Karl Friston - 2016 CCN Workshop: Predictive Coding
Вставка
- Опубліковано 7 жов 2024
- Center for Cognitive Neuroscience at Dartmouth
2016 Workshop: Predictive Coding
KARL FRISTON, UNIVERSITY COLLEGE LONDON
Predictive coding, active inference and belief propagation
Abstract:
I will consider prediction and choice based upon the minimisation of expected free energy. Crucially, (negative) free energy can always be decomposed into pragmatic (extrinsic) and epistemic (intrinsic) value. Minimising expected free energy is therefore equivalent to maximising extrinsic value, while maximising information gain or intrinsic value, i.e., reducing uncertainty about the causes of sensory samples. This decomposition resolves the exploration-exploitation dilemma; where epistemic value is maximised until there is no further resolution of uncertainty, after which exploitation is assured through maximisation of extrinsic value. This is formally consistent with the principle of maximum mutual information, generalising formulations of active vision based upon salience (Bayesian surprise) and optimal decisions based on expected utility and risk sensitive (KL) control - using Hamilton's principle of least action. I will briefly review the normative theory - illustrating the minimisation of expected free energy using simulations and then turn to neuronal processes theories. In brief, the implicit (neuronally plausible) belief propagation offers a form of predictive coding, when hidden causes and outcomes are treated as discrete states.
The content of this presentation is pure gold.
this man is genuis..the implication of his revolutionary theory Free Energy principle will unfold in the near future.
i couldn't agree more
one year and counting
4 years@@JohnDoe-nv2op
05:41 is the main part that shows why RL type approaches cannot work in certain situations. I just loved it!
This video lacks of zooming the presentation :/
optimal action function 6:35; Baysesian suprise 10:53
Thanks so much for sharing
Link to the slides?
Hi all, where can I learn the math used in this presentation, like translating the real world observation into math? Is there a book, or a tutorial for a novice like me.
48:30 savage 'ahh your question is trivial'
The answer to the last unanswered question: "what about surprise seeking" is that when exploitation is complete, the explorer is bored. The rat then seeks surprise to avoid boredom.
I used to admire Mr. Friston until he outed himself as a cheap nationalist in his Guardian interview.
what did he say, i read the interview, and he was very cautious with his choice of words and just presented his results. @Hans Richter
Cheap nationalists are the worst kind.
Typical result whenever neuroscientists try to "explain" how mind works. It's like heart surgeon tries to explain how love works.
yes, damn those neuroscientists who try to explain how our mind works. Damn those psychologists too, wizards
Einstein did not make the universe less beautiful when he explained how it worked.
@@valueengines2184 But he did not explain 'why' we find the universe 'beautiful'. Because Einstein was not concerned with how we find things beautiful as beauty was not his area of expertise. It's a different category.
Similarly, neuroscientists can explain how neurons, synaptic connections, or other biochemical elements work; with lots of maps. These things will not give them an insight into how 'mind' works. Just like nobody can figure out how a city operates just by looking at some Google maps or counting the number of people and cars on the streets.
Different categories need different ways of knowing.
@@trashygit neurons can be easily explained: (Expect value - Actual value)/(Expect value + Actual value) 1= total satisfaction, 0= same, end of learning, boredom, -1 = total dissatisfaction, unlearning
@@valueengines2184 Nobody claimed if neurons can be explained or not: We are talking about 'mind'. If some neuro-centred people seriously think that they can explain how mind works over neurons, then good luck.