Authors: Yichen Zhang Gan He Lei Ma Xiaofei Liu J. J. Johannes Hjorth Alexander Kozlov Yutao He Shenjian Zhang Jeanette Hellgren Kotaleski Yonghong Tian Sten Grillner Kai Du Tiejun Huang Authors: Yichen Zhang Gan He Léo Ma Xiaofei Liu J. J. Johannes Hjorth Alexander Kozlov Yutao He Shenjian Zhang Jeanette Hellgren Kotaleski Yonghong Tian Sten Grillner Kai Du Tiejun Huang Abstract Biophysically detailed multi-compartment models are powerful tools to explore computational principles of the brain and also serve as a theoretical framework to generate algorithms for artificial intelligence (AI) systems. However, the expensive computational cost severely limits the applications in both the neuroscience and AI fields. The major bottleneck during simulating detailed compartment models is the ability of a simulator to solve large systems of linear equations. Here, we present a novel endritic ierarchical cheduling (DHS) method to markedly accelerate such a process. We theoretically prove that the DHS implementation is computationally optimal and accurate. This GPU-based method performs with 2-3 orders of magnitude higher speed than that of the classic serial Hines method in the conventional CPU platform. We build a DeepDendrite framework, which integrates the DHS method and the GPU computing engine of the NEURON simulator and demonstrate applications of DeepDendrite in neuroscience tasks. We investigate how spatial patterns of spine inputs affect neuronal excitability in a detailed human pyramidal neuron model with 25,000 spines. Furthermore, we provide a brief discussion on the potential of DeepDendrite for AI, specifically highlighting its ability to enable the efficient training of biophysically detailed models in typical image classification tasks. D H S Introduction Deciphering the coding and computational principles of neurons is essential to neuroscience. Mammalian brains are composed of more than thousands of different types of neurons with unique morphological and biophysical properties. Even though it is no longer conceptually true, the “point-neuron” doctrine , in which neurons were regarded as simple summing units, is still widely applied in neural computation, especially in neural network analysis. In recent years, modern artificial intelligence (AI) has utilized this principle and developed powerful tools, such as artificial neural networks (ANN) . However, in addition to comprehensive computations at the single neuron level, subcellular compartments, such as neuronal dendrites, can also carry out nonlinear operations as independent computational units , , , , . Furthermore, dendritic spines, small protrusions that densely cover dendrites in spiny neurons, can compartmentalize synaptic signals, allowing them to be separated from their parent dendrites ex vivo and in vivo , , , . 1 2 3 4 5 6 7 8 9 10 11 Simulations using biologically detailed neurons provide a theoretical framework for linking biological details to computational principles. The core of the biophysically detailed multi-compartment model framework , allows us to model neurons with realistic dendritic morphologies, intrinsic ionic conductance, and extrinsic synaptic inputs. The backbone of the detailed multi-compartment model, i.e., dendrites, is built upon the classical Cable theory , which models the biophysical membrane properties of dendrites as passive cables, providing a mathematical description of how electronic signals invade and propagate throughout complex neuronal processes. By incorporating Cable theory with active biophysical mechanisms such as ion channels, excitatory and inhibitory synaptic currents, etc., a detailed multi-compartment model can achieve cellular and subcellular neuronal computations beyond experimental limitations , . 12 13 12 4 7 In addition to its profound impact on neuroscience, biologically detailed neuron models recently were utilized to bridge the gap between neuronal structural and biophysical details and AI. The prevailing technique in the modern AI field is ANNs consisting of point neurons, an analog to biological neural networks. Although ANNs with “backpropagation-of-error” (backprop) algorithm achieve remarkable performance in specialized applications, even beating top human professional players in the games of Go and chess , , the human brain still outperforms ANNs in domains that involve more dynamic and noisy environments , . Recent theoretical studies suggest that dendritic integration is crucial in generating efficient learning algorithms that potentially exceed backprop in parallel information processing , , . Furthermore, a single detailed multi-compartment model can learn network-level nonlinear computations for point neurons by adjusting only the synaptic strength , , demonstrating the full potential of the detailed models in building more powerful brain-like AI systems. Therefore, it is of high priority to expand paradigms in brain-like AI from single detailed neuron models to large-scale biologically detailed networks. 14 15 16 17 18 19 20 21 22 One long-standing challenge of the detailed simulation approach lies in its exceedingly high computational cost, which has severely limited its application to neuroscience and AI. The major bottleneck of the simulation is to solve linear equations based on the foundational theories of detailed modeling , , . To improve efficiency, the classic Hines method reduces the time complexity for solving equations from O(n3) to O(n), which has been widely applied as the core algorithm in popular simulators such as NEURON and GENESIS . However, this method uses a serial approach to process each compartment sequentially. When a simulation involves multiple biophysically detailed dendrites with dendritic spines, the linear equation matrix (“Hines Matrix”) scales accordingly with an increasing number of dendrites or spines (Fig. ), making Hines method no longer practical, since it poses a very heavy burden on the entire simulation. 12 23 24 25 26 1e A reconstructed layer-5 pyramidal neuron model and the mathematical formula used with detailed neuron models. Workflow when numerically simulating detailed neuron models. The equation-solving phase is the bottleneck in the simulation. An example of linear equations in the simulation. Data dependency of the Hines method when solving linear equations in . Le nombre de systèmes d'équations linéaires à résoudre subit une augmentation significative lorsque les modèles sont de plus en plus détaillés. Computational cost (steps taken in the equation solving phase) of the serial Hines method on different types of neuron models. Illustration of different solving methods. Different parts of a neuron are assigned to multiple processing units in parallel methods (mid, right), shown with different colors. In the serial method (left), all compartments are computed with one unit. Computational cost of three methods in when solving equations of a pyramidal model with spines. Run time of different methods on solving equations for 500 pyramidal models with spines. The run time indicates the time consumption of 1 s simulation (solving the equation 40,000 times with a time step of 0.025 ms). p-Hines parallel method in CoreNEURON (on GPU), Branch based branch-based parallel method (on GPU), DHS Dendritic hierarchical scheduling method (on GPU). a b c d c e f g h g i During past decades, tremendous progress has been achieved to speed up the Hines method by using parallel methods at the cellular level, which enables to parallelize the computation of different parts in each cell , , , , , . However, current cellular-level parallel methods often lack an efficient parallelization strategy or lack sufficient numerical accuracy as compared to the original Hines method. 27 28 29 30 31 32 Here, we develop a fully automatic, numerically accurate, and optimized simulation tool that can significantly accelerate computation efficiency and reduce computational cost. In addition, this simulation tool can be seamlessly adopted for establishing and testing neural networks with biological details for machine learning and AI applications. Critically, we formulate the parallel computation of the Hines method as a mathematical scheduling problem and generate a Dendritic Hierarchical Scheduling (DHS) method based on combinatorial optimization and parallel computing theory . We demonstrate that our algorithm provides optimal scheduling without any loss of precision. Furthermore, we have optimized DHS for the currently most advanced GPU chip by leveraging the GPU memory hierarchy and memory accessing mechanisms. Together, DHS can speed up computation 60-1,500 times (Supplementary Table ) compared to the classic simulator NEURON while maintaining identical accuracy. 