top of page

>> DIEP seminars and talks - archives 2022

Below you can find more information on the DIEP seminars of 2022. 

>> DIEP Seminar: Wout Merbis (DIEP)

Emergent information dynamics in many-body interconnected systems | 11am, 15th of December 2022
​

The information implicitly represented in the state of physical systems allows one to analyze them with analytical techniques from statistical mechanics and information theory. In the case of complex networks such techniques are inspired by quantum statistical physics and have been used to analyze biophysical systems, from virus-host protein-protein interactions to whole-brain models of humans in health and disease. Here, instead of node-node interactions, we focus on the flow of information between network configurations. Our results unravel fundamental differences between widely used spin models on networks, such as voter and kinetic dynamics, which cannot be found from classical node-based analysis. Our model opens the door to adapting powerful analytical methods from quantum many-body systems to study the interplay between structure and dynamics in interconnected systems.

Wout_Merbis_edited.jpg

>> DIEP Seminar: Koen van der Zwet (U. Amsterdam)

Opportunistic organisation of illicit supply chains | 11am, 8th of December 2022

 

Opportunistic and small criminal groups are the predominant form in organised criminal markets. As a collective the opportunistic groups form an extensive network of contacts that enables organisation of more sophisticated processes. In the past decade, the network modelling approach has become the novel and focal data-driven method to analyse these criminal organisations. The developed models aim to identify key persons and vulnerabilities of criminal organisations. However, network studies are often limited to analysing static representations. As a consequence, the adaptive capacity of criminal networks remains poorly understood.

 

In this talk I present an agent-based model that incorporates the necessary dimensions to represent a generalisation of the dependencies in illicit supply chains, and model different organisation dynamics that enable individuals to share information and subsequently make strategic decisions. Two network approaches are applied to observe the core concepts of interaction, adaptation, and emergence in the illicit supply chain system. First, a hypergraph approach is used to analyse the opportunistic group interactions. Second, a multilayer network approach is introduced to model dependencies of the illicit activities, enduring social relationships, and fluid transaction relationships.  This simulation-based approach enables us to analyse the effectiveness of the emergent transactions generated by the interplay of organisational structure and dynamics.

uva_profileimagecompound (1)_edited.jpg

>> DIEP Seminar: Astrid Groot (U. Amsterdam)

How to measure evolution? | 11am, 24th of November 2022
​

Evolution is the change through time, where time is measured over generations. Evolutionary histories can be deduced by comparing phylogenetic species trees to genetic changes in genes that have been identified to encode phenotypic traits, such as eye color or odor perception. However, whether and to what extent evolutionary predictions can be made with genetic information is elusive, because a) selection acts on the phenotype, which is determined by genes as well as the environment, and b) different genetic pathways can lead to the same phenotype. Our long-term research on the evolution of chemical communication in relation to speciation in moths (Lepidoptera, Noctuidae) illustrates the difficulty to translate evolution at the gene level to evolution at the organism (phenotype) level. Olfactory receptors determine which odors can be detected by which animals. Since moths find their mates through sex pheromones, the evolution of moth species is hypothesized to be reflected in the evolution of olfactory receptors. However, in our experimental analysis with a moth species that consists of two pheromone strains, we found that not olfactory receptors but a gene involved in neuronal development underlies the difference in response. Even though this result has led to new directions in our experimental work, it poses questions on whether and how information at the genetic level can be used to predict evolution. In our NWA Origins of Life project Predicting evolution, we developed a testable method to predict evolution at the genetic level, using the model organism Caenorhabditis elegans (a nematode, see http://www.wormbook.org/). 

004-groot-astrid-hoogleraar-fnwi-web-fotogillissen0gmail_edited.jpg

>> DIEP Seminar: Eddie Lee (Complexity Science Hub Viena)

Idea engines: Unifying innovation and obsolescence from markets and genetic evolution to science | 11am, 1st of December 2022
​

Innovation and obsolescence describe dynamics of ever-churning and adapting social and biological systems, concepts that encompass field-specific formulations. We formalize the connection with a toy model of the dynamics of the „space of the possible“ (e.g. technologies, mutations, theories) to which agents (e.g. firms, organisms, scientists) couple as they grow, die, and replicate. We predict three regimes: the space is finite, ever growing, or a Schumpeterian dystopia in which obsolescence drives the system to collapse. We reveal a critical boundary at which the space of the possible fluctuates dramatically in size, displaying recurrent periods of minimal and of veritable diversity. When the space is finite, corresponding to physically realizable systems, we find surprising structure. This structure predicts a taxonomy for the density of agents near and away from the innovative frontier that we compare with distributions of firm productivity, covid diversity, and citation rates for scientific publications. Remarkably, our minimal model derived from first principles aligns with empirical examples, implying a follow-the-leader dynamic in firm cost efficiency and biological evolution, whereas scientific progress reflects consensus that waits on old ideas to go obsolete. Our theory introduces a fresh and empirically testable framework for unifying innovation and obsolescence across fields.

