Brain-like approach to Machine Learning

The main aim of this project is to advance the development of hierarchical brain-like network architectures for holistic pattern discovery drawing from the computational insights into neural information processing in the brain in the context of sensory perception, multi-modal sensory fusion, sequence learning and memory association among others (Herman et al., 2017; 2018).

The focus is both on building scalable implementations and studying computational capabilities of the brain-like algorithms. Hence, the intended scope includes some aspects of scaling the proposed brain-like approach as well as comparative analyses with the corresponding state-of-the- art ML methods mainly concerned with semi-supervised and/or reinforcement learning paradigms. In addition, it is envisaged that the Bayesian description of the learning and inference mechanisms in brain-like network architectures will constitute one of key points on the scientific agenda of the project. In this regard, some progress has already been made in relation to the Bayesian interpretation of brain-plausible associative learning rule developed in the lab (Sandberg et al., 2000; Tully et al., 2014).

The application domain envisaged for these developments is very broad including decision support systems in clinical and medical diagnostics, time series prediction/classification/clustering for various types of biosignals/neuroimaging data (brain-computer interfaces, biomarkers for brain disorders, neural correlates of cognitive function etc.), simulations of autonomous agents as well as pattern recognition (classification, clustering, anomaly detection etc.) in the electrical energy sector and to some degree in automotive context. In this project, the focus will be on building e-science infrastructure for advancing brain-like algorithms development and their evaluation will be conducted on concrete problems primarily within brain data analysis but also on a selected subset of tasks in the aforementioned application domains. Both applied and core methodological aspects of this project will provide an opportunity for collaboration with other initiatives within SeRC, mainly ML and DCMD MCPs. In addition, the scientific agenda of this project is expected to create synergies with computational neuroscience research on cortical memories and plasticity in “Brain network architecture and dynamics of short- and long-term memory.”