33 34 1 25 To enable detailed dendritic simulations for use in AI, we next establish the DeepDendrite framework by integrating the DHS-embedded CoreNEURON (an optimized compute engine for NEURON) platform as the simulation engine and two auxiliary modules (I/O module and learning module) supporting dendritic learning algorithms during simulations. DeepDendrite runs on the GPU hardware platform, supporting both regular simulation tasks in neuroscience and learning tasks in AI. 35 Last but not least, we also present several applications using DeepDendrite, targeting a few critical challenges in neuroscience and AI: (1) We demonstrate how spatial patterns of dendritic spine inputs affect neuronal activities with neurons containing spines throughout the dendritic trees (full-spine models). DeepDendrite enables us to explore neuronal computation in a simulated human pyramidal neuron model with ~25,000 dendritic spines. (2) In the discussion we also consider the potential of DeepDendrite in the context of AI, specifically, in creating ANNs with morphologically detailed human pyramidal neurons. Our findings suggest that DeepDendrite has the potential to drastically reduce the training duration, thus making detailed network models more feasible for data-driven tasks. All source code for DeepDendrite, the full-spine models and the detailed dendritic network model are publicly available online (see Code Availability). Our open-source learning framework can be readily integrated with other dendritic learning rules, such as learning rules for nonlinear (full-active) dendrites , burst-dependent synaptic plasticity , and learning with spike prediction . Overall, our study provides a complete set of tools that have the potential to change the current computational neuroscience community ecosystem. By leveraging the power of GPU computing, we envision that these tools will facilitate system-level explorations of computational principles of the brain’s fine structures, as well as promote the interaction between neuroscience and modern AI. 21 20 36 Results Méthode de planification hiérarchique dendritique (DHS) Computing ionic currents and solving linear equations are two critical phases when simulating biophysically detailed neurons, which are time-consuming and pose severe computational burdens. Fortunately, computing ionic currents of each compartment is a fully independent process so that it can be naturally parallelized on devices with massive parallel-computing units like GPUs En conséquence, la résolution d'équations linéaires devient la barrière restante pour le processus de parallélisation (Fig. ). 37 1a–f To tackle this bottleneck, cellular-level parallel methods have been developed, which accelerate single-cell computation by “splitting” a single cell into several compartments that can be computed in parallel , , . However, such methods rely heavily on prior knowledge to generate practical strategies on how to split a single neuron into compartments (Fig. ; Supplementary Fig. ). Hence, it becomes less efficient for neurons with asymmetrical morphologies, e.g., pyramidal neurons and Purkinje neurons. 27 28 38 1g−i 1 We aim to develop a more efficient and precise parallel method for the simulation of biologically detailed neural networks. First, we establish the criteria for the accuracy of a cellular-level parallel method. Based on the theories in parallel computing , we propose three conditions to make sure a parallel method will yield identical solutions as the serial computing Hines method according to the data dependency in the Hines method (see Methods). Then to theoretically evaluate the run time, i.e., efficiency, of the serial and parallel computing methods, we introduce and formulate the concept of computational cost as the number of steps a method takes in solving equations (see Methods). 34 Based on the simulation accuracy and computational cost, we formulate the parallelization problem as a mathematical scheduling problem (see Methods). In simple terms, we view a single neuron as a tree with many nodes (compartments). For parallel threads, we can compute at most nodes at each step, but we need to ensure a node is computed only if all its children nodes have been processed; our goal is to find a strategy with the minimum number of steps for the entire procedure. k k To generate an optimal partition, we propose a method called Dendritic Hierarchical Scheduling (DHS) (theoretical proof is presented in the Methods). The key idea of DHS is to prioritize deep nodes (Fig. ), which results in a hierarchical schedule order. The DHS method includes two steps: analyzing dendritic topology and finding the best partition: (1) Given a detailed model, we first obtain its corresponding dependency tree and calculate the depth of each node (the depth of a node is the number of its ancestor nodes) on the tree (Fig. ). (2) After topology analysis, we search the candidates and pick at most deepest candidate nodes (a node is a candidate only if all its children nodes have been processed). This procedure repeats until all nodes are processed (Fig. ). 2a 2b, c k 2d DHS work flow. DHS processes deepest candidate nodes each iteration. Illustration of calculating node depth of a compartmental model. The model is first converted to a tree structure then the depth of each node is computed. Colors indicate different depth values. Topology analysis on different neuron models. Six neurons with distinct morphologies are shown here. For each model, the soma is selected as the root of the tree so the depth of the node increases from the soma (0) to the distal dendrites. Illustration de l'exécution de DHS sur le modèle en with four threads. Candidates: nodes that can be processed. Selected candidates: nodes that are picked by DHS, i.e., the deepest candidates. Processed nodes: nodes that have been processed before. Parallelization strategy obtained by DHS after the process in . Each node is assigned to one of the four parallel threads. DHS reduces the steps of serial node processing from 14 to 5 by distributing nodes to multiple threads. Coût relatif, c'est-à-dire la proportion du coût de calcul de DHS à celle de la méthode série Hines, lors de l'application de DHS avec des nombres de fils différents sur différents types de modèles. a k b c d b k e d f Take a simplified model with 15 compartments as an example, using the serial computing Hines method, it takes 14 steps to process all nodes, while using DHS with four parallel units can partition its nodes into five subsets (Fig. ): {{9,10,12,14}, {1,7,11,13}, {2,3,4,8}, {6}, {5}}. Because nodes in the same subset can be processed in parallel, it takes only five steps to process all nodes using DHS (Fig. ). 2d 2e Next, we apply the DHS method on six representative detailed neuron models (selected from ModelDB avec un nombre différent de lignes (Fig. ):, including cortical and hippocampal pyramidal neurons , , , cerebellar Purkinje neurons , striatal projection neurons (SPN Les cellules olfactives de la bulbe mitrale , covering the major principal neurons in sensory, cortical and subcortical areas. We then measured the computational cost. The relative computational cost here is defined by the proportion of the computational cost of DHS to that of the serial Hines method. The computational cost, i.e., the number of steps taken in solving equations, drops dramatically with increasing thread numbers. For example, with 16 threads, the computational cost of DHS is 7%-10% as compared to the serial Hines method. Intriguingly, the DHS method reaches the lower bounds of their computational cost for presented neurons when given 16 or even 8 parallel threads (Fig. ), suggesting adding more threads does not improve performance further because of the dependencies between compartments. 39 2f 40 41 42 43 44 45 2f Together, we generate a DHS method that enables automated analysis of the dendritic topology and optimal partition for parallel computing. It is worth noting that DHS finds the optimal partition before the simulation starts, and no extra computation is needed to solve equations. Speeding up DHS by GPU memory boosting DHS computes each neuron with multiple threads, which consumes a vast amount of threads when running neural network simulations. Graphics Processing Units (GPUs) consist of massive processing units (i.e., streaming processors, SPs, Fig. ) for parallel computing . In theory, many SPs on the GPU should support efficient simulation for large-scale neural networks (Fig. ). However, we consistently observed that the efficiency of DHS significantly decreased when the network size grew, which might result from scattered data storage or extra memory access caused by loading and writing intermediate results (Fig. , left). 