EddieLEE_500x500_cStephanieBourke-Altmann-3_edited.jpg

>> DIEP Seminar: Bernd Ensing (U. Amsterdam)

Understanding molecular transitions as pathways in free energy landscapes | 11am, 17th of November 2022
​

After laying out his momentous contribution to the foundation of quantum mechanics, Paul Dirac allegedly said: “… and the rest is chemistry!”. Apparently, Dirac had no idea of the richness and complexity of observable phenomena that emerge as solutions from his Dirac (or Schrödinger’s) equation. In this talk, I will showcase several material properties in soft matter and aqueous systems that are, at first sight, remarkably surprising when just regarding the microscopic chemical constituents. Molecular dynamics simulations can be very helpful to unravel the molecular interactions and mechanisms that underly these observable material properties. Molecular transitions, such as chemical reactions, phase transitions, and self-assembly processes can be comprehensibly represented as pathways on low-dimensional free energy landscapes, but this entails finding the essential reaction coordinate(s) or the few collective variables in the many-particle system that describe the molecular process, which is often far from trivial. During this talk, I will present several in-house developed enhanced sampling methods that we apply for this purpose, including a reinforcement learning approach that does not require prior knowledge of the collective variables.

BerndEnsing_edited.jpg

>> DIEP Seminar: Sara Walker (U. Arizona)

When Time is an Object (and Implications for Emergence) | 4pm, 10th of November 2022
​

Most of the history of physics has been one of removing time from our fundamental descriptions of nature. Newton did this by writing laws of physics that are timeless and exist outside of the universe. Einstein’s universe carries an implication that we live in a universe where time does not pass, and our perceptions of its passage are therefore illusory. In the world of Boltzmann and other statistical mechanists time is an emergent property arising due to the second law. We have not yet had a theory of physics (successful in the long term) where time is fundamental. The goal of this talk is to provoke discussion on why this may the root cause of issues with understanding and quantifying emergence as a physical property, with specific focus on the ‘emergence of life’. In chemistry and biology, we find objects that require memory to produce them because they are too complex to be produced by random chance. This insight forms the foundation of a new theory, called assembly theory, which describes how much memory must exist for a molecular object to come into existence, with the implication that more evolved something is, the more memory is required. Using molecular assembly, we have demonstrated how it is possible to agnostically (in absence of specific details) identify complex molecules as signatures of life empirically using Mass Spec. One of the most interesting implications – if the theory holds – is what it tells us about how time exists in complex objects created by evolution, like the ones we find all around us in the biosphere and technosphere. Objects created by evolutionary processes require time for their formation, because they exist across time – time is a physical attribute. The talk will explore the consequences of re-envisioning time as an observable property of complex matter, including for the general feature of ‘emergence’ being associated to objects that are deep in time.

sara-walker-1_0_edited.jpg

>> DIEP Seminar: Fernando P. Santos (U. Amsterdam)

Reputation dynamics and the emergence of human cooperation | 11am, 3rd of November 2022
​

Indirect Reciprocity – Alice behaves adequately towards Bob; Carol knows about it and thus helps Alice – is a central mechanism to explain the emergence of cooperation between humans. Simultaneously, indirect reciprocity can inspire new ways of engineering cooperation in multiagent (artificial) systems where reputations are paramount. A central challenge in indirect reciprocity is understanding which social norms – here rules defining how reputations should be attributed – lead to the highest levels of cooperation. I will address this challenge through evolutionary game theoretical models that formalize the coupled dynamics of reputations and cooperation as a stochastic process. I will then discuss ways of quantifying the complexity of norms and strategies and ask which social norms promote maximal cooperation at minimal complexity. Finally, I will discuss models of indirect reciprocity on complex networks and discuss the emergence of reputation-driven polarization in these settings.

