Horacio G. Rotstein

Collective decision-making experiments have largely focused on simple scenarios in which groups choose between two options that differ along a single attribute. However, it remains unclear whether principles derived from this single-attribute best-of-2 paradigm generalizes to the multi-attribute, multi-option decisions that groups often face in nature. Here, we use mean-field and agent-based models of cockroach aggregation to study collective decision-making across a range of problem complexities, varying both the number of options (N) and the number of attributes describing each option. Our models make 3 novel predictions: (1) cockroach groups use a compensatory algorithm when integrating attributes, trading off strength in one attribute against weakness in another; (2) decision-making collapses abruptly once N exceeds a critical threshold; and (3) decision time scales non-monotonically with N. Together, these results indicate that dynamics characterized in single-attribute best-of-2 experiments do not extrapolate to multi-attribute best-of-N decision-making. Collective choice under realistic complexity may follow principles not yet captured by existing models.

The knowledge economy is fundamentally driven by human capital, making the “brain drain” phenomenon a critical challenge for national development. Traditionally, policy frameworks have focused on repatriation, circulation or vinculation to address this issue. However, these models often fail to account for more regular and permanent institutional associations between the diaspora of knowledge professionals (KPs) and the science, technology and innovation (STI) institutions of their country of origin (CO). Such associations are increasingly feasibly due to technological advances that render many STI activities independent of geography. This paper advances the notion of KP diaspora integration within the domestic STI institutions of their CO, incorporating a framework focused on structural institutional membership. We provide an in depth analysis and propose a blueprint for policy development aimed at achieving this goal. We use Argentina as a case study for several reasons, including its well-developed STI system, its comparatively large KP diaspora, and its history of KP diaspora organization and interaction with the Argentine STI system. The proposed policy framework opens avenues for the development of specific policies at the institutional level tailored to the unique characteristics, local constraints and goals of each organization. The ideas, core principles, and policy blueprint we present have a broader generality and are applicable mutatis mutandis to other developing countries experiencing the brain drain phenomenon.

Neuronal frequency filters can be thought of as constituent building blocks underlying the ability of neuronal systems to process information, generate rhythms and perform computations. How neuronal filters are generated by the concerted activity of a multiplicity of processes (e.g., electric circuit, history-dependent) and interacting time scales within and across levels of neuronal network organization is poorly understood. In this paper, we use mathematical modeling, numerical simulations and analytical calculations of the postsynaptic response to presynaptic spike trains to address this issue in a basic feedforward network motif in the presence of synaptic short-term plasticity (STP, depression and facilitation). The network motif consists of a presynaptic spike-train, a postsynaptic passive cell, and an excitatory (AMPA) chemical synapse. The dynamics of each network component are controlled by one or more time scales. We explain the mechanisms by which the participating time scales shape the neuronal filters at the (i) synaptic update level (the target of the synaptic variable in response to presynaptic spikes), which is shaped by STP, (ii) the synaptic level, and (iii) the postsynaptic membrane potential (PSP) level. We focus on three metrics that gives rise to three types of profiles (curves of the corresponding metrics as a function of the spike-train input frequency or firing rate): (i) peak profiles, (ii) peak-to-trough amplitude profiles, and (iii) phase profiles. The effects of STP are present at the synaptic update level and are communicated to the synaptic level where they interact with the synaptic time scales. The PSP filters result from the interaction between these variables and time scales and the biophysical properties and time scales of the postsynaptic cell. Band-pass filters (BPFs) result from a combination of low-pass filters (LPFs) and high-pass filters (HPFs) operating at the same or different levels of organization. PSP BPFs can be inherited from the synaptic level (STP-mediated BPFs) or they can be generated across levels of organization due to the interaction between (i) a synaptic LPF and the PSP summation-mediated HPF (PSP peaks), and (ii) a synaptic HPF and the PSP summation-mediated LPF (PSP amplitude). These types of BPFs persist in response to more realistic presynaptic spike trains: jittered (randomly perturbed) periodic spike trains and Poisson-distributed spike trains. The response variability is frequency-dependent and is controlled by STP in a non-monotonic frequency manner. The results and lessons learned from the investigation of this basic network motif are a necessary step for the construction of a framework to analyze the mechanisms of generation of neuronal filters in networks with more complex architectures and a variety of interacting cellular, synaptic and plasticity time scales.

