歐盟HBP（人腦計劃）項目組結構

來自專欄 智能科學與技術

HBP項目主要分為Subprojects、CoDesign Projects及Partnering Projects三類。首先簡要說下這個歐盟HBP：

The Human Brain Project (HBP) is building a research infrastructure to help advance neuroscience, medicine and computing. It is one of the two largest scientific projects ever funded by the European Union.The 10-year Project began in 2013 and directly employs some 500 scientists at more than 100 universities, teaching hospitals and research centres across Europe.Six ICT research Platforms form the heart of the HBP infrastructure: Neuroinformatics (access to shared brain data), Brain Simulation (replication of brain architecture and activity on computers), High Performance Analytics and Computing (providing the required computing and analytics capabilities), Medical Informatics (access to patient data, identification of disease signatures), Neuromorphic Computing (development of brain-inspired computing) and Neurorobotics (use of robots to test brain simulations).The HBP also undertakes targeted research and theoretical studies, and explores brain structure and function in humans, rodents and other species. In addition, the Project studies the ethical and societal implications of HBPs work.

研究與成果如圖：

其12類Subprojects主要如下：

Subproject 1

Mouse Brain Organisation

Human and animal brains share many characteristics, but it is more difficult for scientists to work on the former; there are experiments that cannot be done on humans for ethical reasons and so they are done on mice instead.

The human brain shares many common features with other non-human mammals. These features can be considered as basic building blocks of mammalian brain organisation. Therefore, choosing appropriate experiments to obtain strategic data that can be extrapolated to the human brain is a major goal in SP1. For this purpose, many neuroscientists suggest that the ideal experimental animals at present are rodents, because they can be manipulated to study many aspects from genes to behaviour. Furthermore, we can use relatively large numbers of animals at a relatively low cost.

The next challenge is what can be done with the data and how can it be interpreted. It seems that the most appropriate approach to better understand the brain is to link detailed structural data of the whole brain with genetic, molecular, cellular and physiological data. This integration allows the generation of models that present the data in a form that can be used to rationalise, make predictions and suggest new hypotheses to discover new aspects of the structural and functional organisation of the brain.

organised：

Work Package (WP) 1.1 Subcellular and Molecular: Many different levels of molecular data are required to understand the function of single cells, circuits and brain function. This WP will therefore generate, obtain and integrate data at various subcellular and molecular levels. Since the initial phase of the Project, rapid technological advances have enabled this type of data to be analysed at in much greater detail, and some of these methods have meant that diversity across many cells, brain regions and the whole brain can be mapped.

WP1.2 Cell and Microcircuitry: This WP aims to carry out novel analyses in order to generate the data needed for validated high-fidelity brain models. The work concentrates on four major brain regions: the neocortex, hippocampus, basal ganglia, and cerebellum. The data generated include the number and spatial distribution of neurons, glia, and specific types of synapses, as well as correlation between morphology and physiology.

WP1.3 Whole Brain: We propose to go beyond the state of the art by investigating meso-scale (millimetres to centimetres) multilevel maps of the mouse brain through an integrated view of anatomy and functionality. In terms of anatomy, the intention is to determine the spatial distribution of different cell types, based on the expression of certain proteins, across the entire brain, and to refine the maps produced according to different neuronal types. This will be complemented by imaging of cortical functionality, investigating the functional connectivity involved when specific tasks are performed.

WP1.4 Integration of Micro-Anatomical Data: The various datasets produced by SP1 will be integrated in this WP; integration of external datasets that complement the core data produced by SP1 will also take place. Statistical modelling will reveal relationships between the various datasets that will be crucial for modelling and simulation activities in other HBP Subprojects such as SP4 and SP6. We will also use statistical and machine learning techniques to deduce principles of neuron morphology and neuroanatomical organisation.

WP1.5 Management and Scientific Coordination: This WP aims to ensure that work within SP1 is carried out according to the planned objectives and to coordinate HBP research on strategic mouse brain data, ensuring that the work is efficiently organised and documented and that the research contributes to the overall HBP goals.