Santos-e1568042273205_edited.jpg

>> DIEP Seminar: Fleur Zeldenrust (Radboud University Nijmegen)

Emergent properties of networks in the brain: the relation between structure and computation | 11am, 27th of October 2022
​

The brain continuously processes information. The electrical activity of the neural networks that make up the brain is both irregular and unreliable. How is it possible that the brain can perform stable computations on the basis of such chaotic activity? In this presentation, I will discuss how such chaotic activity emerges, how it is influenced by the nonlinear properties of the network nodes, and different theories on how the brain can perform stable computations on the basis of such unreliable activity. 

nmldrczbwvs-un0assueba_edited.jpg

>> DIEP Seminar: Esmee Geerken (IAS resident artist)

The Holebearers | 11am, 13th of October 2022
​

In her talk, Esmee will introduce some of her current Art-Science projects, touching upon emergence, self-organization, forms and phase-shifts in ‘systems’ at various spatial scales. From human cells interacting with growing crystals to unicellular organisms precipitating their shells from oceanwater to end up in the rocks we use for our building materials, from human-scale architectures (e.g. Amsterdam Science Park) to the ‘shelters’ we build within our human mind, Esmee is interested in how shapes and forms emerge, and in the role and agency of individuals (or individuality) within these larger ‘complex systems’, and in navigating our understanding of the world through storytelling.

​

 

Furthermore, Esmee hopes to engage in a dialogue with the DIEP community, to learn more about inherent chance, entropy and information, and is curious to learn more on how creative, scientific insights emerge in your minds, for an upcoming project. 

1614784296550_edited.jpg

>> DIEP Seminar: Jocelyne Vreede (U. Amsterdam)

From atomic interactions to cellular processes | 11am, 29th of September 2022
​

In this seminar, I will explore how interactions between atoms can provide understanding of cellular processes, using molecular simulation techniques. Signal transduction is a mostly intracellular process that starts with a small trigger and ends with a response on a much larger scale. For example, the detection of a few photons can lead to a response involving the head turning towards the flash of light. Triggers initiating a signal transduction pathway can be many things, including a change in the concentration of a chemical compound. The subsequent response can vary widely, with altered gene expression, triggering an action potential and directional growth as examples. Understanding how a small trigger leads to a cell-wide response requires insights at atomic resolution in both space and time. Models based on atomic interaction potentials in combination with equations of motion can predict macroscopic properties of molecular systems. With molecular dynamics simulations it is possible to indicate, for instance, the various ways in which a protein can bind to DNA. If simulated long enough, such predictions can be quantitative, such as the affinity of a protein for a particular DNA sequence. A problem in molecular simulation of biomolecular systems that simulating long enough is almost always impossible, as the timescale of the event of interest usually is many orders of magnitude larger than the atomic fluctuations at which scale sampling takes place. Focusing on relevant collective coordinates provides a way to overcome this sampling problem and allows for fast predictions. 

​

Using two examples I will illustrate how molecular simulation can aid in the understanding of cellular processes. The first example focuses on elucidating the molecular mechanisms with which plants can detect sodium. Crop loss due to soil salinisation is an increasing threat to agriculture world wide, as plants have an adverse response to sodium. To date, it is unknown how plants sense sodium. Using a molecular simulation protocol, we are making progress in unraveling this signal transduction process. The second example involves the regulation of gene expression. Whether a gene is expressed depends on many factors, including the molecular configuration and shape of the DNA and the presence of regulating proteins. With molecular simulation it is now possible to quantitatively predict the affinity of a protein for a specific nucleotide sequence, opening up new ways of studying gene expression and other processes involving interactions between proteins and DNA.

JVreede_edited.jpg

>> DIEP Seminar: Enrico Cinti (U. Urbino and U. Geneva)

Spacetime emergence inside black holes | 11am, 15th of September 2022
​

The problem of the disappearance of spacetime has long been recognized as one of the most pressing philosophical and conceptual problems facing theories of quantum gravity. In this talk, my goal will be to look at this issue from the point of view of AdS/CFT, focusing in particular on the relationship between the non-perturbative definition of quantum gravity given by the duality and the semiclassical description of gravity given by the effective field theory in the bulk. By thinking in particular about the interior of black holes and the reconstruction map connecting their effective description to their fundamental one, I come to the surprising conclusion that the standard answer to the problem of the disappearance of spacetime, i.e. emergence, is inadequate in this case, at least as standardly formulated. Rather, I suggest that a more flexible and less ontologically demanding approach is required, whose basic features I outline.

