6th Edition of Neurology World Conference 2026

Abstract

Learning and memory are dependent on two key neuromodulatory systems: the dopaminergic (DA) system, which signals reward prediction error, and the acetylcholinergic (ACh) system, which regulates attention, memory formation, and cognitive encoding. These systems together support how the brain acquires and updates information through error-based learning. Similarly, biologically-inspired machine learning (ML) models have drawn from neuroscience principles to improve efficiency, with algorithms like temporal-difference learning and reinforcement learning mirroring functions of DA and ACh. However, the extent to which these models replicate the specific mechanisms of error-based learning and memory encoding found in the human brain remains unclear. A systematic bibliometric analysis was conducted on 100 highly cited articles in neuroscience and ML sourced from Web of Science using R and RStudio’s Biblioshiny. Papers were selected based on relevance to reinforcement learning, neuromodulation, and memory. A subset of 30 high-impact papers underwent in-depth review, providing the basis for a comparative examination of how neurotransmitter functions correspond to analogous ML processes. Several key parallels emerged across the review. Biologically inspired models incorporating replay mechanisms (analogous to hippocampal memory consolidation), neuromodulatory network architectures, and spiking neural dynamics demonstrated improved continual learning, reduced catastrophic forgetting, and greater energy efficiency compared to conventional deep learning approaches. Dopaminergic reward prediction error signaling closely parallels temporal-difference (TD) learning, while ACh memory encoding parallels attention modulation and synaptic plasticity mechanisms in contemporary neural network architectures. Bibliometric analysis revealed that authorship and institutional contributions are broadly distributed across international institutions, reflecting the interdisciplinary nature of this space. AI/ML publications showed a marked surge beginning around 2013, coinciding with the rise of deep learning, and growing overlap with neuroscience publications in recent years reflects increasing cross-disciplinary convergence. Neuromodulatory systems and ML algorithms converge on shared computational strategies for adaptive learning and memory. DA encodes reward prediction errors and drives plasticity, while ACh regulates memory encoding and enhances flexibility under changing reward contingencies. These parallels guide the design of AI systems that learn continuously and flexibly, while also informing therapeutic approaches for neuromodulatory disorders. However, current models lack fully realistic neurotransmitter interactions, and the limited bibliometric scope of 100 papers constrains the generalizability of these findings. Future work should expand the dataset, incorporate newer ML models and evolving AI keywords, and develop models that better simulate complex neurotransmitter dynamics.