Subproject 2

Human Brain Organisation

brain consists of around 86 billion nerve cells, the neurons. Each neuron is connected to between 1 and 200,000 other neurons, resulting in 100 trillion nerve fibres running through the brain. This high complexity is further increased by the inter-individual variability of the brain - each of us is different, has different talents, feelings and their own personality. How the brain is organised, which brain regions are interconnected and which areas work together to execute a certain function is by far not fully answered.

SP2 researchers from 14 research institutions throughout Europe are working on these questions by using many different methods. For example, researchers of the Forschungszentrum Jülich in Germany are working on a map of the human brain (JuBrain atlas) based on differences in the distribution and size of the neurons in brains of body donors. Connections between the neurons, i.e. nerve fibres and bundles can be detected by powerful methods such as polarized light imaging (PLI) and diffusion tensor imaging (DTI), where experts in Jülich and at the CEA in Paris collaborate. Functional magnetic resonance imaging (fMRI) is used by groups in the Netherlands, France, Belgium, UK and Germany to identify regions and networks involved in brain functions like visual and auditory processing or cognition, and reveals more and more details of the functional brain organisation.

One of the greatest goals of SP2 is to develop the HBP Human Brain Atlas, which can be used by neuroscientists all over the world, in neurosurgery and as a basis to understand the differences between the healthy and diseased brain.

organised：

Work Package (WP) 2.1 Human Neurogenomics. This WP aims to understand brain function by correlating genetic variability with brain phenotypic variability in humans. Another aim is to deliver fundamental sets of biological information (DNA, DNA methylation, RNA) for all HBP brain samples by generating 「-omics」 data (e.g. whole genome sequence, whole genome DNA methylation, transcriptome derived from brain regions and single cells).

WP2.2 Morphology and Architecture of the Human Brain. A Multi-Level and Multi-Modal Approach. Here, we study the organisation of the visual, motor and limbic systems and areas of the allo- and neocortex. In addition, participants exploit the advantages of state-of-the-art genetic, cell biological, histochemical, and imaging methodologies in cell typing, fibre orientation (「connectivity」) and protein distribution across brain regions.

WP2.3 Function and Variability. This WP aims to build a holistic view of the brain macroscopic organisation that includes anatomical and functional information gained through in vivo imaging.

WP2.4 Comparative Computational Architecture of Multi-Modal Processing Streams (Systems Physiology). We investigate representations and mechanisms at the different organisational levels in the human and monkey brain, e.g. to advance our understanding on how the brain processes stimuli and how information is integrated within and between visual and auditory processing streams during cognitive tasks.

WP2.5 Integrative Maps and Models. This WP aims to build the multimodal connectome of the human brain to be embedded in a multi-level human brain atlas.

WP2.6 Co-Design/methods and Big Data Analytics. We aim to provide scientists with the tools to project atlas information on to their own datasets, and offer retrieval of quantitative data based on 3D coordinates, macroanatomical landmarks or structural areas.

WP2.7 Coordination and Management. This WP aims to coordinate the scientific activities within SP2 and its interactions with other SPs, the entire HBP, and the larger community.

Subproject 3

Systems and Cognitive Neuroscience

The goal is to uncover neural mechanisms underlying cognitive processes, such as learning, multisensory integration, perception, sleep, consciousness, and associated systems phenomena. The results provide the constraints for the development of computational models of cognitive and systems-level processes that will be implemented in robots and neuromorphic computing systems. SP3 addresses these issues at multiple study levels (cells, groups, networks, brain systems) and works to unify different disciplines.

One of the deepest unsolved problems in science is the nature of consciousness. How is consciousness generated by the brain? Current limitations in answering this question cause several clinical and ethical problems, such as assessing the level of consciousness in patients following brain injury. Novel ways to measure consciousness levels will make us less dependent on purely behavioural measures, which will benefit, for instance, minimally conscious patients.

