Spontaneous learning from a model: adjustment and reinforcement.

Learning is the passage from cerebral knowledge to its realization through the body or through the iteration of experience. On August 3, 2015, McGill University in Montreal published the findings of research on golfers, research that sheds light on how the body interacts with learning and the unexpected.

The golfer model sheds light on the field of learning

"Learning to move

We knew that we owe our sporting performance to a tiny cluster of cells nestled deep in our cerebellum. What we didn't know, and what researchers from the Department of Physiology at McGill University in Canada have just discovered, is that during the learning of new motor gestures (a golf swing, for example), the neurons in the cerebellum perform elegant, quasi-mathematical calculations to compare, in real time, what they feel with what they expected to feel.

They then quickly adjust by altering the strength of connections between other neurons to form new cerebral patterns to perform the task at hand. In other words, individual brain neurons have the ability to dynamically recognize the difference between expected sensory feedback and the information they actually receive during motor learning. The calculated difference is then used to rapidly modify brain patterns and connections between neurons to enable the learning of new motor skills."

Sources: "50 - Learning to move: our brain can predict the unexpected" by Hubert Desrues
August 23, 2015 - https://www.ogolf.fr/50-apprentissage-mouvement-cerveau-sait-prevoir-linattendu/

All learning begins with a theoretical or practical starting point. This is the point of comparison. It will underpin the experience and enrich it each time the experiment is repeated. We can compare this knowledge to a cloud of control or comparison points, whose periodicity from one point to the next will more or less fortify the knowledge and its transmission in the brain.

The more recurrent the information, the more neural routes and highways it will create, conducive to the integration of close or complementary experience. There's an iterative mechanics to learning, a back-and-forth movement from one experience to the next that generates the modelling and integration of intellectual patterns that will foster the opportunity to integrate new physiological, physical and bodily knowledge.

But we mustn't limit living experience to that of the model; it can also generate its own models.

"Our brain can predict

To master a new movement, the brain first estimates the input it should receive from the sensory system. The cerebellum then uses this prediction to calculate the discrepancy between what the person intended to do and what they actually did...

In any case, this discovery confirms one of our intuitions: the sensation of the swing, felt at the end of execution, is indeed one of the active elements of our learning... But our brain needs a model to know what to expect and to compare with the sensation actually felt. Hence the importance of teaching, and of observing "high-level" players..."

CF "50 - Learning to move: our brain knows how to anticipate the unexpected"

The model is not only intellectual, it is also supported by sensory validation points, the model can even be solely sensory and intuitive and therefore not constructed by the intellect. The body's memory can also serve as a predictive model. Sight memory for golf also enters into this game of comparisons. The living model is based on 3 pillars: the intellect, the feeling of the body and observation by the senses.

In fact, we're like machines learning from the living world.

"The algorithms used enable, to a certain extent, a computer-controlled (possibly a robot), or computer-assisted, system to adapt its analyses and behaviors in response, based on the analysis of empirical data from a database or sensors.

The difficulty lies in the fact that the set of all possible behaviors given all possible inputs quickly becomes too complex to describe (this is known ascombinatorial explosion). We therefore entrust programs with the task of adjusting a model to simplify this complexity, and using it operationally. Ideally, learning should be unsupervised, i.e. the nature of the training data is unknown.

Depending on their degree of sophistication, these programs may incorporate capabilities in probabilistic data processing, sensor data analysis, recognition (voice recognition, shape recognition, handwriting recognition, etc.), data-mining, theoretical computer science, etc."

Sources Wikipedia : Machine learning - Chapter - principles
https://fr.wikipedia.org/wiki/Apprentissage_automatique

Human learning thus corresponds to automatic machine learning. The object is different, but the methods are the same. Computer scientists have recreated the human thought process by trying to give machines learning intelligence.

The model simplifies complexity and avoids the phenomenon of combinatorial explosion. Today, our world is becoming more and more complex. Is there an exploratory field to be opened up on the subject? To be continued...

Image source: Pixabay 422737