315ff44855809d54a726e8c586da6618910c4812e16ad3eacc2580fe3adba825_edited.jpg

>> DIEP Seminar: Artemy Kolchinski (U. Tokyo)

Information geometry of fluxes and forces in nonequilibrium thermodynamics | 11am, 8th of September 2022
​

A nonequilibrium system is characterized by a set of thermodynamic forces and fluxes which give rise to entropy production (EP). We show that these forces and fluxes have an information-geometric structure, which allows us to decompose EP into nonnegative contributions from different types of forces. We focus in particular on the excess and housekeeping decomposition, which identifies contributions from conservative and nonconservative forces. Unlike the well-known nonadiabatic/adiabatic (Hatano-Sasa) approach, our decomposition is always well-defined, including in systems with odd variables and nonlinear systems without steady states. It is also operationally meaningful, leading to far-from-equilibrium thermodynamic uncertainty relations and speed limits. (joint work with Andreas Dechant, Kohei Yoshimura, and Sosuke Ito / https://arxiv.org/abs/2206.14599

croppedheadshot2_edited.jpg

>> DIEP Seminar: Sid Redner (Santa Fe Institute)

The Dynamics of Diversity and Polarization | 11am, 30th of June 2022
​

If opinions spread by interactions between reasonable individuals, why is consensus so unlikely?  I present some idealized social dynamics models, which are based on social interactions that decay with distance in opinion space, to understand why diversity is common.  I'll first describe the Axelrod model, in which multi-feature individuals move closer in opinion space only if they agree on some other feature. Another example is a 3-state voter model of leftists, centrists, and rightists, in which a centrist and an extremist influence each other but extremists of the opposite persuasion do not.  Depending on the initial condition, the population may or may not reach consensus.  Similarly, in the bounded compromise model, a population becomes increasingly fragmented as the political range of interaction of an individual decreases.  A related polarization phenomenon arises in the voter model when individuals are also influenced by competing and fixed external news sources.

Sid-Redner-sepia-400x400_9bd750.jpg_edited.jpg

>> DIEP Seminar: Tiziano Squartini (IMT Luca)

From graphs randomization to hypergraphs randomization | 11am, 23rd of June 2022
​

Network theory has emerged as a powerful paradigm to explain phenomena where units interact in a highly non-trivial way. So far, however, research in the field has mainly focused on pairwise interactions, disregarding the possibility that more-than-two constituent units could interact at a time. Hypergraphs represent a class of mathematical objects that could serve the scope of describing this novel kind of many-bodies interactions. In this paper, we propose benchmark models for hypergraphs analysis that generalise the usual Erdos-Renyi and Configuration Model in the simplest possible way, i.e. by randomising the hypergraph incidence matrix while preserving the corresponding connectivity/topological constraints - whose definition is, now, adapted to the novel framework. After exploring the mathematical properties of the proposed benchmark models, we consider two different applications: first, we define a novel quantity, the hyperedge assortativity, whose expected value we theoretically derive for all the introduced null models and which we, then, use to detect deviations in the corresponding real-world hypergraphs; second, we define a principled procedure for testing the statistical significance of the number of hyperedges connecting any two nodes.

ts_picture_edited.jpg

>> DIEP Seminar: Pim van der Hoorn (TU Eindhoven)

Geometry and complex networks. A powerful partnership | 11am, 16th of June 2022
​

Geometry is a powerful tool in many scientific domains, from fundamental physics to social science. It also plays and important role in the study of complex, especially from the point of constructing models for networks. Here nodes represent positions in some geometric space and connections are created based on the distances in that space. In this talk I will highlight results for two key aspects of this use of geometry. In the first part I will discuss a model for complex networks that uses Hyperbolic geometry. This geometry turns out to naturally lead to networks that exhibit key features found in many real-world networks: sparse, power-law degrees and clustering. Here I shall present new results on the clustering of this network model. In the second part I shall touch upon the challenge of recovering geometric information from networks. That is, if we see the resulting network can we identify the underlying geometric space? I will discuss a discrete version of curvature for networks developed by Yann Ollivier. Then I will present results that show that this notion can actually uncover the curvature of hidden geometry and large scale simulation that show how well this works for Euclidean, Spherical and Hyperbolic spaces.

download (30)_edited.jpg

>> DIEP Seminar: Muhammed Sayin (Bilkent University)

Towards the Foundation of Dynamic Multi-agent Learning | 11am, 9th of June 2022
​

Many of the forefront applications of reinforcement learning involve multiple agents and dynamic environments, e.g., playing chess and Go games, autonomous driving, and robotics. Unfortunately, there has been limited progress on the foundation of dynamic multi-agent learning, especially independent learning in which intelligent agents take actions consistent with their individual objectives, e.g., based on behavioral learning models, quite contrary to the plethora of results on algorithmic solutions.