Similarly, how can disparate phenomena such as sleep and wakefulness emerge from the same cortico-thalamic systems in the brain? To answer this question, we will investigate slow-wave activity and simulations of large populations of firing neurons in mice and humans.

SP3 also investigates brain mechanisms of memory. Episodic memory is the memory of our personal, conscious experiences set within space and time. It defines who we are. The brain』s ability to recall objects and experiences from multisensory information, such as vision, audition or touch sensation is a key to understanding memory in humans and animals. We will conduct a coordinated series of experiments to identify the precise neuronal mechanisms behind episodic memory, and validate them by computational models and robotic systems.

In summary, by its cross-disciplinary approach to multiple levels of neural and brain organization, we will work to elucidate mind–brain relationships that have eluded explanation for centuries.

organised：

Work Package (WP) 3.1 Context-Sensitive Multisensory Object Recognition. This WP aims to develop a sophisticated understanding of extensive neural interactions and subsequently develop models incorporating information processing in a realistic theory of how we recognise objects within certain contexts.

WP3.2 Wave Scaling Experiments and Simulations. This WP focuses on slow wave activity (SWA), and asks such questions as how SWA changes when the brain state changes.

WP3.3 Episodic Memory as Multisensory Reconstruction. We conduct experiments to identify the precise neuronal mechanisms behind episodic memory (recollections of personal experiences), validate them using computational models and robotic systems, and test how they fail in old age and dementia.

WP3.4 Experimental and Computational Exploration of Consciousness Mechanisms and Methods in Mice and Humans. We aim to: a) test and improve physiological methods for assessing consciousness; and b) contribute to understanding the nature of consciousness by testing relevant theories. This will be done using experiments to test principles and ideas in mice, and then using the insights from these and from computational modelling to test and develop better non-invasive methods in humans.

WP3.5 Scientific Coordination, Project Management and Communication. This WP coordinates inter-WP work within SP3 and its links with other SPs, the entire HBP and the larger neuroscience community. Its project management and communication brief covers quality assurance, ethics reporting, innovation, outreach, and public engagement.

WP3.6 SP3 Contributions to Co-Design Projects and Infrastructure. This WP supports the HBP Co-Design Projects and Infrastructure coordination beyond that of WP3.5.

Subproject 4

Theoretical Neuroscience

Their researchers in Theoretical Neuroscience work to simulate and capture key biological processes using mathematical models in order to try to better understand the brain。

Theoretical Neuroscience is a link between experimentalists and technology. Brain mechanisms identified in the experimental HBP Subprojects are formalised into mathematical models, which are then made available to the HBP Platform Subprojects

Theoretical Neuroscience is needed for linking scales, another fundamental aspect of brain exploration. Scientists investigate the brain at multiple levels, from the microscopic (synapses, neurons), through mesoscopic (brain circuits) to macroscopic scales (areas of the brain), and each method of investigation has its own specific scale, e.g. single-neuron recordings, imaging methods such as local field potential (LFP), up to large-scale imaging such as EEG, fMRI, etc. One needs theoreticians not only to understand how these signals are generated, but also how to link them together

We also investigate key cognitive mechanisms, such as sensory processing (vision, auditory), learning and memory, spatial navigation or sensorimotor coordination through computational and mathematical models

Our researchers use tools such as Python, Brian, NEST, etc., and ensure compatibility with the HBP Platforms. We use the same software environment, pyNN, as the one used on the Neuromorphic Computing Platform so that the program code of many of our models developed can be directly implemented to simulate neural networks

Many of our results obtained are open-access, and are at the core of the discussions held at the events of the European Institute for Theoretical Neuroscience

organised：

WP4.1 Bridging Scales. Aims to provide models linking different scales of investigation, including linking models over spatial scales, such as synapses (μm) single-cell (tens of μm), local network (mm) or whole brain region (cm), as well as linking between models of different levels of complexity (e.g. detailed vs. simplified neuron models). We also develop models of the different brain signals accessible experimentally, across different scales.

