Free energy principle (original) (raw)
The free energy principle is a mathematical principle in biophysics and cognitive science that provides a formal account of the representational capacities of physical systems: that is, why things that exist look as if they track properties of the systems to which they are coupled. It establishes that the dynamics of physical systems minimise a quantity known as surprisal (which is just the negative log probability of some outcome); or equivalently, its variational upper bound, called free energy. The principle is formally related to variational Bayesian methods and was originally introduced by Karl Friston as an explanation for embodied perception-action loops in neuroscience, where it is also known as active inference.
| Property | Value |
|---|---|
| dbo:abstract | The free energy principle is a mathematical principle in biophysics and cognitive science that provides a formal account of the representational capacities of physical systems: that is, why things that exist look as if they track properties of the systems to which they are coupled. It establishes that the dynamics of physical systems minimise a quantity known as surprisal (which is just the negative log probability of some outcome); or equivalently, its variational upper bound, called free energy. The principle is formally related to variational Bayesian methods and was originally introduced by Karl Friston as an explanation for embodied perception-action loops in neuroscience, where it is also known as active inference. The free energy principle models the behaviour of systems that are distinct from, but coupled to, another system (e.g., an embedding environment), where the degrees of freedom that implement the interface between the two systems is known as a Markov blanket. More formally, the free energy principle says that if a system has a "particular partition" (i.e., into particles, with their Markov blankets), then subsets of that system will track the statistical structure of other subsets (which are known as internal and external states or paths of a system). The free energy principle is based on the Bayesian idea of the brain as an “inference engine.” Under the free energy principle, systems pursue paths of least surprise, or equivalently, minimize the difference between predictions based on their model of the world and their sense and associated perception. This difference is quantified by variational free energy and is minimized by continuous correction of the world model of the system, or by making the world more like the predictions of the system. By actively changing the world to make it closer to the expected state, systems can also minimize the free energy of the system. Friston assumes this to be the principle of all biological reaction. Friston also believes his principle applies to mental disorders as well as to artificial intelligence. AI implementations based on the active inference principle have shown advantages over other methods. Although challenging even for experts, the free energy principle is ultimately quite simple and fundamental, and can be re-derived from conventional mathematics following maximum entropy inference. Indeed, it can be shown that any large enough random dynamical system will display the kind of boundary that allows one to apply the free energy principle to model its dynamics: the probability of finding a Markov blanket in the underlying potential of the system (and therefore, being able to apply the free energy principle) goes to 100% as the size of the system goes to infinity The free energy principle is a mathematical principle of information physics: much like the principle of maximum entropy or the principle of least action, it is true on mathematical grounds. To attempt to falsify the free energy principle is a category mistake, akin to trying to falsify calculus by making empirical observations. (One cannot invalidate a mathematical theory in this way; instead, one would need to derive a formal contradiction from the theory.) In a 2018 interview, Friston explained what it entails for the free energy principle to not be subject to falsification: "the free energy principle is what it is — a principle. Like Hamilton's principle of stationary action, it cannot be falsified. It cannot be disproven. In fact, there’s not much you can do with it, unless you ask whether measurable systems conform to the principle." (en) Le principe d'énergie libre est un cadre utilisé pour décrire comment le cerveau prend des décisions et s'adapte à des circonstances changeantes. Il repose sur l'idée que le cerveau essaie constamment de minimiser son énergie libre, qui est une mesure de la surprise que le cerveau ressent face aux informations sensorielles qu'il reçoit. En minimisant son énergie