- You must login in order to post into this group.
The brain is considered as the most complex structure known in the universe. It is studied at various scales with the ambition to understand how different levels, ranging from molecular to systems level, connect to explain complex behaviour, cognition and the difference between healthy and diseased state. As a wealth of techniques have been developed to measure both structural and dynamical neural phenomena at different temporal and spatial scales, a rapid accumulation of brain data has been observed worldwide. This trend is foreseen not only to continue but also significantly accelerate due to the worldwide big brain project initiatives. An emerging challenge is to integrate all brain data and knowledge, and hence facilitate their use for the neuroscientific community. Thus storage, handling and accessibility issues are crucial. They are largely addressed by international organizations such as International Neuroinformatics Coordinating Facility (INCF, www.incf.org). Still, accumulation and organisation of neuroscience data is not sufficient to pursue the mission aimed at enhancing our understanding of the brain in health and disease. Therefore, the European flagship initiative, the Human Brain Project (HBP), was proposed to build upon the idea that ICT promises to offer great potential to effectively exploit the available data, beyond the data infrastructure perspective, and support efforts in neuroscience. Data-driven modelling is a prime representative of such activities within the HBP. In essence, mathematical models and computer simulations facilitate the integration of heterogeneous multi-scale data and knowledge. In addition, given the sparse nature of the available brain data, predictive informatics approaches are also being developed.
Another important aspect of the cross-fertilization between ICT and brain science is concerned with the advancement of the next generation intelligent systems that mimic the brain’s network infrastructure for multimodal perception, decision making, producing complex behaviour and cognitive function, among others. The ambition is to exploit the accumulating insights into the neural systems and mechanisms underlying robust information processing capabilities in the brain to develop new algorithms, cognitive architectures and ICT systems ready to face the emerging challenges in growing fields of machine learning, data analytics, cognitive robotics, autonomous agents and neuromorphic hardware. Simultaneously, neuromorphic hardware is built to serve as a computational platform for this new class of brain-like computing paradigms.
Within the SeRC coordinated Brain-IT community, the key ongoing activities revolve around: a) data-driven comprehensive modelling of neural systems at different levels of organisation; b) building a computational brain theory through hypothesis-driven modelling; c) experimental neuroscience and data analysis; d) development of novel brain-like machine learning algorithms with diverse applications in data science; e) devising brain-like cognitive architectures for autonomous systems, and f) design of robust neuromorphic hardware implementations and deployment of technological artefacts for real-time applications of brain-like algorithms.
The specific role of SeRC Brain-IT is to provide organisational infrastructure for coordinating these research activities, and also specifically develop e-science approaches to ‘bridge’ between some of these activities, all of which are necessary for understanding the brain and utilize that for brain inspired technologies. In particular, at Stockholm University (O. Eriksson), a modeling workflow for parameter estimation when building data-driven subcellular level models is developed. At Karolinska Institutet, J. Hellgren Kotaleski’s research is aimed at developing approaches for linking such subcellular models with the cellular scales by embedding/integrating subcellular (biochemical) models into whole neuron (electrical) models. Also, methods for analyzing and comparing experimental and modelled data are developed (P. Fransson). Finally, SU/KTH researchers (A. Lansner, P. Herman) develop brain theory, build cortical network models, run large-scale brain simulations on supercomputer platforms, propose new brain-like algorithms for spatio-temporal data analysis, and work towards concrete implementation of neuronal network models on neuromorphic hardware.