WP4.2 Generic Models Of Brain Circuits. Aims to provide theoretical methods for large-scale simulations with generic network models. We will develop simplified large-scale models of specific brain areas and investigate the methodological aspects of model building.

WP4.3 Learning and Memory. Aims to formulate synaptic plasticity algorithms from experimental data. We also aim to develop models of learning and reward, compatible with neuromorphic systems, and develop models of behavioural learning and long-term memory in the brain.

WP4.4 Models of Cognitive Processes. Aims to develop models of elementary cognitive processes that consider the different levels, from the biophysical level 「network states」 to sophisticated functions such as spatial navigation or decisionmaking.

We aim to contribute to a multi-scale brain theory and develop large-scale models of cognitive functions that will bridge 「high-level」 behavioural and imaging data and detailed multilevel models of brain physiology.

WP4.5 Linking Model Activity And Function To Experimental Data. Aims to link theoretical models at different levels of description to create bridges between neuroscience and the models implemented in various HBP Platforms. This will involve mathematical principles and theoretical methods to integrate neuroscience data into models and compare the results with the existing data.

WP4.6 The European Institute For Theoretical Neuroscience. Located in Paris area, the European Institute for Theoretical Neuroscience (EITN) aims to serve as an incubator of ideas and foster the exchange of ideas between theoreticians and experimentalists, and is open to researchers from the field worldwide, whether they are HBP Partners or not.

WP4.7 Scientific Coordination. Coordinates and monitors the scientific activities of SP4, and the interactions with the other SPs.

Subproject 5

Neuroinformatics Platform

Understanding the brain requires that huge amounts of complex data collected at many levels of investigation and with a multitude of methods be combined. Such data integration can be compared to a multi-dimensional puzzle consisting of data about genes and molecules, cells, connections and networks, regions of and the whole brain, plus cognition and behaviour. Fitting data together from this puzzle into meaningful information – a gigantic challenge - is one of the main goals of the Human Brain Project (HBP).

We provide the informatics tools and services that make data integration in HBP possible. Data from the brain produced in other Subprojects are organized and managed, then made available for collaborative use by all researchers in HBP and the wider neuroscience community. The online web services that will make this possible are being established stepwise and we will demonstrate several of the tools and services that are currently in place.

SP5 tools are used to tag the data with important additional information. Such 「metadata」 explain what the data are about, how they are collected, where in the brain they are from, and what they represent. This will make it possible to search and find data, aimed at advanced analysis of new combinations of data. Data are stored at the supercomputer centres in Europe managed by SP7, and the same computing services will embed analytical tools and workflows from SP5, providing increased capacity and capability for analysis of brain data, feeding information into the computational modelling and simulation of the brain taking place in other HBP Subprojects.

organised：

Work Package (WP) 5.1 Data Curation Support Lab. We aim to make HBP data and models discoverable and accessible (already pledged to the HBP community), via metadata enrichment and storage in a federated data infrastructure. Users can curate and share data or models with other HBP researchers in the open data domain and reach high data consistency levels.

WP5.2 Multi-level Atlas of the Rodent Brain. This WP aims to integrate heterogeneous multi-level rodent brain data in common reference atlases, and provide services for exploration, enrichment and analysis. Spatially anchored data, organized in the KnowledgeGraph, can be shared with the research community for use in data mining and predictive neuroinformatics.

WP5.3 Multi-Level Atlas of the Human Brain. This WP does for the human brain what WP5.2 does for that of the mouse.

WP5.4 Data and Atlas Curation Tools. The aim of this WP is to deliver a suite of tools for data curation, spatial integration or 2D and 3D viewing of multi-level human/rodent brain image data.

WP5.5 Community-Driven Neuroinformatics Platform and Infrastructure Operations. We aim to develop and operate the service infrastructure required for an engaging, community centric, multi-level, multi-modal open data ecosystem.