3a, b 46 3c 3d L'architecture GPU et sa hiérarchie de mémoire. Chaque GPU contient des unités de traitement massives (processeurs de flux). Différents types de mémoire ont un débit différent. Architecture of Streaming Multiprocessors (SMs). Each SM contains multiple streaming processors, registers, and L1 cache. Applying DHS on two neurons, each with four threads. During simulation, each thread executes on one stream processor. Memory optimization strategy on GPU. Top panel, thread assignment and data storage of DHS, before (left) and after (right) memory boosting. Bottom, an example of a single step in triangularization when simulating two neurons in . Processors send a data request to load data for each thread from global memory. Without memory boosting (left), it takes seven transactions to load all request data and some extra transactions for intermediate results. With memory boosting (right), it takes only two transactions to load all request data, registers are used for intermediate results, which further improve memory throughput. Run time of DHS (32 threads each cell) with and without memory boosting on multiple layer 5 pyramidal models with spines. Speed up of memory boosting on multiple layer 5 pyramidal models with spines. Memory boosting brings 1.6-2 times speedup. a b c d d e f We solve this problem by GPU memory boosting, a method to increase memory throughput by leveraging GPU’s memory hierarchy and access mechanism. Based on the memory loading mechanism of GPU, successive threads loading aligned and successively-stored data lead to a high memory throughput compared to accessing scatter-stored data, which reduces memory throughput , . To achieve high throughput, we first align the computing orders of nodes and rearrange threads according to the number of nodes on them. Then we permute data storage in global memory, consistent with computing orders, i.e., nodes that are processed at the same step are stored successively in global memory. Moreover, we use GPU registers to store intermediate results, further strengthening memory throughput. The example shows that memory boosting takes only two memory transactions to load eight request data (Fig. En outre, des expériences sur plusieurs nombres de neurones pyramidaux avec des épines et les modèles de neurones typiques (Fig. ; Supplementary Fig. ) show that memory boosting achieves a 1.2-3.8 times speedup as compared to the naïve DHS. 46 47 3d 3e, f 2 Pour tester de manière complète les performances du DHS avec le boost de la mémoire GPU, nous sélectionnons six modèles de neurones typiques et évaluons le temps de fonctionnement de la résolution des équations de câble sur des nombres massifs de chaque modèle (Fig. ). We examined DHS with four threads (DHS-4) and sixteen threads (DHS-16) for each neuron, respectively. Compared to the GPU method in CoreNEURON, DHS-4 and DHS-16 can speed up about 5 and 15 times, respectively (Fig. ). Moreover, compared to the conventional serial Hines method in NEURON running with a single-thread of CPU, DHS speeds up the simulation by 2-3 orders of magnitude (Supplementary Fig. ), tout en conservant la même précision numérique en présence d'épines dense (Figs supplémentaires. and ), active dendrites (Supplementary Fig. ) and different segmentation strategies (Supplementary Fig. ). 4 4a 3 4 8 7 7 Run time of solving equations for a 1 s simulation on GPU (dt = 0.025 ms, 40,000 iterations in total). CoreNEURON: the parallel method used in CoreNEURON; DHS-4: DHS with four threads for each neuron; DHS-16: DHS with 16 threads for each neuron. , à Visualization of the partition by DHS-4 and DHS-16, each color indicates a single thread. During computation, each thread switches among different branches. a b c DHS creates cell-type-specific optimal partitioning To gain insights into the working mechanism of the DHS method, we visualized the partitioning process by mapping compartments to each thread (every color presents a single thread in Fig. ). The visualization shows that a single thread frequently switches among different branches (Fig. ). Interestingly, DHS generates aligned partitions in morphologically symmetric neurons such as the striatal projection neuron (SPN) and the Mitral cell (Fig. ). By contrast, it generates fragmented partitions of morphologically asymmetric neurons like the pyramidal neurons and Purkinje cell (Fig. ), indicating that DHS splits the neural tree at individual compartment scale (i.e., tree node) rather than branch scale. This cell-type-specific fine-grained partition enables DHS to fully exploit all available threads. 4b, c 4b, c 4b, c 4b, c In summary, DHS and memory boosting generate a theoretically proven optimal solution for solving linear equations in parallel with unprecedented efficiency. Using this principle, we built the open-access DeepDendrite platform, which can be utilized by neuroscientists to implement models without any specific GPU programming knowledge. Below, we demonstrate how we can utilize DeepDendrite in neuroscience tasks. We also discuss the potential of the DeepDendrite framework for AI-related tasks in the Discussion section. DHS enables spine-level modelling As dendritic spines receive most of the excitatory input to cortical and hippocampal pyramidal neurons, striatal projection neurons, etc., their morphologies and plasticity are crucial for regulating neuronal excitability , , , , . However, spines are too small ( ~ 1 μm length) to be directly measured experimentally with regard to voltage-dependent processes. Thus, theoretical work is critical for the full understanding of the spine computations. 10 48 49 50 51 We can model a single spine with two compartments: the spine head where synapses are located and the spine neck that links the spine head to dendrites . The theory predicts that the very thin spine neck (0.1-0.5 um in diameter) electronically isolates the spine head from its parent dendrite, thus compartmentalizing the signals generated at the spine head . However, the detailed model with fully distributed spines on dendrites (“full-spine model”) is computationally very expensive. A common compromising solution is to modify the capacitance and resistance of the membrane by a spine factor , instead of modeling all spines explicitly. Here, the spine factor aims at approximating the spine effect on the biophysical properties of the cell membrane . 52 53 F 54 F 54 Inspired by the previous work of Eyal et al. , we investigated how different spatial patterns of excitatory inputs formed on dendritic spines shape neuronal activities in a human pyramidal neuron model with explicitly modeled spines (Fig. ). Noticeably, Eyal et al. employed the spine factor to incorporate spines into dendrites while only a few activated spines were explicitly attached to dendrites (“few-spine model” in Fig. ). The value of spine in their model was computed from the dendritic area and spine area in the reconstructed data. Accordingly, we calculated the spine density from their reconstructed data to make our full-spine model more consistent with Eyal’s few-spine model. With the spine density set to 1.3 μm-1, the pyramidal neuron model contained about 25,000 spines without altering the model’s original morphological and biophysical properties. Further, we repeated the previous experiment protocols with both full-spine and few-spine models. We use the same synaptic input as in Eyal’s work but attach extra background noise to each sample. By comparing the somatic traces (Fig. ) and spike probability (Fig. ) in full-spine and few-spine models, we found that the full-spine model is much leakier than the few-spine model. In addition, the spike probability triggered by the activation of clustered spines appeared to be more nonlinear in the full-spine model (the solid blue line in Fig. ) than in the few-spine model (the dashed blue line in Fig. ). These results indicate that the conventional F-factor method may underestimate the impact of dense spine on the computations of dendritic excitability and nonlinearity. 