TALKS

Shaping the unfolding experience into a memory code.

Lluis Fuentemilla 
University of Barcelona, Cognition and Brain Plasticity Unit. Bellvitge Biomedical Research Institute (IDIBELL), ES

Although our ability to transform the experience of an event episode into a permanent memory trace stands out as one of the most necessary of mental processes, the underlying brain mechanisms still remain poorly understood. This talk will outline emergent findings emphasizing the role of memory reactivation as a brain mechanism that helps address this computational challenge. The talk will cover research on healthy population and neurological patients, combined with EEG/MEG recordings, to guide and document the relevance of memory reactivation during working memory maintenance, memory recollection, and sleep consolidation. Leveraged by the possibility that memory reactivation emerged as a mechanism bridging early stages of memory transformation and long-term memory representation, the talk will finally highlight recent data indicating that the rapid memory reactivation of the just encoded sequence of episodic events influences how experience is carved and represented in long-term memory.

Towards understanding consciousness in computational systems.

Joscha Bach
Harvard Program for Evolutionary Dynamics, US

Currently, we live in what may turn out to be the second wave of Artificial Intelligence. First order AI identified functionality that requires intelligence, and implemented it directly, in the form of human-constructed algorithms. Second order AI builds algorithms that discover the functionality by themselves. These learning systems approximate compositional functions (Deep Learning) and treat intelligence as the ability to make models, i.e. functions that relate observables to each other. Most Deep Learning consists in training artificial neural networks with stochastic gradient descent, but we already see a number of ways in which we can generalize towards other methods of optimization and function approximation. The limits of Deep Learning are less obvious than one might think, but most researchers don’t believe that it will carry us all the way to human-like intelligence and beyond. It is likely that before that, we are going to see third order AI, which is meta-learning, i.e. algorithms that discover functions that learn. It is tempting to see our brains not as learning systems, but as meta-learning systems, and evolution as a (slow and unprincipled) search for meta-learners. Perhaps we will even see a fourth order of AI, which would be the general theory of search.In this lecture, I would like to discuss what distinguishes current approaches and thinking in AI from minds, and how we can conceptualize the most puzzling mental functionality such as attention, emotion, perception, consciousness, intentionality, purpose, meaning and sociality within the frameworks of artificial intelligence.

Insights on the Self from the Spatiotemporal Dynamics of Acute Inflammation In Vivo and In Silico. 

Yoram Vodovotz.
Pittsburg University, US

Acute inflammation is a highly dynamic phenomenon, regulated in part through neural circuits, which ensues as a response to infection or injury, and can be mimicked experimentally by the administration of agents such as Gram-negative bacterial lipopolysaccharide (LPS). Lipopolysaccharide induces an acute inflammatory response across multiple organs, primarily via Toll-like receptor 4 (TLR4). As part of a longstanding effort to measure, model, and modulate the acute inflammatory response, we sought to define novel aspects of the complex spatiotemporal dynamics of LPS-induced inflammation using computational modeling. An analysis of principal drivers of LPS-induced inflammation in the heart, gut, lung, liver, spleen, and kidney to assess organ-specific dynamics, as well as in the plasma (as an assessment of systemic spillover), was carried out using data on 20 protein-level inflammatory mediators measured over 0-48h in both C57BL/6 and TLR4-null mice. Using a suite of computational techniques, including a novel time-interval variant of Principal Component Analysis, we confirm key roles for cytokines such as tumor necrosis factor-α and interleukin-17A, define a temporal hierarchy of organ-localized inflammation, and infer the point at which organ-localized inflammation spills over systemically. Given the known role for neural circuits in regulating inflammation, we hypothesize that systemic spillover occurs once neural control of local inflammation fails. We further hypothesize that there is a TLR4-regulated threshold for neural control. Taken together, our results point to the value of employing a systems biology approach to obtain a novel perspective on the time- and organ-specific components in the propagation of acute systemic inflammation and raise important questions regarding the role of the brain in the spatiotemporal regulation of this vital process.

Dynamics of prefrontal neurons during decision-making tasks.

