A Brief Overview on the Contribution of Machine Learning in Systems Neuroscience

Background: The use of Machine Learning (ML) is witnessing exponential growth in the field of artificial intelligence (AI) and neuroscience, in particular in subdisciplines such as Systems Neuroscience (SN), as a viable alternative to the use of classical statistical techniques. The combination of this interconnection allows a more detailed study of algorithms and neural circuits that emulate core cognitive processes. ML toolbox includes algorithms that are suited to solving problems of classification, regression, clustering, and anomaly detection. Objective: The aim of the present opinion was to exemplify the contribution of ML in the field of SN in three different fields: 1) cognitive modelling; 2) neuroimaging; 3) analysis of clinical datasets. Methods: We gathered evidence from the relevant literature related to the interaction between neuroscience and AI and the impact of ML in SN. Results: ML is specifically suited to the analysis of large clinical neuroscience datasets. Experimental results in neuroscience are hard to replicate for a number of reasons, and ML may contribute to attenuating these replicability issues via the ubiquitous use of cross-validation procedures. While ML modelling is primarily focused on prediction accuracy, one of the drawbacks in ML is the opacity of various algorithms that resist intuitive understanding. Conclusion: Future avenues of research have already been traced and include increased interpretability of currently opaque ML models functioning and causal analysis. Causal analysis is intended to distinguish between spurious associations and cause-effect relationships and is a primary interest in both clinical medicine and basic neuroscience.