51 5a F 5a F 5b, c 5d 5d 5d Nous examinons deux principaux types de modèles: les modèles à peu de colonne vertébrale et les modèles à pleine colonne vertébrale. Les modèles à peu de colonne vertébrale (deux à gauche) sont les modèles qui ont intégré la région de la colonne vertébrale à l'échelle mondiale dans les dendrites et ne joignent que des colonnes individuelles avec des synapses activées. Dans les modèles à pleine colonne vertébrale (deux à droite), tous les colonnes sont explicitement attachés sur des dendrites entières. Nous explorons les effets des entrées synaptiques regroupées et distribuées au hasard sur les modèles à peu de colonne vertébrale et les modèles à pleine colonne verté Somatic voltages recorded for cases in . Colors of the voltage curves correspond to , scale bar: 20 ms, 20 mV. Color-coded voltages during the simulation in at specific times. Colors indicate the magnitude of voltage. Somatic spike probability as a function of the number of simultaneously activated synapses (as in Eyal et al.’s work) for four cases in . Background noise is attached. Run time of experiments in with different simulation methods. NEURON: conventional NEURON simulator running on a single CPU core. CoreNEURON: CoreNEURON simulator on a single GPU. DeepDendrite: DeepDendrite on a single GPU. a b a a c b d a e d In the DeepDendrite platform, both full-spine and few-spine models achieved 8 times speedup compared to CoreNEURON on the GPU platform and 100 times speedup compared to serial NEURON on the CPU platform (Fig. ; Supplementary Table ) while keeping the identical simulation results (Supplementary Figs. and ). Therefore, the DHS method enables explorations of dendritic excitability under more realistic anatomic conditions. 5e 1 4 8 Discussion In this work, we propose the DHS method to parallelize the computation of Hines method and we mathematically demonstrate that the DHS provides an optimal solution without any loss of precision. Next, we implement DHS on the GPU hardware platform and use GPU memory boosting techniques to refine the DHS (Fig. ). When simulating a large number of neurons with complex morphologies, DHS with memory boosting achieves a 15-fold speedup (Supplementary Table ) as compared to the GPU method used in CoreNEURON and up to 1,500-fold speedup compared to serial Hines method in the CPU platform (Fig. ; Supplementary Fig. and Supplementary Table ). Furthermore, we develop the GPU-based DeepDendrite framework by integrating DHS into CoreNEURON. Finally, as a demonstration of the capacity of DeepDendrite, we present a representative application: examine spine computations in a detailed pyramidal neuron model with 25,000 spines. Further in this section, we elaborate on how we have expanded the DeepDendrite framework to enable efficient training of biophysically detailed neural networks. To explore the hypothesis that dendrites improve robustness against adversarial attacks , we train our network on typical image classification tasks. We show that DeepDendrite can support both neuroscience simulations and AI-related detailed neural network tasks with unprecedented speed, therefore significantly promoting detailed neuroscience simulations and potentially for future AI explorations. 55 3 1 4 3 1 56 Decades of efforts have been invested in speeding up the Hines method with parallel methods. Early work mainly focuses on network-level parallelization. In network simulations, each cell independently solves its corresponding linear equations with the Hines method. Network-level parallel methods distribute a network on multiple threads and parallelize the computation of each cell group with each thread , . With network-level methods, we can simulate detailed networks on clusters or supercomputers . In recent years, GPU has been used for detailed network simulation. Because the GPU contains massive computing units, one thread is usually assigned one cell rather than a cell group , , . With further optimization, GPU-based methods achieve much higher efficiency in network simulation. However, the computation inside the cells is still serial in network-level methods, so they still cannot deal with the problem when the “Hines matrix” of each cell scales large. 57 58 59 35 60 61 Cellular-level parallel methods further parallelize the computation inside each cell. The main idea of cellular-level parallel methods is to split each cell into several sub-blocks and parallelize the computation of those sub-blocks , . However, typical cellular-level methods (e.g., the “multi-split” method ) pay less attention to the parallelization strategy. The lack of a fine parallelization strategy results in unsatisfactory performance. To achieve higher efficiency, some studies try to obtain finer-grained parallelization by introducing extra computation operations , , ou faire des approximations sur certains compartiments cruciaux, tout en résolvant des équations linéaires , Ces stratégies de parallélisation à grains plus fins peuvent obtenir une efficacité plus élevée, mais manquent d'une précision numérique suffisante comme dans la méthode Hines originale. 27 28 28 29 38 62 63 64 Unlike previous methods, DHS adopts the finest-grained parallelization strategy, i.e., compartment-level parallelization. By modeling the problem of “how to parallelize” as a combinatorial optimization problem, DHS provides an optimal compartment-level parallelization strategy. Moreover, DHS does not introduce any extra operation or value approximation, so it achieves the lowest computational cost and retains sufficient numerical accuracy as in the original Hines method at the same time. Dendritic spines are the most abundant microstructures in the brain for projection neurons in the cortex, hippocampus, cerebellum, and basal ganglia. As spines receive most of the excitatory inputs in the central nervous system, electrical signals generated by spines are the main driving force for large-scale neuronal activities in the forebrain and cerebellum , . The structure of the spine, with an enlarged spine head and a very thin spine neck—leads to surprisingly high input impedance at the spine head, which could be up to 500 MΩ, combining experimental data and the detailed compartment modeling approach , . Due to such high input impedance, a single synaptic input can evoke a “gigantic” EPSP ( ~ 20 mV) at the spine-head level , , thereby boosting NMDA currents and ion channel currents in the spine . However, in the classic single detailed compartment models, all spines are replaced by the coefficient modifying the dendritic cable geometries . This approach may compensate for the leak currents and capacitance currents for spines. Still, it cannot reproduce the high input impedance at the spine head, which may weaken excitatory synaptic inputs, particularly NMDA currents, thereby reducing the nonlinearity in the neuron’s input-output curve. Our modeling results are in line with this interpretation. 10 11 48 65 48 66 11 F 54 On the other hand, the spine’s electrical compartmentalization is always accompanied by the biochemical compartmentalization , , , resulting in a drastic increase of internal [Ca2+], within the spine and a cascade of molecular processes involving synaptic plasticity of importance for learning and memory. Intriguingly, the biochemical process triggered by learning, in turn, remodels the spine’s morphology, enlarging (or shrinking) the spine head, or elongating (or shortening) the spine neck, which significantly alters the spine’s electrical capacity , , , . Such experience-dependent changes in spine morphology also referred to as “structural plasticity”, have been widely observed in the visual cortex , , somatosensory cortex , , motor cortex , hippocampus , and the basal ganglia in vivo. They play a critical role in motor and spatial learning as well as memory formation. However, due to the computational costs, nearly all detailed network models exploit the “F-factor” approach to replace actual spines, and are thus unable to explore the spine functions at the system level. By taking advantage of our framework and the GPU platform, we can run a few thousand detailed neurons models, each with tens of thousands of spines on a single GPU, while maintaining ~100 times faster than the traditional serial method on a single CPU (Fig. ). Therefore, it enables us to explore of structural plasticity in large-scale circuit models across diverse brain regions. 8 52 67 67 68 69 70 71 72 73 74 75 9 76 5e Another critical issue is how to link dendrites to brain functions at the systems/network level. It has been well established that dendrites can perform comprehensive computations on synaptic inputs due to enriched ion channels and local biophysical membrane properties , , . For example, cortical pyramidal neurons can carry out sublinear synaptic integration at the proximal dendrite but progressively shift to supralinear integration at the distal dendrite . Moreover, distal dendrites can produce regenerative events such as dendritic sodium spikes, calcium spikes, and NMDA spikes/plateau potentials , De tels événements dendriques sont largement observés chez les souris. or even human cortical neurons in vitro, which may offer various logical operations , or gating functions , . Recently, in vivo recordings in awake or behaving mice provide strong evidence that dendritic spikes/plateau potentials are crucial for orientation selectivity in the visual cortex , sensory-motor integration in the whisker system , , and spatial navigation in the hippocampal CA1 region . 