Page 127 - AN-3-3

Advanced Neurology Evaluating plausibility of thalamic model

Our research on the cognitive and electrophysiological simplifications to describe these biological phenomena.
plausibility of a thalamic computational model showcases This may overlook important processes that were
several strengths. The model successfully mimics key not sufficiently considered in the references used
physiological activities of the thalamic circuit, including for program development. In addition, probabilistic
inhibitory facilitation, waveform sculpting in reticular cells, modeling of synapses may not capture all the nuances of
and pattern completion, demonstrating robust biological real synaptic dynamics and requires more experimental
plausibility. Its ability to reconstruct images from partial validation. The assumptions about the linearity and
inputs mirrors human perceptual processes, particularly orthogonality of the basis vectors need to be tested in
in stabilized retinal image experiments. By proposing biological systems, where nonlinearities and noise
the thalamus as a central orthogonalize that transforms are prevalent. Our programming paradigm is matrix-
sensory inputs into orthogonal PCs, the model addresses based, while the morphophysiological organization
interference challenges in Hebbian learning and aligns of the thalamocortical system is columnar, which can
with systems neuroscience theories. In contrast to other impose geometric difficulties in code development. An
computational models of the thalamus, which generally object-oriented programming approach could better
focus on the dynamic properties of neural circuits without manage the structural organization of the system,
emphasizing the probabilistic nature of the computations making code refactoring a future step in our research.
involved, our model bridges the gap between geometric Other limitations include the simplification of cortical
vector spaces and biological neural computation. It offers connections, the reduction in the complexity of
a new perspective on how neural plasticity and learning dendritic trees, and the number of synapses, which can
can be understood from a mathematical standpoint. The limit the accuracy of simulating neuronal interactions.
model integrates a probabilistic pre-synaptic Hebbian The lack of modeling of developmental changes affects
rule, capturing synaptic plasticity dynamics such as LTP the simulation of neuronal maturation and synaptic
and depression, enhancing learning, and adaptability. It plasticity. The simplified representation of stabilized
also replicates experimental findings related to Rs activity, images in the retina and the lack of consideration of
strengthening its credibility. cortical influences limit the understanding of complete
The use of matrix-based programming and an auto- visual processing.
associative neural network structure allows the model to These limitations highlight the need for more
handle large numbers of neurons and adapt dynamically detailed and accurate modeling of neuronal dynamics,
to new inputs, making it scalable and suitable for extensive the integration of cortical influences and developmental
simulations without excessive computational demands. changes, and continuous experimental validation to
Our methodology, enhanced by square weight matrices for strengthen hypotheses and improve model accuracy.
representing the connectivity of the network, captures the A crucial consideration is the presence of a cortico-
regularities of synaptic connections, dynamically adapting thalamic pathway, where layer VI pyramidal neurons
to include new neurons. This facilitates the management of excitatorily project onto thalamic Rs. 22,25,26 In light of the
large neuronal volumes without requiring a proportional described pattern completion process, this pathway could
increase in computational resources. The matrix form enable direct cortical influence on the reticular inhibition
of programming also allows the program to run on process in the relay layer. Phenomena such as perception,
GPUs and quantum computers, significantly boosting its hallucinations, dreams, and consciousness may, to some
17
computational potential. Moreover, our phenomenological extent, find their roots in the processes simulated by our
model aligns with the propositions of Kriegeskorte and network.
61
Douglas on the ideal model for understanding the brain’s Therefore, acknowledging this cortical influence
computational processes. It seeks a balance between in the pattern completion process represents the next
biological and cognitive features, accurately reproducing logical progression in the research. Future studies could
observed phenomena and promoting a deep understanding explore how these computational principles are applied
of brain dynamics. in different neural circuits beyond the thalamus, such as

Although we encounter some significant limitations cortical layers or other subcortical structures involved in
compared to other approaches, as a phenomenological higher cognitive functions. Experimental studies could
model, we must use simplifications to handle aim to directly measure these vector transformations
morphophysiological details. For example, our artificial in vivo, testing the theoretical predictions made by the
neuronal model cannot directly handle the burst and model. By framing our discussion around these points, we
tonic modes of thalamic REs operation, requiring can effectively highlight the significance of our findings

Volume 3 Issue 3 (2024) 15 doi: 10.36922/an.3188