[Pycon] [new paper] "Valerio Nicosia" - Active Learning for Training Set Building

[info a pycon.it]

Title: Active Learning for Training Set Building
Duration: 45 (includes Q&A)
Q&A Session: 15
Language: en
Type: Talk

Abstract: Abstract: Tutorial for Python Data Scientists on Training Set Building using Active Learning Algorithms.

This talk explores the use of active learning techniques using “libact” library in populating the training set in a texts classification scenario.
I will explain how different active learning strategies work for some text classification models.
Afterwards, I will show a case study to compare performances of different active learning strategies, as well as to show how the active learning approach works from scratches to build the training set

Tags: [u'active-learning', u'Text-Mining', u'Python', u'Data-Scientist', u'pydata']