Actions

Algoliterary Encounters: Difference between revisions

From Algolit

Line 16: Line 16:
 
===A few outputs to see how it works===
 
===A few outputs to see how it works===
 
* CHARNN text generator
 
* CHARNN text generator
* [[human & view & power in 5 landscapes]] - Five word2vec graphs, each of them containing the words 'human', 'view' and 'power'. (Before: [[talking_about_machine_learning]] - exploring the vocabulary of machine learning textbooks in 7 stages with word2vec)
+
* [[human & view & power in 5 landscapes]] - Five word2vec graphs, each of them containing the words 'human', 'view' and 'power'.  
 +
<br>(Before: [[talking_about_machine_learning]] - exploring the vocabulary of machine learning textbooks in 7 stages with word2vec)
  
 
===Parts of NN process===
 
===Parts of NN process===

Revision as of 08:31, 17 October 2017


Start of the Algoliterary Encounters catalog.

General Introduction

Algoliterary works

  • Oulipo scripts
  • i-could-have-written-that interfaces
  • Obama, model for a politician
  • ClueBotNG, a special Algolit edition

Algoliterary explorations

A few outputs to see how it works


(Before: talking_about_machine_learning - exploring the vocabulary of machine learning textbooks in 7 stages with word2vec)

Parts of NN process

Datasets

From words to numbers

Different views on the data

  • tensorflow projector visualisation of high dimensional data
  • 5 dimensions 32 graphs
  • GloVe dataset sorted alphabetically

Creating word embeddings using word2vec

Autonomous machine as inspection

  • AI script showing racist bias using supervised classical ML & NN embeddings

Algoliterary Toolkit

Bibliography