5 6 7 77 6 78 6 79 6 79 80 81 82 83 84 85 To establish the causal link between dendrites and animal (including human) patterns of behavior, large-scale biophysically detailed neural circuit models are a powerful computational tool to realize this mission. However, running a large-scale detailed circuit model of 10,000-100,000 neurons generally requires the computing power of supercomputers. It is even more challenging to optimize such models for in vivo data, as it needs iterative simulations of the models. The DeepDendrite framework can directly support many state-of-the-art large-scale circuit models , , , which were initially developed based on NEURON. Moreover, using our framework, a single GPU card such as Tesla A100 could easily support the operation of detailed circuit models of up to 10,000 neurons, thereby providing carbon-efficient and affordable plans for ordinary labs to develop and optimize their own large-scale detailed models. 86 87 88 Recent works on unraveling the dendritic roles in task-specific learning have achieved remarkable results in two directions, i.e., solving challenging tasks such as image classification dataset ImageNet with simplified dendritic networks , and exploring full learning potentials on more realistic neuron , . However, there lies a trade-off between model size and biological detail, as the increase in network scale is often sacrificed for neuron-level complexity , , . Moreover, more detailed neuron models are less mathematically tractable and computationally expensive . 20 21 22 19 20 89 21 Il y a également eu des progrès dans le rôle des dendrites actifs dans les ANN pour les tâches de vision informatique. . proposé une nouvelle architecture ANN avec des dendrites actifs, démontrant des résultats compétitifs dans le multi-task et l'apprentissage continu. used a binary tree to approximate dendrite branching and provided valuable insights into the influence of tree structure on single neurons’ computational capacity. Bird et al. . proposed a dendritic normalization rule based on biophysical behavior, offering an interesting perspective on the contribution of dendritic arbor structure to computation. While these studies offer valuable insights, they primarily rely on abstractions derived from spatially extended neurons, and do not fully exploit the detailed biological properties and spatial information of dendrites. Further investigation is needed to unveil the potential of leveraging more realistic neuron models for understanding the shared mechanisms underlying brain computation and deep learning. 90 91 92 In response to these challenges, we developed DeepDendrite, a tool that uses the Dendritic Hierarchical Scheduling (DHS) method to significantly reduce computational costs and incorporates an I/O module and a learning module to handle large datasets. With DeepDendrite, we successfully implemented a three-layer hybrid neural network, the Human Pyramidal Cell Network (HPC-Net) (Fig. ). This network demonstrated efficient training capabilities in image classification tasks, achieving approximately 25 times speedup compared to training on a traditional CPU-based platform (Fig. ; Supplementary Table ). 6a, b 6f 1 L'illustration du réseau cellulaire pyramidal humain (HPC-Net) pour la classification de l'image. Les images sont transformées en trains de pointe et alimentées dans le modèle réseau. L'apprentissage est déclenché par des signaux d'erreur propagés de la soma aux dendrites. Training with mini-batch. Multiple networks are simulated simultaneously with different images as inputs. The total weight updates ΔW are computed as the average of ΔWi from each network. Comparison of the HPC-Net before and after training. Left, the visualization of hidden neuron responses to a specific input before (top) and after (bottom) training. Right, hidden layer weights (from input to hidden layer) distribution before (top) and after (bottom) training. Workflow of the transfer adversarial attack experiment. We first generate adversarial samples of the test set on a 20-layer ResNet. Then use these adversarial samples (noisy images) to test the classification accuracy of models trained with clean images. Prediction accuracy of each model on adversarial samples after training 30 epochs on MNIST (left) and Fashion-MNIST (right) datasets. Temps d'exécution d'entraînement et de test pour le HPC-Net. La taille du lot est fixée à 16. De gauche, temps d'exécution d'entraînement une époque. Droit, temps d'exécution de test. NEURON parallèle + Python: formation et test sur un seul CPU avec plusieurs cœurs, en utilisant le NEURON parallèle de 40 processus pour simuler le HPC-Net et le code Python supplémentaire pour soutenir la formation mini-batch. DeepDendrite: formation et test du HPC-Net sur un seul GPU avec DeepDendrite. a b c d e f Additionally, it is widely recognized that the performance of Artificial Neural Networks (ANNs) can be undermined by adversarial attacks —intentionally engineered perturbations devised to mislead ANNs. Intriguingly, an existing hypothesis suggests that dendrites and synapses may innately defend against such attacks . Our experimental results utilizing HPC-Net lend support to this hypothesis, as we observed that networks endowed with detailed dendritic structures demonstrated some increased resilience to transfer adversarial attacks compared to standard ANNs, as evident in MNIST and Fashion-MNIST datasets (Fig. ). This evidence implies that the inherent biophysical properties of dendrites could be pivotal in augmenting the robustness of ANNs against adversarial interference. Nonetheless, it is essential to conduct further studies to validate these findings using more challenging datasets such as ImageNet . 93 56 94 95 96 6d, e 97 In conclusion, DeepDendrite has shown remarkable potential in image classification tasks, opening up a world of exciting future directions and possibilities. To further advance DeepDendrite and the application of biologically detailed dendritic models in AI tasks, we may focus on developing multi-GPU systems and exploring applications in other domains, such as Natural Language Processing (NLP), where dendritic filtering properties align well with the inherently noisy and ambiguous nature of human language. Challenges include testing scalability in larger-scale problems, understanding performance across various tasks and domains, and addressing the computational complexity introduced by novel biological principles, such as active dendrites. By overcoming these limitations, we can further advance the understanding and capabilities of biophysically detailed dendritic neural networks, potentially uncovering new advantages, enhancing their robustness against adversarial attacks and noisy inputs, and ultimately bridging the gap between neuroscience and modern AI. Methods Simulation with DHS CoreNEURON simulator ( Utiliser le neurone Nous mettons en œuvre notre méthode Dendritic Hierarchical Scheduling (DHS) dans l’environnement CoreNEURON en modifiant son code source. Tous les modèles qui peuvent être simulés sur la GPU avec CoreNEURON peuvent également être simulés avec DHS en exécutant la commande suivante : 35 https://github.com/BlueBrain/CoreNeuron 25 coreneuron_exec -d /path/to/models -e time --cell-permute 3 --cell-nthread 16 --gpu The usage options are as in Table . 1 Accuracy of the simulation using cellular-level parallel computation Pour assurer l'exactitude de la simulation, nous devons d'abord définir la correctité d'un algorithme parallèle au niveau cellulaire pour juger s'il générera des solutions identiques par rapport aux méthodes sérieuses correctes prouvées, comme la méthode Hines utilisée dans la plate-forme de simulation NEURON. , a parallel algorithm will yield an identical result as its corresponding serial algorithm, if and only if the data process order in the parallel algorithm is consistent with data dependency in the serial method. The Hines method has two symmetrical phases: triangularization and back-substitution. By analyzing the serial computing Hines method , we find that its data dependency can be formulated as a tree structure, where the nodes on the tree represent the compartments of the detailed neuron model. In the triangularization process, the value of each node depends on its children nodes. In contrast, during the back-substitution process, the value of each node is dependent on its parent node (Fig. ). Thus, we can compute nodes on different branches in parallel as their values are not dependent. 