In this talk, I will present a new framework and several new independent learning dynamics. These dynamics converge almost surely to an equilibrium or converge selectively to an equilibrium maximizing the social welfare in certain but important classes of Markov games -- ideal models for decentralized multi-agent reinforcement learning. These results can also be generalized to the cases where agents do not know the model of the environment, do not observe opponent actions, and can adopt different learning rates. I will conclude my talk with several remarks on possible future research directions for the framework presented.

MUHAMMED-SAYIN-600x600-2-300x300_edited.jpg

>> DIEP Seminar: Sabin Roman (U. Cambridge)

Modelling the long-term evolution and collapse of societies | 11am, 2nd of June 2022
​

The talk will cover some key modelling issues that come up when considering the long-term development of societies. Of particular importance is the topic of societal collapse as the archaeological record has numerous instances of the phenomenon. Sabin will discuss some of the general modelling philosophy, relevant literature, his own work on ancient societies (Easter Island, the Maya, Roman Empire and Chinese dynasties) and implications for modern society. There are several modelling considerations unique to modern society that will be highlighted.

sabin_edited.jpg

>> DIEP Seminar: Manilo de Domenico (University of Padua)

Emergent phenomena in human networks and dynamics: infodemics and epidemics | 11am, 19th of May 2022
​

Complex systems are characterized by constituents -- from neurons in the brain to individuals in a social network -- which often exhibit a special structural organization and nonlinear dynamics. As a consequence, a complex system can not be understood by studying its units separately because their interactions lead to unexpected emerging phenomena, from collective behavior to phase transitions. In the last decade, we have discovered that a new level of complexity characterizes a variety of natural and artificial systems, where units interact, simultaneously, in distinct ways. For instance, this is the case of multimodal transportation systems (e.g., metro, bus and train networks) or of social networks, whose interactions might be of different type (e.g. trust, trade, virtual, etc.).  The unprecedented newfound wealth of data allows to categorize system's interdependency by defining distinct "layers", each one encoding a different network representation of the system. The result is a multilayer network model. In this talk we will discuss the most salient features of multilayer systems, with special attention to empirical human communication/mobility networks responsible for emergent phenomena such as infodemics and epidemics.

manlio_dedomenico_small_edited.jpg

>> DIEP Seminar: Greg Stephens (VU Amsterdam)

Bridging timescales in partially observed dynamical systems | 11am, 12th of May 2022
​

In approaches such as Langevin, combining fast fluctuations with slower dynamics has proven remarkably powerful. But how do we proceed in generally out-of-equilibrium systems for which we lack underlying equations?  Here we construct short-time, maximally-predictive states by concatenating measurements in time, partitioning the resulting sequences using maximum entropy, and choosing the sequence length to maximize short-time predictive information. We use transitions between these states to analyze reconstructed dynamics through transfer operators, revealing timescale separation with long-lived collective modes through the operator spectrum. Applicable to both deterministic and stochastic systems, we illustrate our approach through partial observations of the Lorenz system and the stochastic dynamics of a particle in a double-well potential. Applied to the behavior of the nematode worm C. elegans, we bridge sub-second posture fluctuations and long range effective diffusion in foraging behavior, recovering the ``ballistic-to-diffusive'' transition in the worm's centroid trajectories. We use transfer operators to reveal long-lived ``run-and-pirouette'' behaviors, and predict additional subtle subdivisions of the worm's ``run'' dynamics.

GregStephens_edited.jpg

>> DIEP Seminar: Katarzyna Sznajd-Weron (Wroclaw University)

Private Truths, Public Lies” within agent-based modeling | 11am, 28th of April 2022
​

The title of this talk was inspired by Timur Kuran's book entitled "Private Truths, Public Lies. The Social Consequences of Preference Falsification".  During the presentation I will discuss the idea of Preference Falsification (PF) and propose a simple binary agent-based model to describe PF by introducing two levels of opinion: public and private. I will show how the second (hidden) level of opinion affects the phase transitions occurring in the system, and thus how it can help to understand the origin of the so-called social hysteresis. This research is a part of the project  Towards understanding of the social hysteresis: an agent-based approach, NCN 2019/35/B/HS6/02530.

download (29)_edited.jpg

>> DIEP Seminar: Hendrik Baier (TU Eindhoven)

A Vision for Explainable Search, and Some First Steps Towards It| 11am, 21st of April 2022
​

