Everything is in Python 2.7 (sorry). See requierements.txt for dependencies.
(Can be executed for a little example)
- echo_state_network.py : Definition of EchoStateNetwork class, which can only learn offline.
- FORCE_learning.py : Definition of EchoStateNetworkRLS, which provides RLS online learning.
-
sentence_processing_network.py : Definition of class SentenceProcessingNetwork, which takes sentences as input and can learn with RLS.
-
sentence_grounding_network.py : Definition of class SentenceGroundingNetwork, which takes sentences as input and returns a list of object having different caracteristics, and provides a method to do cross situational learning.
(Can be executed)
-
sentence_grounding_test.py : Test of the sentence gounding network. A training is done creating a false vision and randomly combining sentences from sentences_to_predicate. Can be modified changing reservoir parameters, and CATEGORIES, POSITIONS, COLORS in sentence_grounding_test_parameters.py file.
-
sentence_grounding_test_plots.py : Plots the saved performance curves, or computes them.
-
sentence_grounding_test_parameters.py: uses grammar_manipulation.py to create the grammar used for the test
-
recognized_object.py : Definition of object returns by the module, representing an object seen on the picture
-
sentence_to_predicate.py : Definition of the class WordPredicate, wich is construct by reading an annotated sentence from sentence_to_role.py. Definition of a set of sentences annotaded with the predicate they contain.
-
one_hot_encoder.py : To transform object from a given set into a (0,...,1,...,0) vector.