34 55 1d Based on the data dependency of the serial computing Hines method, we propose three conditions to make sure a parallel method will yield identical solutions as the serial computing Hines method: (1) The tree morphology and initial values of all nodes are identical to those in the serial computing Hines method; (2) In the triangularization phase, a node can be processed if and only if all its children nodes are already processed; (3) In the back-substitution phase, a node can be processed only if its parent node is already processed. Once a parallel computing method satisfies these three conditions, it will produce identical solutions as the serial computing method. Computational cost of cellular-level parallel computing method To theoretically evaluate the run time, i.e., efficiency, of the serial and parallel computing methods, we introduce and formulate the concept of computational cost as follows: given a tree and threads (basic computational units) to perform triangularization, parallel triangularization equals to divide the node set of into subsets, i.e., = { , , … } where the size of each subset | | ≤ , i.e., at most nodes can be processed each step since there are only Le processus de la phase de triangularisation suit l'ordre : → → … → , and nodes in the same subset can be processed in parallel. So, we define | La taille (la taille de l’ensemble) , i.e., here) as the computational cost of the parallel computing method. In short, we define the computational cost of a parallel method as the number of steps it takes in the triangularization phase. Because the back-substitution is symmetrical with triangularization, the total cost of the entire solving equation phase is twice that of the triangularization phase. T k V T n V V1 V2 Vn Vi k k k V1 V2 Vn Vi V V n Mathematical scheduling problem Based on the simulation accuracy and computational cost, we formulate the parallelization problem as a mathematical scheduling problem: Given a tree = { , } and a positive integer , where is the node-set and is the edge set. Define partition ( ) = { , , … }, | | ≤ , 1 ≤ ≤ n, where | | indicates the cardinal number of subset , i.e., the number of nodes in , and for each node ∈ Tous ses enfants nœuds Voyage ∈children( )} must in a previous subset , where 1 ≤ < . Our goal is to find an optimal partition ( ) whose computational cost | ( )| is minimal. T V E k V E P V V1 V2 Vn Vi k i Vi Vi Vi v Vi c c v Vj j i P* V P* V Here subset consists of all nodes that will be computed at -th step (Fig. ), so | | ≤ indicates that we can compute nodes each step at most because the number of available threads is . The restriction “for each node ∈ , all its children nodes { | ∈children( doit dans un sous-ensemble précédent , where 1 ≤ » ” indicates that node can be processed only if all its child nodes are processed. Vi i 2e Vi k k k v Vi c c v Vj j i v DHS implementation We aim to find an optimal way to parallelize the computation of solving linear equations for each neuron model by solving the mathematical scheduling problem above. To get the optimal partition, DHS first analyzes the topology and calculates the depth ( ) for all nodes ∈ . Then, the following two steps will be executed iteratively until every node ∈ is assigned to a subset: (1) find all candidate nodes and put these nodes into candidate set . A node is a candidate only if all its child nodes have been processed or it does not have any child nodes. (2) if | | ≤ , i.e., the number of candidate nodes is smaller or equivalent to the number of available threads, remove all nodes in and put them into , otherwise, remove deepest nodes from and add them to subset . Label these nodes as processed nodes (Fig. ). After filling in subset , go to step (1) to fill in the next subset . d v v V v V Q Q k Q V*i k Q Vi 2d Vi Vi+1 Correctness proof for DHS After applying DHS to a neural tree = { , }, we get a partition ( et = { , , … }, | | ≤ , 1 ≤ ≤ . Nodes in the same subset will be computed in parallel, taking steps to perform triangularization and back-substitution, respectively. We then demonstrate that the reordering of the computation in DHS will result in a result identical to the serial Hines method. T V E P V V1 V2 Vn Vi k i n Vi n The partition ( ) obtained from DHS decides the computation order of all nodes in a neural tree. Below we demonstrate that the computation order determined by ( ) satisfies the correctness conditions. ( ) is obtained from the given neural tree . Operations in DHS do not modify the tree topology and values of tree nodes (corresponding values in the linear equations), so the tree morphology and initial values of all nodes are not changed, which satisfies condition 1: the tree morphology and initial values of all nodes are identical to those in serial Hines method. In triangularization, nodes are processed from subset to . As shown in the implementation of DHS, all nodes in subset are selected from the candidate set , and a node can be put into only if all its child nodes have been processed. Thus the child nodes of all nodes in are in { , , … }, meaning that a node is only computed after all its children have been processed, which satisfies condition 2: in triangularization, a node can be processed if and only if all its child nodes are already processed. In back-substitution, the computation order is the opposite of that in triangularization, i.e., from to . As shown before, the child nodes of all nodes in are in { , , … }, so parent nodes of nodes in are in { , , … }, which satisfies condition 3: in back-substitution, a node can be processed only if its parent node is already processed. P V P V P V T V1 Vn Vi Q Q Vi V1 V2 Vi-1 Vn V1 Vi V1 V2 V1 à Vi pour + 1 Vi+2 Vn Optimality proof for DHS L’idée de la preuve est que s’il existe une autre solution optimale, elle peut être transformée en notre solution DHS sans augmenter le nombre de étapes requis par l’algorithme, ce qui indique que la solution DHS est optimale. For each subset in ( ), DHS moves (thread number) deepest nodes from the corresponding candidate set to . If the number of nodes in est plus petit que , move all nodes from to . To simplify, we introduce , indicating the depth sum of deepest nodes in . All subsets in ( ) satisfy the max-depth criteria (Supplementary Fig. ): . We then prove that selecting the deepest nodes in each iteration makes an optimal partition. If there exists an optimal partition = { , , … } containing subsets that do not satisfy the max-depth criteria, we can modify the subsets in ( ) so that all subsets consist of the deepest nodes from and the number of subsets ( | ( )|) remain the same after modification. Vi P V k Qi Vi Qi k Qi Vi Di k Qi P V 6a P(V) P*(V) V*1 V*2 V*s P* V Q P* V Without any loss of generalization, we start from the first subset not satisfying the criteria, i.e., . There are two possible cases that will make not satisfy the max-depth criteria: (1) | | < and there exist some valid nodes in that are not put to ; (2) | | = Les nodes dans are not the Les nœuds les plus profonds . V*i V * I V*i k Qi V*i V*i k V*i k Qi For case (1), because some candidate nodes are not put to , these nodes must be in the subsequent subsets. As | | , we can move the corresponding nodes from the subsequent subsets to , which will not increase the number of subsets and make satisfy the criteria (Supplementary Fig. , top). For case (2), | | = , these deeper nodes that are not moved from the candidate set into must be added to subsequent subsets (Supplementary Fig. , bottom). These deeper nodes can be moved from subsequent subsets to through the following method. Assume that after filling , is picked and one of the -th deepest nodes is still in , thus will be put into a subsequent subset ( > ). We first move from to + puis modifier le sous-ensemble + as follows: if | + | ≤ and none of the nodes in + is the parent of node , stop modifying the latter subsets. Otherwise, modify + comme suit (Fig. complémentaire ): if the parent node of is in + , move this parent node to + à ; else move the node with minimum depth from + à to + . After adjusting , modify subsequent subsets + , à + , … with the same strategy. Finally, move from to . V*i V*i < k V*i V*i 6b V*i k Qi V*i 6b V*i V * I v k v’ Qi v’ V * J j i v V*i V*i 1 V*i 1 V*i 1 k V*i 1 v V*i 1 6c v V*i 1 V*i 2 V*i 1 V * I 2 V * I V*i 1 V*i 2 V*j-1 v’ V*j V*i With the modification strategy described above, we can replace all shallower nodes in with the -th deepest node in and keep the number of subsets, i.e., | ( )| the same after modification. We can modify the nodes with the same strategy for all subsets in ( ) that do not contain the deepest nodes. Finally, all subsets ∈ ( ) can satisfy the max-depth criteria, and | ( L’Etat ne change pas après la modification. V*i k Qi P* V P* V V*i P* V P* V In conclusion, DHS generates a partition ( ), and all subsets ∈ ( ) satisfy the max-depth condition: . For any other optimal partition ( ) we can modify its subsets to make its