In electrically coupled networks, the coupling coefficient (CC) quantifies the strength of the connectivity or the degree to which two participating nodes are coupled in response to an external input to one of them. The CC is measured by computing the relative responses of the indirectly activated (post-J) and the directly activated (pre-J) nodes. In response to time-dependent inputs, the CC is frequency-dependent and has two components capturing the contributions of the amplitude and phase frequency profiles of the participating nodes (quotient of the amplitudes and phase-difference, respectively). The properties and mechanisms of generation of the frequency-dependent CCs (FD-CCs) are largely unknown beyond electrically coupled passive cells and their electrical circuit equivalents. Being linear and 1D, the FD-CCs for passive cells are relatively simple, consisting of low-pass filters (amplitude) and positive and monotonically increasing phase-difference profiles. In linear systems, the FD-CCs depend on the properties of the pre-J cell and the connectivity and are independent of the properties of the post-J cell and the input amplitude. There is a gap in our understanding of the FD-CCs are shaped by (i) how the presence of intrinsic cellular positive and negative feedback currents and the resulting amplification and resonance phenomena, and (ii) the presence of cellular nonlinearities that incorporates the dependence of the FD-CC on the post-J node in addition to the pre-J one. In this paper we address these issues by using biophysically plausible (conductance-based) mathematical modeling, numerical simulations, analytical calculations and dynamical systems tools. We conduct a systematic analysis of the properties of the FD-CC in networks of two electrically connected nodes receiving oscillatory inputs, which is the minimal network architecture that allows for a systematic study of the biophysical and dynamic mechanisms that shape the FD-CC profiles. The participating neurons are either passive cells (low-pass filters) or resonators (band-pass filter) and exhibit lagging or mixed leading-lagging phase-shift responses as the input frequency increases. The formalism and tools we develop and use in this paper can be extended to larger networks with an arbitrary number of nodes, to spatially extended multicompartment neuronal models, and to neurons having a variety of ionic currents. The principles that emerge from our study are directly applicable to these scenarios. Our results make experimentally testable predictions and have implications for the understanding of spike transmission, synchronized firing and coincidence detection in electrically coupled networks in the presence of oscillatory inputs. For clarity, the paper includes an extensive supplementary material section.