WP5.6 Data Mining and Analysis Neuroinformatics Capabilities. This WP brings learning-based image analysis to HBP neuroscientists by creating the ilastik toolkit, a simple, user-friendly tool for interactive image classification, segmentation, and analysis of neural images.

WP5.7 Tools and Curation for Integrated Parallelized Analysis of Activity Data. We aim to enable users to work with dynamic functional data from experiments or simulations by providing tools and services to integrate and analyse activity data, e.g. those from neuron spiking.

WP5.8 Management and Coordination. This WP coordinates SP activities and maintains an efficient and proactive relationship with other SPs and the broader science community, including securing integration of the Neuroinformatics Platform into other relevant services.

Subproject 6

Brain Simulation Platform

The Brain Simulation Platform (BSP) is an internet-accessible collaborative platform designed for the digital reconstruction and simulation of brain models. Researchers can access the BSP to reconstruct and simulate models of the brain at different levels of detail to study their structure and function. In June 2017, the second version of the BSP was released. This new version is more user-centric and user-friendly, so that users with different levels of neuroscientific or technical expertise can benefit from the BSP for their research or just for curiosity!

Along with the new version of the BSP, we released the first MOOC (massive open online course), on simulation neuroscience. There will be three MOOCs in total that will teach participants how to use the state-of-the-art modelling tools of the BSP to simulate neurons, build neural networks, and perform simulation experiments. In the first MOOC, you can learn how to digitally reconstruct a single neuron. The first run of the first MOOC has finished, but the self-paced course will be available again in the autumn. If you are interested in our goal to reconstruct and simulate the brain, do follow the MOOC!

organised：

Work Package (WP) 6.1 Subcellular and Molecular Modelling. This WP creates multiscale scaffold models of key signalling cascades in neurons for use in bootstrap data-driven community modelling efforts. It uses inputs from SP1, SP2 and molecular dynamics to parameterise subcellular models and it integrates models of key signalling cascades into single neuron models, which should make it possible to model neuromodulation and plasticity.

WP6.2 Cellular-Level and Whole-Brain Modelling. We build cellular level models of target areas of the rodent brain and a point neuron Whole Mouse Brain Model, adapting tools and workflows developed for the somatosensory cortex for use in other brain regions (cerebellum, hippocampus, basal ganglia), and the whole mouse brain. In parallel, we will continue the exploratory modelling of human neurons started in the Ramp-Up Phase.

WP6.3 Reconstruction and Simulation Tools. This WP is building tools and workflows for data-driven reconstruction and simulation of brain models at different levels of biological organisation, exploiting data available through the Neuroinformatics Platform (SP5). The tools, which will facilitate building of scaffold and community models elsewhere in the SP, will include a Hodgkin-Huxley Neuron Builder and tools for in silico experimentation.

WP6.4 Brain Simulation Platform. We design, implement and operate the HBP Brain Simulation Platform. This comprises a collection of Apps, APIs and Platform Foundation Software, which support collaborations to build, simulate, analyse, validate and disseminate data-driven brain models. Part of the software underlying the Platform will be developed in WP6.3; other parts will come from efforts of WP6.1 and WP6.2 and from community activities. We will seed the development of Apps and APIs for a subset of this software, which has reached a high level of maturity. In the medium term, it is expected that most Apps will come from the community.

WP6.5 Coordination and Community Outreach. This WP coordinates the work of the SP6 WPs and their interaction with the HBP management, other SPs and the wider community.

Subproject 7

High-Performance Analytics and Computing Platform

One of today』s science challenges is to understand how our brains work. Usually new endeavours require new tools and technologies to get to the next level. Neuroscientists in the Human Brain Project (HBP) collect a lot of data, develop models based on this data that try to explain how mechanisms in the brain work, and finally they simulate these models. In these simulations, neural networks (parts of the neurons and their connections in the brain) are built in the computer, get some input stimulation (like the actual brain gets input from our senses) and then these 「digital」 neurons and neural networks react to it. The scientists analyse the simulation results and compare them to experimental data to improve their models.

