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

Meta-learning

No description
by

Pavel Kordik

on 16 October 2014

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Meta-learning

What is
meta-learning?

Meta-learning in recommender systems
Meta-learning in predictive modeling
See our chapter on self organization of supervised models
Our actual research
Practical impacts
(in Modgen)
Algorithm footprints
: Kate Smith-Miles -Monash university, Australia
Meta-learning
Computational Intelligence research Group at CTU in Prague
Company offering precise predictive models and recommender systems
Pavel Kordík
Faculty of Information Technology,
CTU in Prague

Meta-learning in optimization
Knowledge based approaches
Ensemble methods
Several methods
Definitions
Machine learning systems that utilize information from previous (related) problems (runs) to improve learning in future.
Machine learning
meta-data
can be used
in several ways.
We will focus on
ensemble methods
and
knowledge base approaches.
http://expdb.cs.kuleuven.be/expdb/index.php
Meta-data
Statistical:
Statistical Measures (e.g. mean of numerical attributes)
Simple Measures (e.g. number of attributes, classes )
Information-based measures (e.g. entropy of classes)
Histograms based
information regarding the distribution of values of attributes with relational nature (e.g. mutual information between symbolic attributes and class)
Landmarking
use the performance of simple (fast) learners to predict the performance of candidate algorithms
StatLog project, Metal project: www.metal-kdd.org

Application field
To select the suitable algorithm (algorithm recommendation)
To combine algorithms in a clever way: ensembling
To use the meta-information in order to construct better learning algorithm in general

OpenML results for the BreastCancer UCI data
Boosting
Bagging - bootstrap aggregating
Bias-variance decomposition of error
ceur-ws.org/Vol-1201/MetaSel2014-complete.pdf
Diversity in ensembles
Recommender systems
http://predictorfactory.com/doku.php/algorithm
KNN and AR on LastFM data
Clustering
Predictive models
Automation in data preparation
Meta optimization for TSP problems
Oleg Kovářík, CIG
https://www.researchgate.net/publication/41942762_Meta-learning_approach_to_neural_network_optimization
Jan Černý, CIG
Tomáš Řehořek, CIG
Tomáš Bartoň, CIG
Jan Motl, CIG
Metadatabáze
BI Hackaton:
http://enterprise.hackathon.bi/

CIG meetings:
http://cig.fit.cvut.cz/doku.php?id=public:meetings:root

Invitations
Full transcript