3.5.7 Learning

Learning Including Computational Models Both in Normal Learning and in Pathology (Associative Learning by Hebbian Adaptation, Unsupervised vs. Supervised, Reinforcement)

Learning is a complex process that involves the formation of new neural connections, as well as the strengthening or weakening of existing connections, in the brain. This process, known as neural plasticity, allows the brain to adapt and change in response to new experiences and information (Kolb & Whishaw, 2016).

The neural circuits that are involved in learning are typically located in the hippocampus, a region of the brain that is important for memory and spatial navigation (Eichenbaum, 2014). The hippocampus is connected to various other brain regions, including the prefrontal cortex and the amygdala, which play important roles in learning and emotion (Bechara, Tranel, & Damasio, 2005).

When an experience or piece of information is first encountered, it is processed by the hippocampus and then consolidated into long-term memory in the neocortex (Squire, Wixted, & Clark, 2004). This process involves the strengthening of the neural connections between neurons, a process that is thought to be mediated by the release of neurotransmitters such as glutamate and GABA (Lisman & Grace, 2005).

Overall, the neural circuits of learning are complex and involve the interaction of multiple brain regions and neurotransmitter systems (Kolb & Whishaw, 2016).

Computational models:

Computational models are mathematical or algorithmic representations of a system or process, and they can be used to study and understand normal learning and learning in pathology.

In normal learning, computational models can be used to understand how the brain processes and stores new information. For example, one computational model of memory consolidation proposes that new information is initially stored in the hippocampus and then gradually transferred to the neocortex for long-term storage (McClelland, McNaughton, & O’Reilly, 1995). This model has been supported by a number of experimental studies and has helped researchers to understand how the brain consolidates new memories over time.

Computational models can also be used to study learning in pathologies, such as in the case of individuals with brain injuries or neurological disorders. For example, computational models have been used to study learning and memory impairments in individuals with Alzheimer’s disease (AD), a condition characterized by the accumulation of amyloid-beta plaques and tau tangles in the brain (Jack, Knopman, & Jagust, 2013). These models have helped researchers to understand the specific cognitive and neural changes that occur in AD and have provided insights into potential treatment strategies.

Computational models can be powerful tools for studying normal and pathological learning, as they allow researchers to test and refine hypotheses about how the brain processes and stores new information.

Associative learning by Hebbian adaptation:

Associative learning is a type of learning that occurs when an animal or person learns to associate a particular stimulus or event with a particular response. One type of associative learning that has been extensively studied is Hebbian adaptation, which is also known as Hebbian learning or Hebb’s rule.

Hebbian adaptation is a form of learning that occurs through the strengthening of the connection between two neurons (or between a neuron and a muscle) that are active at the same time. This strengthening is thought to occur through the release of neurotransmitters and the activation of certain signalling pathways.

Hebbian adaptation is often described as “neurons that fire together, wire together,” which means that the more often two neurons are active at the same time, the stronger their connection becomes. This type of learning is thought to be important for the formation of associations between stimuli and responses in the brain, as well as for the formation of long-term memories.

There is evidence that Hebbian adaptation plays a role in a variety of learning processes, including classical conditioning (Pavlov, 1927), instrumental conditioning (Thorndike, 1911), and habituation (Gibbon & Balsam, 2004). However, it is important to note that Hebbian adaptation is just one of many mechanisms that contribute to learning and memory, and it is likely that a combination of multiple mechanisms are involved in most learning situations.

Unsupervised and supervised learning:

Both unsupervised and supervised learning can play a role in normal learning.

Unsupervised learning is a type of machine learning where the learner is not given any explicit instructions or labelled examples but is instead expected to discover patterns and relationships in the data on its own. In the context of human learning, unsupervised learning may occur when an individual is exposed to a new environment or situation and is able to explore and learn about the environment on their own. For example, a child who is given free play time in a playground may learn about their surroundings and the properties of different objects through unsupervised exploration.

Supervised learning, on the other hand, is a type of machine learning where the learner is given labelled examples and explicit instructions about what to learn. In the context of human learning, supervised learning may occur when an individual is taught a specific skill or task by a teacher or mentor. For example, a child who is given explicit instructions on how to play a musical instrument is engaging in supervised learning.

Both unsupervised and supervised learning play important roles in normal learning and development. Unsupervised learning allows individuals to explore and learn about their surroundings on their own, while supervised learning allows individuals to learn specific skills and knowledge through instruction and guidance (Pascalis, 2005).

Reinforcement learning:

Reinforcement learning is a type of learning that occurs when an animal or person learns to associate a particular behaviour with a reinforcing consequence. Reinforcing consequences are those that increase the likelihood that a behaviour will be repeated while punishing consequences are those that decrease the likelihood that a behaviour will be repeated.

In normal learning, reinforcement learning is thought to play a role in a variety of learning processes, including classical conditioning (Pavlov, 1927), instrumental conditioning (Thorndike, 1911), and habituation (Gibbon & Balsam, 2004). For example, an animal that is reinforced with a food reward for pressing a lever will be more likely to repeat the lever-pressing behaviour in the future.

Reinforcement learning can also play a role in learning in pathology. For example, individuals with addiction disorders may engage in behaviours that are reinforced by the use of drugs or alcohol, leading to a cycle of reinforcement that can be difficult to break. In these cases, reinforcement learning principles may be used to design interventions that aim to change the reinforcing consequences of the behaviour and break the cycle of reinforcement.

Reinforcement learning is a fundamental process that plays a role in both normal learning and learning in pathology, and it is an active area of research in both psychology and neuroscience.

References:

(1) Bechara, A., Tranel, D., & Damasio, H. (2005). The somatic marker hypothesis: A neural theory of economic decision. Games and Economic Behavior, 52(2), 336-372.

(2) Eichenbaum, H. (2014). The hippocampus: A unique interface between memory and the environment. Nat Rev Neurosci, 15(6), 529-539.

(3) Gibbon, J., & Balsam, P. D. (2004). Habituation: a non-associative learning process. Scholarpedia, 9(2), 3378.

(4) Jack, C. R., Knopman, D. S., & Jagust, W. J. (2013). The epidemiology of Alzheimer’s disease: risk factors and prevention. Cold Spring Harb Perspect Med, 3(3), a011451.

(5) Kolb, B., & Whishaw, I. Q. (2016). Fundamentals of human neuropsychology. New York, NY: Worth Publishers.

(6) Lisman, J. E., & Grace, A. A. (2005). The hippocampal-VTA loop: controlling the entry of information into long-term memory. Nat Rev Neurosci, 6(11), 773-783.

(7) McClelland, J. L., McNaughton, B. L., & O’Reilly, R. C. (1995). Why there are complementary learning systems in the hippocampus and neocortex: Insights from the successes and failures of connectionist models of learning and memory. Psychol Rev, 102(3), 419-457.

(8) Pascalis, O., & Slater, A. (2005). Plasticity of face processing in infancy. Trends in Cognitive Sciences, 9(5), 254-259.

(9) Pavlov, I. P. (1927). Conditioned reflexes. Oxford, UK: Oxford University Press.

(10) Squire, L. R., Wixted, J. T., & Clark, R. E. (2004). The medial temporal lobe. Annu Rev Neurosci, 27, 279-306.

(11) Thorndike, E. L. (1911). Animal intelligence: An experimental study of the associative processes in animals. New York, NY: Macmillan.