
Clustering: group items into clusters
according to a similarity measure
Type of Machine Learning
Algorithms
Linear Models
Decision Trees
Type of Machine Learning
Algorithms
Probabilistic /
Statistical Models
Neural Networks /
Deep Learning
Important Concepts in ML
Besides the Machine Learning…
● Data gathering / importation
● Data preprocessing
● Feature extraction
● Feature selection
● Performance evaluation (testing)
Applications
Natural Language Processing
Text Mining
Speech to Text
Applications:
Computer Vision
Face Recognition
OCR
Applications
Data Mining / Predictive Analytics
Recommendation Engines
Medicine
Applications
Intelligent Agents
Robotics
Game Players
Why use Machine Learning?
● Solve problems that manually would be extremely
difficult or impossible.
● Make predictions.
● Automatically process huge amounts of information and
sources: big data.
● Intelligent apps => improve UX => improve conversion
rates => $$$
● Great companies use it...
Why use a Cloud Saas ML platform?
● Avoid to deploy and maintain the full stack.
● Be cross platform.
● Not all programming languages have ML
tools.
● ML requires huge amounts of computer
power.
● Just solve it: good, fast, easy.
Machine Learning Platforms
As with other problems (eg: payments,
communications) is a trend to go SaaS.
Machine Learning
Microsoft Azure ML
● http://azure.microsoft.com/enus/services/machine-learning/
● Launched preview version on June 2014.
● Cloud based ML platform to build predictive
numerical applications.
● Technologies used in Xbox and Bing.
Machine Learning
Microsoft Azure ML
●
●
●
●
●
Easy to scale, Azure infrastructure.
Users can build custom R modules.
GUI and APIs.
More oriented to Data Scientists.
Pricing: pay as you go.
Machine Learning
MonkeyLearn
● http://monkeylearn.com/
● Launched private alpha on April 2014
● Cloud based, focused on Text Mining:
extract and classify information from text.
MonkeyLearn
● Easy to use.
● Pre-trained modules for different
applications.
● GUI and APIs.
● More oriented to developers.
● Pricing: freemium, pay as you go.
Conclusions
● Machine Learning can allow
us to make intelligent apps.
● It's a trendy topic…
● New ML platforms are
emerging, allowing any
developer to incorporate ML
technologies.