Understanding how neuronal networks process oscillatory inputs is key for deciphering the brain’s information processing dynamics. Neuronal filters describe the frequency-dependent relationship of neuronal outputs (e.g., membrane potential amplitude, firing rate) and their inputs for the level of neuronal organization (e.g., cellular, network) considered. Band-pass filters are associated to the notion of resonance and reflect the system’s ability to respond maximally to inputs at a nonzero (resonant) frequency or a limited (resonant) frequency band. The complementary notion of phasonance refers to the ability of a system to exhibit a zero-phase response for a nonzero (phasonant) input frequency. The biophysical and dynamic mechanisms that shape neuronal filters and give raise to preferred frequency responses to oscillatory inputs are poorly understood beyond single cells. Moreover, the mechanisms that control the frequency-dependent communication of information across cells in a network remain unclear. Here, we use mathematical modeling, analytical calculations, computational simulations and dynamical systems tools to investigate how the complex and nonlinear interaction of the systems’s biophysical properties and interacting time scales shape neuronal filters in minimal network models receiving oscillatory inputs with frequencies ( f ) within some range. The minimal networks consist of one directly stimulated cell (cell 1) connected to another (not directly stimulated) cell (cell 2) via graded chemical synapses. Individual cells are either passive or resonators and chemical synapses are either excitatory or inhibitory. The network outputs consist of the voltage peak envelopes and the impedance amplitude and phase profiles (as a function of f ) for the two cells. We introduce the frequency-dependent amplitude K ( f ) and phase ΔΦ( f ) communication coefficients, defined as the ratio of the amplitude responses of the indirectly and directly stimulated cells and the phase difference between these two cells, respectively. Extending previous work, we also introduce the K -curve, parametrized by f , in the phase-space diagram for the voltage variables of the two participating cells. This curve joins the peak voltage values of the two cells in response to the oscillatory inputs and is a geometric representation of the communication coefficient. It allows to interpret the results and explain the dependence of the properties of the communication coefficient in terms of the biophysical and dynamic properties of the participating cells and synaptic connectivity when analytical calculations are not possible. We describe the conditions under which one or the two cells in the network exhibit resonance and phasonance and the conditions under which the network exhibits K -resonance and ΔΦ-phasonance and more complex network responses depending as the complexity of the participating cells increases. For linear networks (linear nodes and linear connectivity), K is proportional to the impedance of the indirectly activated cell 2 and ΔΦ is equal to the phase of the indirectly stimulated cell 2, independent of the directly stimulated cell 1 in both cases. We show that the presence of nonlinear connectivity in the network creates (nonlinear) interactions between the two cells that give rise to K -resonance, ΔΦ-phasonance and more complex responses that are absent in the corresponding linear networks. The results and methods developed in this paper have implications for the processing of information in more complex networks.

Background: Influenza virus is responsible for a yearly epidemic in much of the world. To better predict short-term, seasonal variations in flu infection rates and possible mechanisms of yearly infection variation, we trained a Long Short-Term Memory (LSTM)-based deep neural network on historical Influenza-Like-Illness (ILI), climate, and population data. Methods: Data were collected from the Centers for Disease Control and Prevention (CDC), the National Center for Environmental Information (NCEI), and the United States Census Bureau. The model was initially built in Python using the Keras API and tuned manually. We explored the roles of temperature, precipitation, local wind speed, population size, vaccination rate, and vaccination efficacy. The model was validated using K-fold cross validation as well as forward chaining cross validation and compared to several standard algorithms. Finally, simulation data was generated in R and used for further exploration of the model. Results: We found that temperature is the strongest predictor of ILI rates, but also found that precipitation increased the predictive power of the network. Additionally, the proposed model achieved a +1 week prediction mean absolute error (MAE) of 0.1973. This is less than half of the MAE achieved by the next best performing algorithm. Additionally, the model accurately predicted simulation data. To test the role of temperature in the network, we phase-shifted temperature in time and found a predictable reduction in prediction accuracy. Conclusions: The results of this study suggest that short term flu forecasting may be effectively accomplished using architectures traditionally reserved for time series analysis. The proposed LSTM-based model was able to outperform comparison models at the +1 week time point. Additionally, this model provided insight into the week-to-week effects of climatic and biotic factors and revealed potential patterns in data series. Specifically, we found that temperature is the strongest predictor of seasonal flu infection rates. This information may prove to be especially important for flu forecasting given the uncertain long-term impact of the SARS-CoV-2 pandemic on seasonal influenza.