We support neuroscientists to do this research by developing the tools and technology that they need for it. We make huge storage available at four centres in Europe to store the data. The storage at one site alone would be enough to store 3–4 billion books or for 250–300 years of high definition movies. At the same centres, there are also supercomputers, which are among the most powerful computers worldwide. The human brain is so complex that a normal computer in the scientist』s office is not enough to simulate even a fraction of the human brain. One of the supercomputers is as powerful as about 350,000 standard computers.

Our job in the HBP is not only to make this hardware available to the scientists, but also to develop software that supports neuroscientists in their endeavour, e.g. to manage their huge datasets, to simulate models most efficiently on the supercomputers (getting better results as fast as possible) or to look at 「visualisations」 of the datasets. A visualisation turns the columns of numbers produced by a simulation into a graphical representation like pictures or even 3D objects.

organised：

WP 7.1 Simulation Technology. Here, we develop the concepts, numerical algorithms, and software technologies for the implementation of new features of the simulation codes of the HBP and focus on the extension of the functionality of the simulation engines.

WP7.2 Data-Intensive Supercomputing. We aim to link extreme scale data processing challenges to the exploitation of scalable computer resources. Our work is driven by specific use cases coming from different areas of the HBP to ensure that R&D work is guided towards enabling infrastructure for future neuroscience research.

WP7.3 Interactive Visualisation. We aim to develop a software infrastructure for timely user-centric visual data analysis for the HBP using new and existing visualization tools.

WP7.4 Dynamic Resource Management. We seek to combine different tools and techniques to achieve a novel approach to dynamic resource management in high-performance computing facilities, therefore having a direct impact on how neuroscience applications are executed.

WP7.5 High-Performance Analytics and Computing Platform. We are responsible for HBP high-end computer and data analytics Platform services, from HBP internal and external sources, as the HBP Research Infrastructure becomes operational.

WP7.6 Management and Coordination. We manage the High Performance Analytics and Computing Platform. Our objective is to coordinate and validate the technology and infrastructure development in the Subproject, ensuring that the work is aligned with the overall HBP objectives, meets actual user needs and is efficiently organized and documented.

Subproject 8

Medical Informatics Platform

Our goal is to provide a collaborative open source platform, the Medical Informatics Platform (MIP), that allows researchers worldwide to share medical data, enabling the use of machine-learning tools for brain-related diseases, while strictly preserving patient confidentiality.

A combination of medicine and computer science, we aim to break down barriers between patient care, brain science, and clinical research to minimize the delays involved in diagnosis of brain diseases and institution of the most effective treatments.

As in many IT contexts, data security is treated as a top priority, the need for which is made all the more pressing due to the long-standing commitment of the medical profession to patient confidentiality. The Subproject aims to preserve hospital ownership and control of data by developing a federated query engine within the hospitals, leaving patient data in its original location and format. This is a fundamental change, compared to traditional schemes in which data are moved to accommodate the needs of the query engine. The research team is also developing techniques to ensure that it will not be possible to infer personal information about patients from query results while performing advance machine learning analytics.

We urgently need better diagnostic tools and treatments for brain-related diseases. People are living longer, thanks to improved sanitation, nutrition and treatments for infectious diseases. As a consequence, chronic diseases – which include most brain-related diseases – form the largest part of the overall health burden, and their share is growing. In 2010, the annual cost of managing the 500 or so brain disorders, from migraine through degenerative diseases like Alzheimer』s and Parkinson』s, to conditions like autism, were estimated at EUR 800 billion in the EU alone, and that cost will continue to increase.

organised：

WP8.1 Federated Clinical Data Infrastructure (FCDI). This WP develops tools for harmonizing heterogeneous clinical databases; tools for data anonymization, ontology based query interfaces, and federated search and intensive distributed analysis of clinical data. These form the Hospital Bundle that will run at every hospital participating in the MIP.

