Automatic Machine Learning (AUTO-AUTO-ENCODER!) - Revision history

Slawek at 17:55, 14 October 2018

2018-10-14T17:55:54Z

Seppas: Seppas moved page Automatic Machine Learn (AUTO-AUTO-ENCODER!) to Automatic Machine Learning (AUTO-AUTO-ENCODER!)

2017-09-27T14:14:41Z

Seppas moved page Automatic Machine Learn (AUTO-AUTO-ENCODER!) to Automatic Machine Learning (AUTO-AUTO-ENCODER!)

Seppas: Seppas moved page Name of the new projectAutomatic Machine Learn (AUTO-AUTO-ENCODER!) to Automatic Machine Learn (AUTO-AUTO-ENCODER!): renamed

2017-09-27T14:14:17Z

Seppas moved page Name of the new projectAutomatic Machine Learn (AUTO-AUTO-ENCODER!) to Automatic Machine Learn (AUTO-AUTO-ENCODER!): renamed

Seppas: Created page with "{{StudentProjectTemplate |Summary=Automatic configuration algorithm for autoencoders |Keywords=Deep learning, autoencoder, meta learning, AutoML |TimeFrame=Winter 2016 - Sprin..."

2017-09-27T14:08:18Z

Created page with "{{StudentProjectTemplate |Summary=Automatic configuration algorithm for autoencoders |Keywords=Deep learning, autoencoder, meta learning, AutoML |TimeFrame=Winter 2016 - Sprin..."

New page

{{StudentProjectTemplate
|Summary=Automatic configuration algorithm for autoencoders
|Keywords=Deep learning, autoencoder, meta learning, AutoML
|TimeFrame=Winter 2016 - Spring 2017
|References=The following paper summarises the algorithm configuration in the different domain :
http://aad.informatik.uni-freiburg.de/papers/16-AUTOML-AutoNet.pdf

This paper presents the initial idea behind Bayesian optimization for estimating parameter:
https://www.cs.ubc.ca/~hutter/papers/10-TR-SMAC.pdf

Previous master thesis on applying autoencoder for histogram data:
Robin Ng, “Efficient Implementation of Histogram Dimension Reduction using Deep Learning”, 2017.

|Prerequisites=Artificial Intelligence and Learning Systems courses,
|Supervisor=Sławomir Nowaczyk, Sepideh Pashami,
|Level=Master
|Status=Open
}}
For anyone who is tired of machine learning algorithm’s configuration.

Algorithm configuration plays an important role in the performance of machine learning methods. In addition, data scientists every day spend a lot of time with a little guidance to choose the parameters of the algorithm. Further, learning the parameter of the algorithm as automatic as possible enables the use of machine learning for a wider range of science and technology.

Usually, state of the art methods target supervised classification machine learning tasks. This project focuses on parameter configurations of autoencoders for variously available datasets. Autoencoder is unsupervised feature extraction technique based on the neural network which trains in a supervised fashion. Following explains the necessary steps toward achieving an automatic autoencoder during this project.
- Studying recent advances in meta-learning, transfer learning, algorithm selection, and algorithm configuration.
- Studying and implementing autoencoder
- Adapting existing algorithm configurations for autoencoder and comparing their performance

@@ Line 2: / Line 2: @@
 |Summary=Automatic configuration algorithm for autoencoders
 |Keywords=Deep learning, autoencoder, meta learning, AutoML
-|TimeFrame=Winter 2016 - Spring 2017
+|TimeFrame=Fall 2018
 |References=The following paper summarises the algorithm configuration in the different domain :
 http://aad.informatik.uni-freiburg.de/papers/16-AUTOML-AutoNet.pdf
@@ Line 11: / Line 11: @@
 Previous master thesis on applying autoencoder for histogram data:
 Robin Ng, “Efficient Implementation of Histogram Dimension Reduction using Deep Learning”, 2017.
 |Prerequisites=Artificial Intelligence and Learning Systems courses,
-|Prerequisites=Artificial Intelligence and Learning Systems courses,
+|Supervisor=Sławomir Nowaczyk, Sepideh Pashami,
 |Level=Master
 |Status=Open

← Older revision	Revision as of 14:14, 27 September 2017
(No difference)