Finding the next step…
piatra . institutepathwayspiatra . institutepathwaysFinding the next step…
An agent can change observations through action; inferring a hidden state and selecting an action are different operations.
The short introduction helps with the question. For the mathematical reading, bring probability distributions, Bayes’ rule and basic calculus; the Probability route or equivalent experience is useful. Obtain the specified text before starting.
Build a two-state observation table and compare choosing an informative observation with choosing an immediately rewarding action.
Start here if the background is new. Equivalent experience is enough.
Introduces the question and vocabulary used in the longer selected reading.
Bayes’ rule and probability distributions support the selected mathematical reading; equivalent experience is welcome.
Thomas Parr · Giovanni Pezzulo · Karl J. Friston · book · 2022
Introduction or preface and opening chapter. Record the edition and translator before reading.
Free full text
Open the readingActive Inference. Neuroscience and mathematical frameworks for inference, perception and action. Begin with a small probabilistic model before tackling the formal free-energy principle.
An agent tries to decide whether a door is open. Name a hidden state, an observation and an action in this example.
Keep the world, its sensory evidence and the agent’s intervention separate.
The door’s open/closed condition is a hidden state. An image of the doorway is an observation. Moving to view it from another angle is an action. The model also needs assumptions connecting states to observations.
What to look for
Common mistake: Calling the image itself the hidden state.
Start this pathway to keep your answers in My learning.
Thomas Parr · Giovanni Pezzulo · Karl J. Friston · book · 2022
Locate two passages about “How can an action change the information available to an agent” using the contents or index. Record their chapter/section names in your notes; this route assigns selections, not the whole work.
Free full text
Open the readingActive Inference. Neuroscience and mathematical frameworks for inference, perception and action. Begin with a small probabilistic model before tackling the formal free-energy principle.
Would one accurate prediction establish that active inference is a complete explanation of behaviour?
Compare success on a task with support for a whole framework.
No. We would need a specified model, alternative explanations and observations that discriminate between them. A fit on one example supports a much narrower claim.
What to look for
Common mistake: Treating successful prediction as unique confirmation.
Start this pathway to keep your answers in My learning.
Is variational free energy simply the energy used by the brain?
Ask what quantity each use of “energy” measures.
No. Variational free energy is a quantity in an inference objective. Metabolic energy is a physical quantity. Connecting them requires an argument and assumptions; a shared word does not identify them.
What to look for
Common mistake: Replacing a mathematical objective with a metabolic measurement.
Start this pathway to keep your answers in My learning.
Background, different viewpoints, and further reading.
Neuroscience and mathematical frameworks for inference, perception and action. Begin with a small probabilistic model before tackling the formal free-energy principle.
Introduction or preface and opening chapter. Record the edition and translator before reading.
Read the complete work after the selected passages, using the recorded edition and continuous chapter coverage.