WP8.2 Data Selection and Community Engagement. We coordinate data selection across the MIP federated network, develop unified data governance and ensure that quality data are delivered securely, and liaise with participating hospitals and users (clinicians and researchers).

WP8.3 Data Features, Tools and Biological Signatures of Disease. This WP uses all available hospital and medical datasets to identify and define replicable disease signatures. The analysis of data from different domains will require the tailoring of existing tools and the development of new analysis tools and methods.

WP8.4 Theory, Disease Models & Big Data Engineering. Here, we provide a theoretical framework for querying, analysing and integrating existing data into a model-based approach for discovery of brain disease signatures. It builds on data features obtained in WP8.1-8.2 and uses mathematical and statistical methods developed in WP8.3 to predict disease characteristics.

WP8.5 The Medical Informatics Platform. We provide online evidence-based medicine tools via a web portal to users such as neuroscientists, computational scientists, epidemiologists, and the pharmaceutical industry. Its analyses integrate open-access research data repositories with brain disease features generated in participating hospitals using MIP tools.

Subproject 9

Neuromorphic Computing Platform

The Neuromorphic Computing Platform takes two fundamentally different paths to support scientific research and applications.

The BrainScaleS system, based in Heidelberg, employs a mixed signal approach employing analogue electronics to model 4 million neurons and 1 billion synapses, as well as their connections and intercellular communications, using digital communications protocols. It is targeted to the emerging field of bio-inspired AI as well as a better understanding of the learning and development in the brain. The system is a direct, silicon-based image of the neuronal networks found in nature and runs 10,000 times faster than its biological archetype, allowing a day of biological development to be compressed into 10 seconds.

The SpiNNaker system, based in Manchester, is a massively parallel computing platform, targeted towards neuroscience, robotics and computer science. For robotics, SpiNNaker provides mobile, low power computation, and makes possible the simulation of networks of tens of thousands of spiking neurons, as well as processing sensory input and generate motor output, all in real time and in a low power system. The system is unconventional in that SpiNNaker nodes communicate using simple messages (spikes) that are inherently unreliable. This break with determinism not only offers new challenges, but also the potential to discover powerful new principles of massively parallel computation.

Both approaches are involved in technological next generation development (both hardware and software) and integration. The establishment of principles for brain-like computation, computational capabilities through learning and large scale organisation of cognitive computation are also focuses of interest. Outreach activities include user training, support, and coordination for effective application of the Platform.

organised：

WP 9.1 Platform Software and Operations. This WP operates the Neuromorphic Computing Platform constructed in the HBP Ramp-Up Phase, maintains and further develops the software methods and tools required for the neuromorphic hardware systems, and integrates them with the HBP Collaboratory, other HBP Platforms and, where possible, external resources.

WP9.2 Next-Generation Physical Model Implementation. We develop, prototype, manufacture, assemble, test and operate next-generation hardware systems to implement massively parallel, physical models of brain cells, circuits and networks. These prototype neuromorphic physical model (NM-PM) systems will be integrated in future versions of the Neuromorphic Computing Platform.

WP9.3 Next-Generation Many Core Implementation. This WP develops, prototypes, manufactures, assembles, tests and operates next-generation hardware systems to implement massively parallel, many-core implementations of brain cells, circuits and networks. These prototype neuromorphic many-core (NM-MC) systems will be integrated in future versions of the Neuromorphic Computing Platform.

WP9.4 Computational Principles. In this WP, we use brain activity and plasticity data to develop principles that enable brain-like computation, cognition, and learning in neuromorphic systems. This work supports emulation of specific brain functions or cognitive processes in existing neuromorphic hardware and will guide the design of next-generation neuromorphic systems.

WP9.5 Platform Training and Coordination. This WP provides training and documentation for SP9』s various neuromorphic systems and ensures that these are accessible via the Neuromorphic Computing Platform. It also coordinates the SP』s R&D work and its integration with the rest of the HBP.