Encarni Marcos
Instituto de Neurociencia CSIC, Alicante, ES

We continuously need to coordinate multiple processes to accomplish our current goals. By means of its connection with multiple other areas of the brain, the prefrontal cortex (PF) plays an important role in this matter. Neurons in PF have shown to represent task-relevant information and to maintain in memory the current goal until the proper action can be selected and executed. In this talk, I will present our recent results on the coding of different task variables by prefrontal neurons and on the activity reconfiguration of the prefrontal network during the transition from the mnemonic representing of a goal to its use for action selection. I will also present a computational model that accounts for the heterogeneity of prefrontal activity and that is able to produce both stability and susceptibility needed for persistent activity during memory and for flexibility to encode the required behavioural output, respectively.

What is a brain system, what are the brain’s systems, and how do they work together?

Bjorn MerkerTutorialIt would seem obvious to anyone with even passing acquaintance with the brain that it is composed of a set of subsystems which together accomplish the many complex tasks with which the brain is charged. But what exactly are those systems, how do we find them, and how do we determine their place in the brain’s over-all functional division of labor? The difficulty here is that such functional assignments are theory-dependent. Different commitments regarding what kind of machine the brain in fact is will promote different conceptions of a given structure’s or system’s functional role. Consider the hippocampus. If one thinks that the brain employs a “temporal code” – say in the frequency domain – as an essential operational mechanism, the conspicuous hippocampal oscillatory activity in the theta frequency range will likely be taken as the key to its functional role. If instead, one is committed to spatial mapping as the brain’s basic mode of representing information, hippocampal place fields are readily interpretable as elements of a high level (cognitive, sparse) map serving navigation in familiar circumstances and associated memory functions. In this tutorial we will discuss a number of issues related to this over-arching question of what kind of machine the brain, in fact, is as a framework for understanding it at the systems level, using a number of functional issues and anatomical arrangements as exercises and illustrations.

Exploiting the hypodopaminergic state with Transcranial Magnetic Stimulation in addiction. 

Marco Diana
Universita’ degli Studi di Sassari, IT

Repetitive Transcranial Magnetic Stimulation (rTMS) of the dorsolateral prefrontal cortex may affect neuro-adaptations associated with alcohol addiction, potentially influencing drug craving and intake. Previous pre-clinical and clinical evidence suggests a tonically reduced functioning of the mesolimbic dopamine system leading to hypothesize that ‘boosting’ the hypofunctional system may yield clinical benefits. Here we show that rTMS reduces alcohol and cocaine intake in alcoholics and cocaine addicts. We investigated alcohol intake and dopamine transporter (DAT) availability by Single Photon Emission Computed Tomography (SPECT) in the striatum, in Alcohol Use Disorder (AUD) patients before and after deep rTMS.  The study results suggest a modulatory effect of deep rTMS on dopaminergic terminals and a potential clinical efficacy in reducing alcohol intake in AUD patients.
Similarly, a study with cocaine addicts (DSM-V) receiving bilateral repetitive TMS r(TMS) stimulation showed a lasting reduction of cocaine-intake significantly more in 10 Hz treated patients vs. SHAM. While further studies are required to confirm these encouraging, preliminary findings they support the notion that DA can be considered a useful biomarker to be targeted by rTMS in addicts.

 Building brains with iqr.

Ulysses Bernardet Aston University in Birmingham, UKTutorial

iqr is a graphical environment for building and simulating neuronal models. Uniquely, iqr is geared towards large-scale, architectural models that can be interfaced with physical sensors and robots. In this tutorial, we will introduce the concepts behind iqr and do hands-on practical exercises with the software.  At the end of the tutorial, participants should understand the basics of constructing iqr systems, and be able to build and run their own neuronal simulations.

Motivation and emotion in AI systems.

Joscha Bach
Harvard Program for Evolutionary Dynamics, US

Current AI research mostly focuses on learning, which is the creation of models that maximize the accuracy of expectations (unsupervised learning), the accuracy of classifications (supervised learning) or the expected reward (reinforcement learning). Here, we are discussing the nature and architecture of rewards in humans, with respect to their translation into AI systems. We will look at physiological, social and cognitive rewards, how to harness them to control behavior and how to even understand and model personality. We will also discuss emotion as a modulation of cognition, including the different types of emotional modulators, and their integration into a cognitive architecture.

Setting the bar for conscious will. 