In his presentation, Hendrik is going to sketch one of the research themes he is developing - explainable search. The first part of this talk is going to define and motivate the problem: While AI agents based on search or online planning are state of the art in many challenging domains, famously in board games, current approaches lack the ability to explain, summarize, or visualize their plans and decisions. Users struggle to understand how observable behavior is derived from considering complex spaces of possible futures, contingencies, and eventualities, spanned by the available actions of the agent. This limits human trust in high-stakes scenarios, as well as effective human-AI collaboration. Hendrik will outline the proposed research direction of explainable search, and important related research challenges.
The second part of the talk is then going to move from challenges to concrete first steps towards solving them. Focusing on algorithms in the Monte Carlo Tree Search (MCTS) family, at the heart of many recent breakthroughs in AI, Hendrik is going to present his ongoing exploration of possible explanations for sequential decision-making and behavior. This work is tackling for the first time some of the challenges previously posed for explainable search, such as: meaningfully summarizing the space of possible futures spanned by the available actions of the AI and their possible consequences, in order to explain how the AI’s choices between them emerge; considering such explanations not only as static objects but as interactive conversations between user and AI; and understanding explanation not only as a one-way information flow from the AI to the user, but as a tool for human-AI collaboration and for leveraging both AI and human capabilities in problem solving.

download (28)_edited.jpg

>> DIEP Seminar: Chris Reinders Folmer (U. Amsterdam)

Beyond narrow perspectives on prosocial behavior: benefits of aggregation of game behavior | 11am, 14th of April 2022
​

When using economic games to study prosocial, cooperative behavior, research tends to zoom in on highly specific settings: a specific type of game, represented with a specific set of parameters, and a specific, narrow subset of treatments. However, some research suggests that behavior in such narrow settings is only modestly related to other indices of prosociality, such as prosocial personality traits and mundane prosocial behaviors. In this research, we advocate a broader approach that moves beyond such narrow perspectives, by aggregating choice behavior across different representations and types of games. We find that aggregation across different settings strengthens associations with prosocial personality, and with prosocial behaviors in mundane settings. Based on this, we recommend for research to move beyond narrow approaches to prosocial behavior, to broad approaches based on aggregation. 

uva_profileimagecompound_edited.jpg

>> DIEP Seminar: Maurice Weiler (U. Amsterdam)

Coordinate independent convolutional networks - Isometry and gauge equivariant convolutions on Riemannian manifolds | 11am, 7th of April 2022
​

Motivated by the vast success of deep convolutional networks, there is a great interest in generalizing convolutions to non-Euclidean manifolds. A major complication in comparison to flat spaces is that it is unclear in which alignment a convolution kernel should be applied on a manifold. The underlying reason for this ambiguity is that general manifolds do not come with a canonical choice of reference frames (gauge). Kernels and features therefore have to be expressed relative to arbitrary coordinates. We argue that the particular choice of coordinatization should not affect a network's inference - it should be coordinate independent. A simultaneous demand for coordinate independence and weight sharing is shown to result in a requirement on the network to be equivariant under local gauge transformations (changes of local reference frames). The ambiguity of reference frames depends thereby on the G-structure of the manifold, such that the necessary level of gauge equivariance is prescribed by the corresponding structure group G. Coordinate independent convolutions are proven to be equivariant w.r.t. those isometries that are symmetries of the G-structure. The resulting theory is formulated in a coordinate free fashion in terms of fiber bundles. To exemplify the design of coordinate independent convolutions, we implement a convolutional network on the Möbius strip. The generality of our differential geometric formulation of convolutional networks is demonstrated by an extensive literature review which explains a large number of EuclideanCNNs, spherical CNNs and CNNs on general surfaces as specific instances of coordinate independent convolutions.

citations (4)_edited.jpg

>> DIEP Seminar: Jaap Kaandorp (U. Amsterdam)

Modelling embryogenesis and biomineralization in sea anemones, jelly fish and corals | 11am, 24th of March 2022
​

Rather than being directed by a central control mechanism, embryogenesis can be viewed as an emergent behavior resulting from a complex system in which several sub-processes on very different temporal and spatial scales (ranging from nanometer and nanoseconds to cm and days) are connected into a multi-scale system. In our research we have been focusing on the embryogenesis of basal organisms like the non-calcifying sea anemone Nematostella vectensis and the hydrozoan Clytia hemisphaerica and the calcifying coral Acropora millepora. We have developed methods for analysing spatio-temporal gene expression patterns, methods for spatio-temporal modelling and inferring gene regulatory networks from gene expression data (qPCR data and in–situ hybridizations) and a cell-based mechanical model of early embryogenesis. In the coral we have developed a model of calcification physiology controlling biomineralization. Currently we are investigating how the different levels of organization (gene regulation controlling embryogenesis, cell mechanics and biomineralization) can be coupled into a multi-scale model of embryogenesis.