libre, le cerveau peut faire des prédictions sur le monde qui l'entoure et prendre les actions appropriées. La minimisation implicite de l'énergie libre variationnelle est formellement liée aux méthodes bayésiennes variationnelles et a été introduite à l'origine par Karl Friston comme explication de la perception incarnée dans les neurosciences, où elle est aussi connue sous le nom d'inférence active ou de codage prédictif. En termes généraux, le principe de l'énergie libre est utilisé pour décrire un système tel que défini comme étant enfermé dans une couverture de Markov - essaie de minimiser la différence entre son modèle du monde et la perception de ses capteurs. Cette différence peut être qualifiée de « surprise » et minimisée par une mise à jour constante du modèle du monde. En tant que tel, le principe est basé sur l'idée bayésienne du cerveau comme « moteur d'inférence ». Friston a ajouté une deuxième voie à la minimisation : l'action. En changeant activement le monde dans l'état attendu, les systèmes peuvent aussi minimiser l'énergie libre du système. Friston suppose que c'est le principe de toute réaction biologique. Friston considère que son principe s'applique aussi bien aux troubles mentaux qu'à l'intelligence artificielle. Les implémentations de l'IA basées sur le principe de l'inférence active ont montré des avantages par rapport à d'autres méthodes. Il est également utilisé pour expliqué comment un systèmes biologiques se maintiennent en état de non-équilibre quasi-stationnaire en se limitant à un nombre limité d'états (fr) O princípio da energia livre é uma declaração formal que explica como os sistemas vivos e não vivos permanecem em estados estacionários de não equilíbrio, restringindo-se a um número limitado de estados. Estabelece que os sistemas minimizam uma função de energia livre de seus estados internos (não deve ser confundida com energia livre termodinâmica), o que implica crenças sobre estados ocultos em seu ambiente. A minimização implícita da energia livre está formalmente relacionada aos métodos variacionais Bayesianos e foi originalmente introduzida por Karl Friston como uma explicação para a percepção incorporada na neurociência, onde também é conhecida como inferência ativa. O princípio da energia livre explica a existência de um determinado sistema modelando-o através de um Envoltório de Markov que tenta minimizar a diferença entre seu modelo de mundo e seu sentido e percepção associados. Essa diferença pode ser descrita como "surpresa" e é minimizada pela correção contínua do modelo de mundo do sistema. Como tal, o princípio é baseado na ideia bayesiana do cérebro como um “motor de inferência”. Friston adicionou um segundo caminho para a minimização: a ação. Ao mudar ativamente o mundo para o estado esperado, os sistemas também podem minimizar a energia livre do sistema. Friston assume que este é o princípio de toda reação biológica. Friston também acredita que seu princípio se aplica tanto aos transtornos mentais quanto à inteligência artificial. Implementações de IA baseadas no princípio de inferência ativa mostraram vantagens sobre outros métodos. O princípio da energia livre tem sido criticado por ser muito difícil de entender, mesmo para especialistas, e a consistência matemática da teoria tem sido questionada por estudos recentes. As discussões do princípio também foram criticadas por invocar suposições metafísicas muito distantes de uma previsão científica testável, tornando o princípio infalsificável. Em uma entrevista de 2018, Friston reconheceu que o princípio da energia livre não é adequadamente falsificável: "o princípio da energia livre é o que é - um princípio. Como o princípio de ação estacionária de Hamilton, não pode ser falsificado. Não pode ser refutado. Na verdade, não há muito o que fazer com ele, a menos que você pergunte se os sistemas mensuráveis estão de acordo com o princípio." (pt) |
| dbo:thumbnail | wiki-commons:Special:FilePath/MarokovBlanketFreeEnergyFigure.jpg?width=300 |
| dbo:wikiPageExternalLink | http://journals.cambridge.org/action/displayAbstract%3FfromPage=online&aid=8918803 |
| dbo:wikiPageID | 39403556 (xsd:integer) |
| dbo:wikiPageLength | 40972 (xsd:nonNegativeInteger) |
| dbo:wikiPageRevisionID | 1124746214 (xsd:integer) |