Subproject 10

Neurorobotics Platform

The human brain is one of the most astonishing, complex and tremendously powerful creations of nature. But what makes is so efficient, so flexible, so intelligent? We know it』s not just the sophisticated 「design」, but also the ability to constantly learn. The brain makes the body perform an action, then the body perceives the results of this action, and finally the brain interprets the results and changes its behaviour accordingly, so that the next action can be more effective.

What we do in our SP is create an opportunity to give any simulated brain model its own robotic 「body」 — virtual or even real — that can make it feel as if it had a real body, capable of performing appropriate actions, gathering perceptions and learning. Our team has also developed tools to create very detailed simulated environments — 「virtual realities」 — in which to test brain models and robots. Our tools for creating virtual robots and environments can be found in our Neurorobotics Platform, which is public, online and available to all researchers who wish to test their brain models or build future brain-inspired robots.

The Neurorobotics Platform is constantly evolving, thanks to inputs from researchers from all over the world, and our team is collaborating with them to help them implement their experiments using the Platform. The experiments include the iCub humanoid robot balancing a ball towards the centre of a board that it holds in its hand or the NeuroSnake robot faced with a similar challenge. They show that, in such simulated environments, robots equipped with the ability to perceive their surroundings can construct their own effective and powerful learning rules, almost like living creatures. Moreover, we are in the process of creating a complete virtual mouse, with eyes, whiskers, skin, a brain and a body, with bones and muscles that function like those of its natural counterpart. We also see tremendous potential in using simulated brain-inspired systems, such as the 「virtual mouse」, in neuroscience and medical research.

organised：

WP10.1 Closed-Loop Experiments (Data-Driven Brain Models). This WP develops strategic closed-loop experiments chosen to reconstruct the sensory motor loops (between brain and body models) for the most important systems.

WP10.2 Closed-Loop Experiments (Functional/Control Models). This WP focuses on top-down models of sensory-motor processing to control virtual and physical robots.

WP10.3 Components of Closed-Loop Experiments. This WP provides all components (robots, bodies, sensors, actuators, environments) needed in the closed-loop experiments of WPs 10.1 and 10.2 as well as in the Co-Design Projects.

WP10.4 Translational Neurorobotics. This WP translates virtual robots and brain-derived controllers to physical prototypes to contribute to future robotics research. Unlike WPs 10.1 and 10.2, WP10.4 tries to transfer results from brain research and neurorobotics to future robotics.

WP10.5 Simulation and Visualization Tools for Neurorobotics. This WP aims to build a suite of tools and workflows to plan, run and analyse in silico experiments with Neurorobotics Systems.

WP10.6 Neurorobotics Platform. This WP is building a collection of Apps, application programming interfaces and Platform foundation software that supports collaborative approaches to building and sharing neurorobotic models.

WP10.7 Scientific Coordination and Community Outreach. This WP coordinates quality assurance, organisation of meetings and workshops, as well as reporting, within SP10 and its interactions with the wider scientific community within and outside the HBP.

Subproject 11

Management & Coordination

SP11 supports HBP decision-making, operates the management structure and European Research Programme, ensures transparency and accountability toward funders and stakeholders, and maintains standards of quality and performance. Its primary responsibilities include coordination of the scientific roadmap, and in particular the supervision of the Milestones and Deliverables for the HBPs ICT Platforms. Other areas include:

?Coordinating the HBPs governance, leadership and decision making mechanisms, ensuring balanced representation for stakeholders while remaining lean enough to promote decisive action;

?Monitoring the Projects performance and ensuring that all governance, management, and administrative processes run smoothly;

?Providing the Project and the HBP Consortium with centralised support for administration, IT Services, media and communications, innovation and technology transfer, and science and technology coordination;

?Developing a framework for collaboration;

?Designing and launching the HBP Competitive Calls Programme;

?Designing and coordinating a programme of transdisciplinary education, training young European scientists to exploit the convergence between ICT and neuroscience, and creating new capabilities for European industry and academia;

?Providing administrative management for projects selected through the HBP Competitive Call.

?Coordinating the projects gender equality activities.