JaapKaandorp_edited.jpg

>> DIEP Seminar: Fernando N. Santos (Amsterdam UMC)

Emergence of High Order Hubs in the Human Connectome | 11am, 17th of March 2022
​

In this talk, I would like to share early results on our ongoing work with colleagues at IAS and VUmc Amsterdam on high-order interactions in functional brain networks. Network theory is predominantly based on pairwise relationships between nodes, which is not realistic for most complex systems. In particular, it does not accurately capture nondyadic interactions in the brain. Over the past years, I have been interested in inferring high-order interactions from brain signals and exploring their consequences for our understanding of the human brain as a high-order network.

 

The talk will be divided in two parts: i) methodology and ii) application. I will first introduce the methodology we developed to analyze the brain as a high-order network, in particular the multivariate signal processing pipeline that can define high-order interactions and connectivity in rs-fMRI  images of brain networks (or any other time series signal) in an intuitive way. I will then present preliminary result from our current applied work, in which we searched for high-order "hubs" in a cohort of 100 individuals from the Human Connectome project as a proof of concept. In fact, even though we did not consider any prior knowledge about the functionality of specific brain regions in our analysis, we found that well-known integration and segregation patterns in the brain emerge spontaneously from the high-order hubs and can be considered emergent properties of functional brain networks. In this context, each type of high-order interdependency is compatible with distinct systems in the brain. For instance, three-point interactions seem to manifest segregation and integration principles consistent with the sensory-motor and visual systems. We believe that this opens up exciting avenues for further research.

fernando-nobrega-santos_edited.jpg

>> DIEP Seminar: Peter Sloot (U. Amsterdam)

Discussion on "The calculi of emergence" | 1pm, 3rd of March 2022 
​

Discussion led by Peter Sloot on the paper "The calculi of emergence: computation, dynamics and induction" by James P. Crutchfield.

SlootP_edited.jpg

>> DIEP Seminar: Mark Golden (U. Amsterdam)

How the photoelectric effect on rusty copper can test AdS-CFT in real life | 10th of February 2022
​

When the (many) electrons in crystals interact strongly with each other the system behaves in ways that condensed matter theory often cannot predict, and even to try to brings enormous computational cost. The discovery of high temperature superconductivity in a copper oxide (hence the rust) is a poster-child for such unexpected behaviour, and ‘emergent’ has been used for decades to describe the behaviour of such strongly correlated electron systems.

Usually a superconductor is born out of a metallic quantum state called a Fermi liquid, on lowering the temperature. In the rusty-copper high temperature superconductors the parent high temperature state is not a normal metal. As it breaks many electron behavioural rules and we don’t (yet) understand it, the name it gets is a ‘strange metal’.

In this seminar I’ll try to explain how a simple experiment (photoelectric effect, see pic) can spy on the mysterious inner workings of quantum materials such as the strange metals. In the case of our crystal of rusty copper, in which the electrons live essentially in 2 spatial dimensions (and time), the data show a behaviour so strange that we allow ourselves to look further than condensed matter theory for help. The cavalry on the hillside riding to our rescue take the form of gravitational problems connected to black holes in 3+1 dimensions, soluble using the tools of string theory.

Really. 

So, perhaps simple lab experiments taking place here in the SciencePark seem to be able to connect with and test the correspondence between general relativity in anti-de-Sitter space and conformal field theory (AdS-CFT). It seems the emergent geometries of AdS-CFT seem to help explain the strange metal behaviour of emergent high temperature superconductors.

Though many have been involved in this research I’d like to highlight the PhD researchers Steef Smit (lead experimentalist; UvA) and Enea Mauri (with Henk Stoof lead theorists @UU). 

