Not true, learning does occur but most likely a bit differently from what happens in the cortex.
“Although pallial structures exist in amphibians and fish, reptiles and mammals are the only vertebrates to have a cerebral cortex with a clear, though simple, three-layered structure, similar to that of mammalian allocortex.” Naumann, Robert K et al. “The reptilian brain.” Current biology : CB vol. 25,8 (2015): R317-21. doi:10.1016/j.cub.2015.02.049
The bottom line is that fish and amphibians can learn–no cortex.
The classical notion that the cerebellum and the basal ganglia are dedicated to motor control is under dispute given increasing evidence of their involvement in non-motor functions. Is it then impossible to characterize the functions of the cerebellum, the basal ganglia and the cerebral cortex in a simplistic manner? This paper presents a novel view that their computational roles can be characterized not by asking what are the “goals” of their computation, such as motor or sensory, but by asking what are the “methods” of their computation, specifically, their learning algorithms. There is currently enough anatomical, physiological, and theoretical evidence to support the hypotheses that the cerebellum is a specialized organism for supervised learning, the basal ganglia are for reinforcement learning, and the cerebral cortex is for unsupervised learning.
This paper investigates how the learning modules specialized for these three kinds of learning can be assembled into goal-oriented behaving systems. In general, supervised learning modules in the cerebellum can be utilized as “internal models” of the environment. Reinforcement learning modules in the basal ganglia enable action selection by an “evaluation” of environmental states. Unsupervised learning modules in the cerebral cortex can provide statistically efficient representation of the states of the environment and the behaving system. Two basic action selection architectures are shown, namely, reactive action selection and predictive action selection. They can be implemented within the anatomical constraint of the network linking these structures. Furthermore, the use of the cerebellar supervised learning modules for state estimation, behavioral simulation, and encapsulation of learned skill is considered. Finally, the usefulness of such theoretical frameworks in interpreting brain imaging data is demonstrated in the paradigm of procedural learning.
Hi,
This post of yours prompts me to mention my concEPT (short for ÆPT concept) “actention”, and that I’ve used it to contrive a both seriously meant and mirth-inviting alternative to “brain” or “nervous system”; namely (that we humans and all other animals have a neural) Actention Selection Serving System.
I maximally truncated this concEPT to the “AS” that I put (to acronym-building work) right in the middle of EAVASIVE (the most integrative of all the concEPTs I contrived to be expressible as acronyms.
The ASSS (the concEPT’s minimal truncation) is an ingredient of a by me arrived at effectively philosophy terminating but also satisfyingly science-aligned ‘atheistically enlightening perspective, textualised’; thereof and from similar enough (but never derogatory) word-strings have I seen fit to derive the facetious looking and by me flexibly raised flag/naming-gimmick ‘ÆPT’.
P.S. It should be obvious that ÆPT was informally formulated, and that I’ve had fun putting it on electronic paper’ (using Wix), and that I wrote it all partly but prominently with help of “MAD”-inspired sem_antics.
What else ÆPT contains (much of which is still awaiting editing/cleaning up) can be found-out at www.aeimcinternetional.org
Flace P, Livrea P, Basile GA, Galletta D, Bizzoca A, Gennarini G, Bertino S, Branca JJV, Gulisano M, Bianconi S, Bramanti A and Anastasi G
2021
doi: 10.3389/fnsys.2021.650614
Abstract
In the central nervous system (CNS), dopamine (DA) is involved in motor and cognitive functions. Although the cerebellum is not been considered an elective dopaminergic region, studies attributed to it a critical role in dopamine deficit-related neurological and psychiatric disorders [e.g., Parkinson’s disease (PD) and schizophrenia (SCZ)]. Data on the cerebellar dopaminergic neuronal system are still lacking. Nevertheless, biochemical studies detected in the mammalians cerebellum high dopamine levels, while chemical neuroanatomy studies revealed the presence of midbrain dopaminergic afferents to the cerebellum as well as wide distribution of the dopaminergic receptor subtypes (DRD1-DRD5). The present review summarizes the data on the cerebellar dopaminergic system including its involvement in associative and projective circuits. […]
My Commentary
Before I argued that all four categories of learning were implemented in various areas of the brain. However the full picture is more nuanced: each area of the brain likely uses multiple types of learning. This article shows that the cerebellum contains a lot of dopaminergic signals, which according to the current theory represent a reinforcement learning signal. In addition, other researchers demonstrated learning in the cerebellum in the absence of the supervised learning signals, which they concluded means that it likely can perform unsupervised learning. Finally, the cerebellum has distinct regions with significant anatomical differences between them, and such specializations are clearly the result of evolution.
In terms of the difficulty of evolving this: making new learning systems is difficult, but taking an existing system and repurposing it is comparatively easy. The cerebellar dopaminergic system appears to use the same molecular mechanisms as the primary dopaminergic systems in the basal ganglia. The dopamine receptors in the cerebellum evolved first in the basal ganglia and then later those genes were simply turned on in new parts of the brain.
I think this mapping to cortex, BG, cerebellum is vaguely suggestive. Except for evolutionary learning, it’s not done in the brain. But there is no clean separation, all three types of learning are performed in all areas, just in different proportion.
More importantly, there is a confusion between an actor and a target.
Supervisor is basically any inherited stimulus-pattern-specific neural architecture, while supervised learning is done in all areas that receive the outputs. Both are present in all areas, including the cortex. The difference is that cortical density of both is much lower than in brainstem, and they are more abstract.
Same thing for RL, except that it’s downstream from supervision: the architecture here is not innate but persistently modulated by innate response patterns.
In pure unsupervised learning, the innate architecture is uniform, all the specifics come from value-free sensory input.
Re cerebellum, it’s a largely homogenous organ, which is similar to cortex. That means the density of specific innate patterns is low, so it’s intrinsic learning mode is also unsupervised. But the density of modulating external inputs is much higher than in the cortex. And learning here is far more detailed, vs. generalized, the structure is flat vs. hierarchical in the cortex.
I think that active learning could be useful as a pattern of behavior for learning things; helpful strategy for studying. I haven’t seen any evidence suggesting that it does or does not happen in the brain, I don’t think it’s something many researchers have considered. But I don’t understand how it could be implemented in a biologically plausible way, especially considering that it requires two participants and animals are capable of learning alone, in isolation.
I agree that active learning can’t be the only style of learning. It does seem to be something different from the other styles. In “systems thinking” one of the suggestions for understanding a system is to reconsider the boundaries of the system under study. To expand the boundary to the level of the system in which the system of interest is embedded. I am fairly convinced this is the only way to understand human intelligence - it is a social process. This will be a challenge for most researchers because it requires study in the humanities and we tend to produce specialists. Grossberg is one researcher who sees this but I don’t think he published a model. Another neuroscientist is Daniel Bullock e.g. Socializing the Theory of Intellectual Development