structure the same as ( ), i.e., each subset consists of the deepest nodes in the candidate set, and keep | ( ) the same after modification. So, the partition ( ) obtenu à partir de DHS est l'une des partitions optimales. P V Vi P V P* V P V P* V | P V GPU implementation and memory boosting To achieve high memory throughput, GPU utilizes the memory hierarchy of (1) global memory, (2) cache, (3) register, where global memory has large capacity but low throughput, while registers have low capacity but high throughput. We aim to boost memory throughput by leveraging the memory hierarchy of GPU. GPU employs SIMT (Single-Instruction, Multiple-Thread) architecture. Warps are the basic scheduling units on GPU (a warp is a group of 32 parallel threads). A warp executes the same instruction with different data for different threads . Correctly ordering the nodes is essential for this batching of computation in warps, to make sure DHS obtains identical results as the serial Hines method. When implementing DHS on GPU, we first group all cells into multiple warps based on their morphologies. Cells with similar morphologies are grouped in the same warp. We then apply DHS on all neurons, assigning the compartments of each neuron to multiple threads. Because neurons are grouped into warps, the threads for the same neuron are in the same warp. Therefore, the intrinsic synchronization in warps keeps the computation order consistent with the data dependency of the serial Hines method. Finally, threads in each warp are aligned and rearranged according to the number of compartments. 46 When a warp loads pre-aligned and successively-stored data from global memory, it can make full use of the cache, which leads to high memory throughput, while accessing scatter-stored data would reduce memory throughput. After compartments assignment and threads rearrangement, we permute data in global memory to make it consistent with computing orders so that warps can load successively-stored data when executing the program. Moreover, we put those necessary temporary variables into registers rather than global memory. Registers have the highest memory throughput, so the use of registers further accelerates DHS. Full-spine and few-spine biophysical models We used the published human pyramidal neuron . The membrane capacitance m = 0.44 μF cm-2, membrane resistance m = 48,300 Ω cm2, and axial resistivity a = 261.97 Ω cm. In this model, all dendrites were modeled as passive cables while somas were active. The leak reversal potential l = -83.1 mV. Ion channels such as Na+ and K+ were inserted on soma and initial axon, and their reversal potentials were Na = 67.6 mV, K = -102 mV respectively. All these specific parameters were set the same as in the model of Eyal, et al. , for more details please refer to the published model (ModelDB, access No. 238347). 51 c r r E E E 51 In the few-spine model, the membrane capacitance and maximum leak conductance of the dendritic cables 60 μm away from soma were multiplied by a spine factor to approximate dendritic spines. In this model, spine was set to 1.9. Only the spines that receive synaptic inputs were explicitly attached to dendrites. F F In the full-spine model, all spines were explicitly attached to dendrites. We calculated the spine density with the reconstructed neuron in Eyal, et al. . The spine density was set to 1.3 μm-1, and each cell contained 24994 spines on dendrites 60 μm away from the soma. 51 The morphologies and biophysical mechanisms of spines were the same in few-spine and full-spine models. The length of the spine neck neck = 1.35 μm and the diameter neck = 0.25 μm, whereas the length and diameter of the spine head were 0.944 μm, i.e., the spine head area was set to 2.8 μm2. Both spine neck and spine head were modeled as passive cables, with the reversal potential = -86 mV. The specific membrane capacitance, membrane resistance, and axial resistivity were the same as those for dendrites. L D Le Synaptic inputs We investigated neuronal excitability for both distributed and clustered synaptic inputs. All activated synapses were attached to the terminal of the spine head. For distributed inputs, all activated synapses were randomly distributed on all dendrites. For clustered inputs, each cluster consisted of 20 activated synapses that were uniformly distributed on a single randomly-selected compartment. All synapses were activated simultaneously during the simulation. AMPA-based and NMDA-based synaptic currents were simulated as in Eyal et al.’s work. AMPA conductance was modeled as a double-exponential function and NMDA conduction as a voltage-dependent double-exponential function. For the AMPA model, the specific rise and decay were set to 0.3 and 1.8 ms. For the NMDA model, rise and decay were set to 8.019 and 34.9884 ms, respectively. The maximum conductance of AMPA and NMDA were 0.73 nS and 1.31 nS. τ τ τ τ Background noise We attached background noise to each cell to simulate a more realistic environment. Noise patterns were implemented as Poisson spike trains with a constant rate of 1.0 Hz. Each pattern started at start = 10 ms and lasted until the end of the simulation. We generated 400 noise spike trains for each cell and attached them to randomly-selected synapses. The model and specific parameters of synaptic currents were the same as described in , except that the maximum conductance of NMDA was uniformly distributed from 1.57 to 3.275, resulting in a higher AMPA to NMDA ratio. t Synaptic Inputs Exploring neuronal excitability Nous avons étudié la probabilité de pike lorsque plusieurs synapses étaient activées simultanément. Pour les entrées distribuées, nous avons testé 14 cas, de 0 à 240 synapses activées. Pour les entrées clusterées, nous avons testé 9 cas au total, en activant respectivement de 0 à 12 clusters. Chaque cluster était composé de 20 synapses. Pour chaque cas dans les entrées distribuées et clusterées, nous avons calculé la probabilité de pike avec 50 échantillons aléatoires. La probabilité de pike a été définie comme le rapport du nombre de neurones déclenchés au nombre total d’échantillons. Tous les 1150 échantillons ont été simulés simultanément sur notre plateforme DeepDendrite, réduisant le temps de simulation de jours à minutes. Performing AI tasks with the DeepDendrite platform Conventional detailed neuron simulators lack two functionalities important to modern AI tasks: (1) alternately performing simulations and weight updates without heavy reinitialization and (2) simultaneously processing multiple stimuli samples in a batch-like manner. Here we present the DeepDendrite platform, which supports both biophysical simulating and performing deep learning tasks with detailed dendritic models. DeepDendrite se compose de trois modules (Fig. complémentaire. ): (1) an I/O module; (2) a DHS-based simulating module; (3) a learning module. When training a biophysically detailed model to perform learning tasks, users first define the learning rule, then feed all training samples to the detailed model for learning. In each step during training, the I/O module picks a specific stimulus and its corresponding teacher signal (if necessary) from all training samples and attaches the stimulus to the network model. Then, the DHS-based simulating module initializes the model and starts the simulation. After simulation, the learning module updates all synaptic weights according to the difference between model responses and teacher signals. After training, the learned model can achieve performance comparable to ANN. The testing phase is similar to training, except that all synaptic weights are fixed. 