For the first paper out on this, see: Smit et al., arxiv.org/pdf/2112.06576

markgolden.jpeg

>> DIEP Seminar: Peter Bolhuis (U. Amsterdam)

Understanding emergent rare event behavior in high dimensional systems | 3rd of February 2022
​

Microscopic dynamics governs the macroscopic behavior of complex molecular systems, from materials to living cells. To predict this emerging behavior, physics-based atomistic models combined with well-established equations of motion provide a high dimensional dynamical system, which, given an initial condition, can be time-integrated to yield the long-time behavior for the system, and hence its macroscopic properties. In principle, one could use accurate quantum mechanics for this time evolution, but in all but a few cases this is utterly unfeasible and intractable. Fortunately, coarse-graining in time and/or space, neglecting or integrating out degrees of freedom leads to simpler descriptions, which make the problem tractable.  This leads to molecular dynamics (MD) based on effective (empirical) force fields, which can be designed to reproduce experimental or (quantum mechanically) computed observables. However, even in such empirical MD simulations timescales can be quite long, caused by the presence of dynamical bottlenecks between metastable states.  Exploring these transitions by direct MD is extremely costly, for example in the case of protein folding or nucleation.  Further coarse graining towards simpler force fields could solve this problem, but then important molecular details might get lost. A better option is to coarse-grain in time and make use of the emergent Markovianity of the dynamics, which leads to a probabilistic Markov state description. Still, this requires knowledge of the state-to-state transition rates that are hard to obtain by direct MD as these transitions occur very rarely on the timescale of the simulation. This rare event problem can be solved by computing the probability to find the system at the dynamical bottleneck, the transition state, by slowly restraining the system towards it. This in turn requires a way to identify this transition state by a collective variable or reaction coordinate (sometimes also called the importance function). This very important reaction coordinate (RC) is ideally  a low dimensional representation of the high-dimensional space. Finding it requires a dimensionality reduction of the phase space, but is often elusive. In fact, it is kind of a chicken and egg problem: To study the rare event, one needs the RC, but to find the RC one needs to observe the rare event first.  

A solution to this problem is to use the concept of trajectory sampling. One can sample trajectories between easily identifiable metastable states, and extract the RC from the resulting path ensemble by a dimensionality reduction, e.g. using machine learning.  The thus obtained knowledge can then be used to better describe the kinetics and dynamics in complex molecular systems.   

Of course, this approach still relies on accurate force fields. Recent work demonstrated that it is possible to correct for force field inaccuracies by incorporating experimental kinetic constraints into the computed trajectory ensemble, through the use of the maximum (path) entropy principle, even further improving our understanding of rare events in complex systems.

PeterBolhuis.jpeg

>> DIEP Seminar: Tomas Veloz (Vrije Universiteit Brussel)

Chemical organisation theory, structural changes, and emergence | 27th January 2022
​

Chemical Organization Theory (COT) is an approach to compute closed and self-maintaining reaction networks (RNs) called organizations. Organizations represent an abstraction of the attractors of the dynamics of the RN. In this sense, they reflect the stable-enough-to-be-observed parts of the RN and thus might serve as a model of emergent structures. Organizations form a partial ordered set (POS),  which in many cases can be equipped with operators to become a lattice. This connection between RN and mathematical logic opens interesting and new issues. One of them is that RNs can be perturbed by adding or subtracting reactions (meaning externally or evolutionarily induced structural changes), and such changes modify the lattice properties. I will present a decomposition theorem for organizations which allows to study in more detail the impact of a structural perturbations on the organizations lattice, and the potential changes at a logical level that can occur.

tveloz_edited.jpg

>> DIEP Seminar: André de Roos (U. Amsterdam)

Everything is a network, but the network is not everything: Dynamics of stage-structured food webs | 20th of January 2022
​

Ecological communities are traditionally viewed as networks of negative and positive interactions between species, while the species themselves are seen as collections of identical individuals. Using this conceptualisation, however, it had been difficult to explain the importance of biodiversity. In fact, analysis of the topology of ecological interaction networks has shown that “diversity begets instability”: more complex and more diverse communities tend to be dynamically unstable and hence not persist. Complex communities are predicted to be stable only when species growth rates are mostly limited by intraspecific self-regulation (within-node limitation) rather than by interactions with resources, competitors, and predators.

In this talk I show how adding a second axis of complexity to the study of ecological interaction networks yields contrasting predictions about the relation between community diversity and stability. This second axis recognises that conspecific individuals are different from one another, first and foremost, because they are in different stages of development. Using food web models that account for juvenile and adult individuals of species, I show that commonly observed differences between juveniles and adults in foraging capacity and predation risk result in larger, more complex communities than predicted by models without stage structure. Based on their species interaction networks these complex and diverse communities would be expected to be unstable, but these destabilizing effects of species interactions are overruled by stabilizing changes in juvenile–adult stage structure. Differences between juvenile and adult individuals hence offer a natural resolution to the diversity–stability enigma of ecological communities.

GBP_3036-scaled_edited.jpg
bottom of page