en
0.25
0.5
0.75
1.25
1.5
1.75
2
Free energy and active inference
Published on Oct 16, 201213014 Views
How much about our interactions with - and experience of - our world can be deduced from basic principles? This talk reviews recent attempts to understand the self-organised behaviour of embodied agen
Related categories
Chapter list
Abstract00:00
From the Helmholtz machine to the Bayesian brain and self-organization01:41
Difference between a snowflake and a bird?02:48
The basic ingredients07:05
The principle of least free energy (minimising surprise)09:38
How can we minimize surprise (prediction error)?17:27
Action as inference – the “Bayesian thermostat” (1)19:49
How might the brain minimise free energy (prediction error)?22:50
Free energy minimisation->Generative model->Predictive coding with reflexes23:10
From models to perception27:32
Predictive coding with reflexes30:04
Summary33:52
Generating bird songs with attractors34:45
Predictive coding37:17
Perceptual categorization42:42
Action as inference – the “Bayesian thermostat” (2)46:56
Action with point attractors48:27
Heteroclinic cycle (central pattern generator)48:29
Where do I expect to look?48:53
Sampling the world to minimise uncertainty49:39
Sensations (1)50:24
Sensations (2)50:27
Sensations (3)50:30
Thank you52:22
Article: Dark Room52:42
Time-scale52:45