Aaron Schurger
INSERM, Neurospin, FR

Over the past four decades, a handful of highly influential studies have produced evidence that appears to be incompatible with the possibility of conscious volition – the notion that our conscious
intentions are the causes of our actions. These studies have generated a great deal of controversy, in part because empirical approaches to this question remain problematic. What kind of “evidence” do studies like these provide, once we get beyond the sound bites and look closely at the relevant details in a clear and methodical way? The age of multivariate decoding and brain-computer interfaces opens up new avenues for research on volition and the initiation of voluntary movement, making this question even more pressing. The bottom line is
this: If by looking at your brain activity, I can reliably predict your decisions before you are conscious of having made them, then I call into question whether or not you possess “conscious will.” But what does it mean to “predict” behavior and how can we evaluate the ability to do so? How, in general, are we to evaluate evidence that purportedly bears on the question of conscious will? In this talk, I will cover the key studies that have defined this field of research as well as more recent work that has called into question some of its assumptions. Through a critical analysis of prior work, we will try to establish some criteria for evaluating evidence that bears on the question of conscious volition.

Tactual Curiosities. 

Vincent Hayward. 
School of Advanced Study, University of London and ISIR, Sorbonne Université, Paris.

The astonishing variety of phenomena resulting from the contact between fingers and objects can be regarded as a formidable trove of information that can be extracted by organisms to learn about the nature and the properties of objects. This richness is likely to have fashioned our somatosensory system at all levels of its organisation, from early mechanics to cognition. The talk will illustrate this idea through examples and show how the physics of mechanical interactions shape the messages that are sent to the brain; and how the early stages of the somatosensory system en route to the primary areas through the lemniscus pathway are organised to process these messages.

Internal models and counterfactual cognition predicting our environment.

Lars Muckli
University of Glasgow, UK.

Historically, the cognitive turn was the shift away from behaviorism – the understanding that brains create a rich internal world that performs at will many different specialized tasks and shifts between them. The predictive coding framework departs from the divide of perception, cognition, and action, and provides a unifying framework in which the organism minimizes surprises. We investigated internal models and gathered evidence for cognitive processes in early sensory systems. Normal brain function involves the interaction of internal processes with incoming sensory stimuli. We have created a series of brain imaging experiments that sample internal models and feedback mechanisms in the early visual cortex. Primary visual cortex (V1) is the entry-stage for cortical processing of visual information. We can show that there are two information counter-streams concerned with: (1) retinotopic visual input and (2) top-down predictions of internal models generated by the brain. Our results speak to the conceptual framework of predictive coding, whereby internal models amplify and disamplify incoming information. Healthy brain function strikes a balance between the precision of prediction and prediction update based on prediction error. Our results incorporate state of the art, layer-specific ultra-high field fMRI, and other imaging techniques. A future direction will be to investigate internal models that mind-wander away from the here and now and simulate alternative scenarios.

Astrocytes revisited: from brain homeostasis to systems neuroscience

Elena Galea
Institute of Neurosciences, Universitat Autònoma de Barcelona, ICREA , ES

Although brain anatomy is currently somewhat discredited in favor of molecular analyses, the way cells look and move, as well as their relative abundance, gives important insights into their function (1). Thus, astrocytes are less abundant than neurons—the average ratio of astrocytes to neurons in humans is 2 to 10—and, despite their name, astrocytes do not look like stars at all. The term ‘astrocyte’ was coined at the beginning of the 2oth century because the dyes used to identify non-neuronal cells only stain cellular cytoskeletons, giving the erroneous impression that astrocytes are star-like cells bearing a small cell body from which long, thin processes emerge. Rather, astrocytes are huge, bushy cells with a dense, intricate network of processes and lamellae that virtually wrap all other cellular structures of the brain. Further, astrocytes are highly territorial, meaning that their processes do not overlap with those of other astrocytes, giving rise to a characteristic tiled arrangement—although there is astrocyte-to-astrocyte communication via gap junctions. Thus, astrocytes resemble a pipe system with selectively permeable walls, plausibly representing, after the vascular system, the second largest interface for molecule exchange and signal transmission in the brain. Astrocyte morphology appears well suited to sustaining global functions affecting several neural circuits at once. The goal of the class is to convey a contemporary view of astrocytic functions beyond textbook narrative. First, we will update the traditional view of the compartmentalization of homeostatic functions between astrocytes and neurons with regards to ion and neurotransmitter recycling, generation and use of energy fuels (2), and the recently discovered glymphatic system (3). Second, we will analyze how unbiased molecular profiling at single-cell and population levels has transformed our view of astrocytes in health and disease, as well as astrocyte-based drug discovery (4). Third, we will address the emerging notion that astrocytes may contribute to complex behaviors as a building block of neural circuits, as shown by Ca2+ imaging, chemogenetics and optogenetics (5, 6). Open questions to be discussed during the class are: could astrocytes encode by way of Ca2+ transients, if so, what variables do astrocytes encode, what is the function of astrocyte tiling, do astrocytes control brain state, and are there instances in which brain computations may specifically call for astrocytes?