Degeneracy in dynamic models refers to these situations where multiple combinations of parameter values produce identical patterns for the observable variable. We investigate this phenomenon in two qualitatively different adaptive circuit mechanisms: nonlinear feedback loop (NFBL) and incoherent feedback loop (IFFL). We use minimal models of these circuit types together with analytical calculations, regular perturbation analysis, dynamical systems tools and numerical simulations. In response to constant (or step-constant) inputs, NFBLs and IFFLs produce and overshoot allowing the observable variable to return to a value closer to baseline than the peak (adaptation). We identify the dynamic principles underlying the emergence of degeneracy in adaptive patterns both within and across circuit types in representative NFBL and IFFL models in terms of biologically plausible parameters. We identify the conditions under which degeneracy persists in response to oscillatory inputs with arbitrary frequencies, giving rise to resonance and phasonance degeneracy. This naturally extends to the response of adaptive systems to time-dependent inputs within a relatively large class. By using phase-plane analysis, we provide a mechanistic, dynamical systems-based interpretation of degeneracy. Our results have implication for the understanding of adaptive systems, for the relationship between adaptive and resonant/phasonant systems, for the understanding of complex biochemical circuits, for neuronal computation, and for the development of methods for circuit and dynamical systems reconstruction based on experimental or observational data.

Neuronal frequency filters can be thought of as constituent building blocks underlying the ability of neuronal systems to process information, generate rhythms and perform computations. How neuronal filters are generated by the concerted activity of a multiplicity of processes (e.g., electric circuit, history-dependent) and interacting time scales within and across levels of neuronal network organization is poorly understood. In this paper, we use mathematical modeling, numerical simulations and analytical calculations of the postsynaptic response to presynaptic spike trains to address this issue in a basic feedforward network motif in the presence of synaptic short-term plasticity (STP, depression and facilitation). The network motif consists of a presynaptic spike-train, a postsynaptic passive cell, and an excitatory (AMPA) chemical synapse. The dynamics of each network component are controlled by one or more time scales. We explain the mechanisms by which the participating time scales shape the neuronal filters at the (i) synaptic update level (the target of the synaptic variable in response to presynaptic spikes), which is shaped by STP, (ii) the synaptic level, and (iii) the postsynaptic membrane potential (PSP) level. We focus on three metrics that gives rise to three types of profiles (curves of the corresponding metrics as a function of the spike-train input frequency or firing rate): (i) peak profiles, (ii) peak-to-trough amplitude profiles, and (iii) phase profiles. The effects of STP are present at the synaptic update level and are communicated to the synaptic level where they interact with the synaptic time scales. The PSP filters result from the interaction between these variables and time scales and the biophysical properties and time scales of the postsynaptic cell. Band-pass filters (BPFs) result from a combination of low-pass filters (LPFs) and high-pass filters (HPFs) operating at the same or different levels of organization. PSP BPFs can be inherited from the synaptic level (STP-mediated BPFs) or they can be generated across levels of organization due to the interaction between (i) a synaptic LPF and the PSP summation-mediated HPF (PSP peaks), and (ii) a synaptic HPF and the PSP summation-mediated LPF (PSP amplitude). These types of BPFs persist in response to more realistic presynaptic spike trains: jittered (randomly perturbed) periodic spike trains and Poisson-distributed spike trains. The response variability is frequency-dependent and is controlled by STP in a non-monotonic frequency manner. The results and and lessons learned from the investigation of this basic network motif are a necessary step for the construction of a framework to analyze the mechanisms of generation of neuronal filters in networks with more complex architectures and a variety of interacting cellular, synaptic and plasticity time scales.

The classical Hodgkin–Huxley (HH) point-neuron model of action potential generation is four-dimensional. It consists of four ordinary differential equations describing the dynamics of the membrane potential and three gating variables associated to a transient sodium and a delayed-rectifier potassium ionic currents. Conductance-based models of HH type are higher-dimensional extensions of the classical HH model. They include a number of supplementary state variables associated with other ionic current types, and are able to describe additional phenomena such as subthreshold oscillations, mixed-mode oscillations (subthreshold oscillations interspersed with spikes), clustering and bursting. In this manuscript we discuss biophysically plausible and phenomenological reduced models that preserve the biophysical and/or dynamic description of models of HH type and the ability to produce complex phenomena, but the number of effective dimensions (state variables) is lower. We describe several representative models. We also describe systematic and heuristic methods of deriving reduced models from models of HH type.