Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

Machine Learning - an introduction

No description
by

Benjamin Lindberg

on 3 February 2016

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Machine Learning - an introduction

Dog, Ball, Snow
LabeledPictures
Sound to text
Internet using patterns
Machine Learning
Main machine learning algorithms:

- Supervised learning
- Unsupervised leaning
Summary
Definition
Working with machine learning
Parts in a machine learning algorithm
big DATA
Representation: ANN

Evaluation: Classification error

Optimization: Gradient Descent on weights
Representation

Evaluation

Optimization
Thank you for
listening

Artificial Neural Networks
Home assignment 4
A set of actions it can take in the world.

A way to predict what effect an action has. (internal model of reality)

A way to evaluate the different effects these actions have.
Proposed by Pedro Domingos in his paper "A Few Useful Things to Know about Machine Learning", in 2012
1959, Arthur Samuel
1998, Tom Mitchell
"Field of study that gives computers the ability to learn without being explicitly programmed".
"A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E".
History and connection to other subjects
Developed from Artificial Intelligence.

Machines that can learn from data.

Artificial Intelligence lost interest in data analysis, statistics and neural networks.

Separate sub field and started to flourish in the 1990s.

Borrowed methods and models from statistics and probability theory.

Sub field to Computer Science and Artificial intelligence, strongly connected to Statistics, Probability, Neuroscience, Information Theory and Signal Processing.
Supervised Learning
Regression
Classification
Unsupervised Learning
Clustering

Linear regression
Nonlinear regression
General Artificial Intelligence
General AI on the NES
Predict the world by emulating the game for each set of inputs an amount of frames into the future. Then executing the set of inputs that improves the utility the most.
Learn evaluation of world states and reasonable sequences of inputs, from real player input.
Tom Murphy, "The First Level of Super Mario Bros. is Easy with Lexicographic Orderings and Time Travel . . . after that it gets a little tricky." for the SIGBOVIK Conference 2013
The definition

History and relations to other fields

Supervised

Unsupervised learning

Important parts of machine learning algorithms

Big data and applications
Presentation
The definition

History and relations to other fields

Supervised and unsupervised learning

Important parts of machine learning algorithms

Big data and applications

Short summary
Human Genome
Computer vision




Face recognition
and other medical data
Missing data

British "100,000 genome project"
Labeled data such as audio books

Phones partially based on machine learning
ImageNet: Database of 14,197,122 labeled images.
1. What do and don't we want machines to be taught?

2. Are there subjects which we can't feasibly teach a machine?

3. What negative effects can machine learning applications have?

4. Since we now have the capability to process large amounts of data, are there any fields would be especially benefited by the collection of data?

5. According to the presentation, a good way to divide a learning algorithm is: Representation, Evaluation, Optimization. What other divisions can you think of?

Questions:
jhf
Full transcript