  1. Masgrau, R., Guaza, C., Ransohoff, R., and Galea, E. (2017). Should we stop saying ‘glia’ and ‘neuroinflammation’? Trends in Molecular Neuroscience 23(6): 486-500. doi:10.1016/j.molmed.2017.04.005.
  2. Eraso-Pichot, A., Brasó-Vives, M., Golbano, A., Menacho, C., Claro, E., Galea, E., Masgrau, R. (2018). GSEA of mouse and human mitochondriomes reveals fatty acid oxidation in astrocytes. GLIA 66(8):1724, 1735; org/10.1002/glia.23330.
  3. Xie, L., Kang, H., Xu, Q., Chen, M.J., Liao, Y., Thiyagarajan, M., O’Donnell, J., Christensen, D.J, Nicholson, Ch., Iliff, J.J., Takano, T., Deane, R., and Nedergaard, M. (2013). Sleep drives metabolic clearance from the adult brain. Science 18(342), 373-377. doi: 10.1126/science.1241224.
  4. Pardo, L., Schlüter, A., Valor, L.M., Barco, A., Giral, , Golbano, A., Hidalgo, J., Jia, P., Zhao, Z., Jové, M., Portero-Otin, M., Ruiz, M., Giménez-Llort, L., Masgrau, R., Pujol, A., Galea, E. (2016). Targeted activation of CREB in reactive astrocytes is neuroprotective in focal acute cortical injury. GLIA 64(5):853-74. doi: 10.1002/glia.22969.
  5. Poskanzer, K. E., & Yuste, R. (2016). Astrocytes regulate cortical state switching in vivo. Proc Natl Acad Sci U S A 113(19), E2675-2684. doi:10.1073/pnas.1520759113
  6. Martin-Fernandez, M., Jamison, S., Robin, L.M., Zhao, Z, Martin, E.D., Aguilar, J., Benneyworth, M.A., Marsicano, G., Araque, A. (2017). Synapse-specific astrocyte gating of amygdala-related behavior. Nat Neurosci. 20(11):1540-1548. doi: 10.1038/nn.4649.

Accompanying responsible technological innovation.

Christine Aicardi
King’s College,  London, UK

New developments at the intersection of neuroscience and information technology have the potential to be socially and ethically disruptive. The talk will introduce ideas of social responsibility in science. It will then present, based on practical examples drawn from the Human Brain Project, how responsible innovation principles can be deployed, in practice, into multidisciplinary research collaborations, allowing for a continuous and participative risk assessment. It will also reflect on some of the limitations experienced by the speaker over the course of her participation in the Human Brain Project.

Self- organizing models of brain and behaviour.

Stuart Wilson
University of Sheffield, UK

The cortex sets the brains of mammals apart from the brains of other vertebrates, and the complex organisation of the cortical sheet sets the brains of primates apart from the brains of other mammals. The cortical sheet is partitioned into regions that process information about different modalities. Within each region, nearby neurons tend to respond selectively to similar patterns of input, such that cortical regions are functionally organised in terms of topological maps. Topological maps can be recreated in computer simulations by assuming that cells interact only locally and that their activities are driven by external input. The success of these simple models in recreating map features suggests that cortical maps emerge during postnatal development according to similar self-organising principles. Cortical maps self-organise to reflect the statistical structure in patterns of early sensory and motor experience, and by influencing behaviour, the organisation of cortical maps, in turn, shapes these early experiences. I have been trying to understand the nature of this interaction between brain and behaviour by allowing models of map self-organisation to guide the movement of simulated animal bodies, as they grow, develop, and interact with others. In this lecture, I will provide a tutorial introduction to self-organizing models of cortical development, and I will report my progress on modelling the interaction between brain and behaviour.

The social robot: emotions and drives.