| dbo:wikiPageWikiLink | dbr:Calculus dbr:Calculus_of_variations dbr:Bayesian_inference dbr:Bellman_equation dbr:Predictive_coding dbr:Principle dbr:Principle_of_least_action dbr:Partially_observable_Markov_decision_process dbr:Bayes'_theorem dbr:Approximate_Bayesian_computation dbr:Joint_probability_distribution dbr:Perception dbr:Cybernetics dbr:Degrees_of_freedom_problem dbr:Inference_engine dbr:Information_content dbr:Mental_disorder dbr:Generalized_filtering dbr:Principle_of_minimum_energy dbr:Cognitive_neuroscience dbr:Generative_model dbr:Model_selection dbr:Mutual_information dbr:Optimal_decision dbr:Embodied_cognition dbr:Hamilton's_principle dbr:Helmholtz_decomposition dbr:Helmholtz_free_energy dbr:Self-assembly dbr:Pattern_formation dbr:Synaptic_plasticity dbr:Autopoiesis dbc:Computational_neuroscience dbc:Systems_theory dbr:Hebbian_theory dbr:Ergodic_theory dbr:Falsifiability dbr:Fluctuation_theorem dbr:Energy_function dbr:Kalman_filter dbr:Kolmogorov_equations dbr:Probability_density_function dbr:Variational_free_energy dbr:Posterior_probability dbr:Good_Regulator dbr:Hermann_von_Helmholtz dbr:Marginal_likelihood dbr:Artificial_intelligence dbc:Mathematical_and_theoretical_biology dbc:Biological_systems dbr:Synergetics_(Haken) dbr:Dopamine dbr:Marginal_distribution dbr:Maximum_entropy_principle dbr:Kullback–Leibler_divergence dbr:Optimal_control dbr:Second_law_of_thermodynamics dbr:Self-organisation dbr:Markov_blanket dbr:Sense dbr:Utility dbr:Variational_Bayesian_methods dbr:Neuroscience dbr:Unconscious_inference dbr:Occam's_razor dbr:Karl_Friston dbr:Random_dynamical_system dbr:Self-information dbr:Bayesian_brain dbr:Practopoiesis dbr:Surprisal dbr:File:MarokovBlanketFreeEnergyFigure.jpg dbr:File:PESAIM.jpg |
| dbp:wikiPageUsesTemplate | dbt:Annotated_link dbt:Failed_verification dbt:Reflist dbt:Short_description |
| dcterms:subject | dbc:Computational_neuroscience dbc:Systems_theory dbc:Mathematical_and_theoretical_biology dbc:Biological_systems |
| rdf:type | yago:Artifact100021939 yago:Instrumentality103575240 yago:Object100002684 yago:PhysicalEntity100001930 yago:System104377057 yago:Whole100003553 yago:WikicatSystems |
| rdfs:comment | The free energy principle is a mathematical principle in biophysics and cognitive science that provides a formal account of the representational capacities of physical systems: that is, why things that exist look as if they track properties of the systems to which they are coupled. It establishes that the dynamics of physical systems minimise a quantity known as surprisal (which is just the negative log probability of some outcome); or equivalently, its variational upper bound, called free energy. The principle is formally related to variational Bayesian methods and was originally introduced by Karl Friston as an explanation for embodied perception-action loops in neuroscience, where it is also known as active inference. (en) Le principe d'énergie libre est un cadre utilisé pour décrire comment le cerveau prend des décisions et s'adapte à des circonstances changeantes. Il repose sur l'idée que le cerveau essaie constamment de minimiser son énergie libre, qui est une mesure de la surprise que le cerveau ressent face aux informations sensorielles qu'il reçoit. (fr) O princípio da energia livre é uma declaração formal que explica como os sistemas vivos e não vivos permanecem em estados estacionários de não equilíbrio, restringindo-se a um número limitado de estados. Estabelece que os sistemas minimizam uma função de energia livre de seus estados internos (não deve ser confundida com energia livre termodinâmica), o que implica crenças sobre estados ocultos em seu ambiente. A minimização implícita da energia livre está formalmente relacionada aos métodos variacionais Bayesianos e foi originalmente introduzida por Karl Friston como uma explicação para a percepção incorporada na neurociência, onde também é conhecida como inferência ativa. (pt) |
| rdfs:label | Free energy principle (en) Principe de l'énergie libre (fr) Princípio da energia livre (pt) |
| owl:sameAs | freebase:Free energy principle yago-res:Free energy principle wikidata:Free energy principle dbpedia-fr:Free energy principle dbpedia-pt:Free energy principle https://global.dbpedia.org/id/fSmm |
| prov:wasDerivedFrom | wikipedia-en:Free_energy_principle?oldid=1124746214&ns=0 |
| foaf:depiction | wiki-commons:Special:FilePath/MarokovBlanketFreeEnergyFigure.jpg wiki-commons:Special:FilePath/PESAIM.jpg |
| foaf:isPrimaryTopicOf | wikipedia-en:Free_energy_principle |
| is dbo:knownFor of | dbr:Karl_J._Friston |
| is dbo:wikiPageDisambiguates of | dbr:Free_energy |
| is dbo:wikiPageRedirects of | dbr:The_free_energy_principle dbr:Active_inference |
| is dbo:wikiPageWikiLink of | dbr:Bayesian_approaches_to_brain_function dbr:Predictive_coding dbr:Sam_Edwards_(physicist) dbr:Biological_neuron_model dbr:Interoception dbr:Self-organization dbr:Abductive_reasoning dbr:Captive_bubble_method dbr:Free_energy dbr:Karl_J._Friston dbr:Synergetics_(Haken) dbr:Mental_time_travel dbr:The_free_energy_principle dbr:Information_theory dbr:Neats_and_scruffies dbr:Markov_blanket dbr:Unconscious_inference dbr:Occam's_razor dbr:Active_inference |
| is dbp:knownFor of | dbr:Karl_J._Friston |
| is foaf:primaryTopic of | wikipedia-en:Free_energy_principle |