Vicky Vouloutsi
SPECS_lab, IBEC Barcelona, ES

In a world of constant technological development, it is only natural to observe a shift from industrial robots to robots that are social and act in human environments. The challenge that robots must face in the integration in the integration in the human society is multidimensional: safety is important, as robots will have to be at a relatively close range with humans while interacting; furthermore, they will need to be understood and be accepted by humans as communication partners and understand the environment in which humans live.. The introduction of robots in our daily lives requires them to adopt social roles in dynamic environments that rarely involve well-defined tasks. Several domains could benefit from having social robots including the field of education, entertainment as well as healthcare. Indeed, humans could benefit from social robots that can act as caretakers for the elderly or even have the role of companions in general. This raises a fundamental question: do humans interact with robots in a similar way to how they interact with other humans especially with their peers?

Translational Neuroscience: from neurons to large-scale networks and virtual brains.

Viktor Jirsa
Institute de Neurosciences des Systèmes, CNRS, Fr.

Over the past decade, we have demonstrated that the fusion of subject-specific structural information of the human brain with mathematical dynamic models allows building biologically realistic brain network models, which have a predictive value, beyond the explanatory power of each approach independently. This workflow requires considerable High Performance Computing (HPC) resources and makes use of the neuroinformatics platforms in the Human Brain Project (HBP). The virtual brain workflow is as follows: The network nodes hold neural population models, which are derived in (HBP-SP4: Theoretical Neuroscience) using mean field techniques from statistical physics expressing ensemble activity via collective variables. This approach has been successfully applied to the modeling of the resting state dynamics of individual human brains, validated in mouse brain models, and translated to clinical applications in stroke and epilepsy. The open source neuroinformatics platform The Virtual Brain (http://www.thevirtualbrain.org) is integral part of the Human Brain Project (available also in the HBP collaboratory) and enables the complete workflow from the neuroimages to the simulation of patient-specific brain signals. Here we will illustrate the workflow along the example of epilepsy: we reconstruct personalized connectivity matrices of human epileptic patients using Diffusion Tensor weighted Imaging (DTI). Subsets of brain regions generating seizures in patients with refractory partial epilepsy are referred to as the epileptogenic zone (EZ). During a seizure, paroxysmal activity is not restricted to the EZ, but may recruit other brain regions and propagate activity through large brain networks, which comprise brain regions that are not necessarily epileptogenic. The identification of the EZ is crucial for candidates for neurosurgery and requires unambiguous criteria that evaluate the degree of epileptogenicity of brain regions. Stability analyses of propagating waves provide a set of indices quantifying the degree of epileptogenicity and predict conditions, under which seizures propagate to non-epileptogenic brain regions, explaining the responses to intracerebral electric stimulation in epileptogenic and non-epileptogenic areas. These results provide guidance in the presurgical evaluation of epileptogenicity based on electrographic signatures in intracerebral electroencephalograms and have been validated in small-scale clinical trials. The example of epilepsy nicely underwrites the predictive value of personalized large-scale brain network models.

Mechanisms of neurotransmission.

Pau Gorostizia
Institute for Bioengineering of Catalonia, Barcelona, ES

TBA

Space, memory, action: insights from behaviour and neurophysiology.

Daniel Pacheco
Institute for Bioengineering of Catalonia, Barcelona, ES

TBA

Understanding automatic and deliberate control of action.

Giovanni Maffei
Institute for Bioengineering of Catalonia, Barcelona, ES

Automatic and deliberate processes are at the core of brain functions, however, to acquire a complete knowledge of their relevance for behavior it is necessary to take into consideration their embodied nature. In an interdisciplinary effort which integrates methods from computational modeling, robotics, and electrophysiology, I will presents a series of studies that aim at advancing the understanding of the brain mechanisms responsible for the automatic and deliberate control of action. Through the formulation of a biologically constrained control architecture engaged in a real-world foraging task, we lay the ground for modeling and analyzing complex behavior which emerges from the interplay between the automatic cerebro-cerebellar system and the deliberate fronto-hippocampal system.  We explore the properties of the automatic control system and advance a novel hypothesis on the role of the cerebellum by recasting its computation in the perceptual domain. We ask how the automatic and deliberate systems interact during unexpected situations that require a sudden change of plans. By analyzing the neural dynamics of the human frontal cortex during switching between automatic and deliberate actions, we support the role of low-frequency oscillations in orchestrating goal-oriented behavior. Altogether, these results contribute to our understanding of how automatic and deliberate processes in the brain control behavior and advance novel insights that challenge or extend current theories.