5 HPC-Net model Image classification is a typical task in the field of AI. In this task, a model should learn to recognize the content in a given image and output the corresponding label. Here we present the HPC-Net, a network consisting of detailed human pyramidal neuron models that can learn to perform image classification tasks by utilizing the DeepDendrite platform. HPC-Net has three layers, i.e., an input layer, a hidden layer, and an output layer. The neurons in the input layer receive spike trains converted from images as their input. Hidden layer neurons receive the output of input layer neurons and deliver responses to neurons in the output layer. The responses of the output layer neurons are taken as the final output of HPC-Net. Neurons between adjacent layers are fully connected. For each image stimulus, we first convert each normalized pixel to a homogeneous spike train. For pixel with coordinates ( ) in the image, the corresponding spike train has a constant interspike interval ISI( ) (in ms) which is determined by the pixel value ( ) as shown in Eq. ( ). x, y τ x, y p x, y 1 In our experiment, the simulation for each stimulus lasted 50 ms. All spike trains started at 9 + ISI ms and lasted until the end of the simulation. Then we attached all spike trains to the input layer neurons in a one-to-one manner. The synaptic current triggered by the spike arriving at time is given by τ t0 where is the post-synaptic voltage, the reversal potential syn = 1 mV, the maximum synaptic conductance max = 0.05 μS, and the time constant de 0,5 ms. v E g τ Neurons in the input layer were modeled with a passive single-compartment model. The specific parameters were set as follows: membrane capacitance m = 1.0 μF cm-2, membrane resistance m = 104 Ω cm2, axial resistivity a = 100 Ω cm, reversal potential of passive compartment l = 0 mV. c r r E The hidden layer contains a group of human pyramidal neuron models, receiving the somatic voltages of input layer neurons. The morphology was from Eyal, et al. , and all neurons were modeled with passive cables. The specific membrane capacitance m = 1.5 μF cm-2, membrane resistance m = 48,300 Ω cm2, axial resistivity a = 261.97 Ω cm, and the reversal potential of all passive cables l = 0 mV. Input neurons could make multiple connections to randomly-selected locations on the dendrites of hidden neurons. The synaptic current activated by the -th synapse of the -th input neuron on neuron ’s dendrite is defined as in Eq. ( ), where is the synaptic conductance, is the synaptic weight, is the ReLU-like somatic activation function, and is the somatic voltage of the -th input neuron at time . 51 c r r E k i j 4 gijk Wijk i t Neurons in the output layer were also modeled with a passive single-compartment model, and each hidden neuron only made one synaptic connection to each output neuron. All specific parameters were set the same as those of the input neurons. Synaptic currents activated by hidden neurons are also in the form of Eq. ( ). 4 Image classification with HPC-Net For each input image stimulus, we first normalized all pixel values to 0.0-1.0. Then we converted normalized pixels to spike trains and attached them to input neurons. Somatic voltages of the output neurons are used to compute the predicted probability of each class, as shown in equation , where is the probability of -th class predicted by the HPC-Net, is the average somatic voltage from 20 ms to 50 ms of the -th output neuron, and indicates the number of classes, which equals the number of output neurons. The class with the maximum predicted probability is the final classification result. In this paper, we built the HPC-Net with 784 input neurons, 64 hidden neurons, and 10 output neurons. 6 pi i i C Synaptic plasticity rules for HPC-Net Inspired by previous work , we use a gradient-based learning rule to train our HPC-Net to perform the image classification task. The loss function we use here is cross-entropy, given in Eq. ( ), where is the predicted probability for class , indicates the actual class the stimulus image belongs to, = 1 if input image belongs to class , and = 0 if not. 36 7 pi i yi yi i yi When training HPC-Net, we compute the update for weight (the synaptic weight of the -th synapse connecting neuron to neuron ) at each time step. After the simulation of each image stimulus, is updated as shown in Eq. ( ): Wijk k i j Wijk 8 Here is the learning rate, is the update value at time , , are somatic voltages of neuron and respectively, is the -th synaptic current activated by neuron Les neurones , its synaptic conductance, is the transfer resistance between the -th connected compartment of neuron on neuron ’s dendrite to neuron ’s soma, s = 30 ms, e = 50 ms are start time and end time for learning respectively. For output neurons, the error term can be computed as shown in Eq. ( ). For hidden neurons, the error term is calculated from the error terms in the output layer, given in Eq. ( ). t VJ vi i j Iijk k i j gijk royaume k i j j t t 10 11 Since all output neurons are single-compartment, equals to the input resistance of the corresponding compartment, . Transfer and input resistances are computed by NEURON. Mini-batch training is a typical method in deep learning for achieving higher prediction accuracy and accelerating convergence. DeepDendrite also supports mini-batch training. When training HPC-Net with mini-batch size batch, we make batch copies of HPC-Net. During training, each copy is fed with a different training sample from the batch. DeepDendrite first computes the weight update for each copy separately. After all copies in the current training batch are done, the average weight update is calculated and weights in all copies are updated by this same amount. N N Résistance aux attaques adversaires avec HPC-Net To demonstrate the robustness of HPC-Net, we tested its prediction accuracy on adversarial samples and compared it with an analogous ANN (one with the same 784-64-10 structure and ReLU activation, for fair comparison in our HPC-Net each input neuron only made one synaptic connection to each hidden neuron). We first trained HPC-Net and ANN with the original training set (original clean images). Then we added adversarial noise to the test set and measured their prediction accuracy on the noisy test set. We used the Foolbox , to generate adversarial noise with the FGSM method . ANN was trained with PyTorch , and HPC-Net was trained with our DeepDendrite. For fairness, we generated adversarial noise on a significantly different network model, a 20-layer ResNet . The noise level ranged from 0.02 to 0.2. We experimented on two typical datasets, MNIST and Fashion-MNIST . Results show that the prediction accuracy of HPC-Net is 19% and 16.72% higher than that of the analogous ANN, respectively. 98 99 93 100 101 95 96 Reporting summary Further information on research design is available in the linked to this article. Nature Portfolio Reporting Summary Data availability The data that support the findings of this study are available within the paper, Supplementary Information and Source Data files provided with this paper. The source code and data that used to reproduce the results in Figs. – are available at . The MNIST dataset is publicly available at . The Fashion-MNIST dataset is publicly available at Le . are provided with this paper. 3 6 https://github.com/pkuzyc/DeepDendrite http://yann.lecun.com/exdb/mnist https://github.com/zalandoresearch/fashion-mnist Source data Code availability The source code of DeepDendrite as well as the models and code used to reproduce Figs. – in this study are available at . 3 6 https://github.com/pkuzyc/DeepDendrite References McCulloch, W. S. & Pitts, W. A logical calculus of the ideas immanent in nervous activity. , 115–133 (1943). Bull. Math. Biophys. 5 LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. , 436–444 (2015). Nature 521 Poirazi, P., Brannon, T. & Mel, B. W. Arithmetic of subthreshold synaptic summation in a model CA1 pyramidal cell. , 977–987 (2003). Neuron 37 London, M. & Häusser, M. Dendritic computation. , 503–532 (2005). Annu. Rev. Neurosci. 28 Branco, T. & Häusser, M. The single dendritic branch as a fundamental functional unit in the nervous system. , 494–502 (2010). Curr. Opin. Neurobiol. 20 Stuart, G. J. & Spruston, N. Dendritic integration: 60 years of progress. , 1713–1721 (2015). Nat. Neurosci. 18 Poirazi, P. & Papoutsi, A. Illuminating dendritic function with computational models. , 303–321 (2020). Nat. Rev. Neurosci. 21 Yuste, R. & Denk, W. Dendritic spines as basic functional units of neuronal integration. , 682–684 (1995). Nature 375 Engert, F. & Bonhoeffer, T. Dendritic spine changes associated with hippocampal long-term synaptic plasticity. , 66–70 (1999). Nature 399 Yuste, R. Dendritic spines and distributed circuits. , 772–781 (2011). Neuron 71 Yuste, R. Electrical compartmentalization in dendritic spines. , 429–449 (2013). Annu. Rev. Neurosci. 36 Rall, W. Branching dendritic trees and motoneuron membrane resistivity. , 491–527 (1959). Exp. Neurol. 1 Segev, I. & Rall, W. Computational study of an excitable dendritic spine. , 499–523 (1988). J. Neurophysiol. 60 Silver, D. et al. Mastering the game of go with deep neural networks and tree search. , 484–489 (2016). Nature 529 Silver, D. et al. A general reinforcement learning algorithm that masters chess, shogi, and go through self-play. , 1140–1144 (2018). Science 362 McCloskey, M. & Cohen, N. J. Catastrophic interference in connectionist networks: the sequential learning problem. , 109–165 (1989). Psychol. Learn. Motiv. 24 French, R. M. Catastrophic forgetting in connectionist networks. , 128–135 (1999). Trends Cogn. Sci. 3 Naud, R. & Sprekeler, H. Sparse bursts optimize information transmission in a multiplexed neural code. , E6329–E6338 (2018). Proc. Natl Acad. Sci. USA 115 Sacramento, J., Costa, R. P., Bengio, Y. & Senn, W. Dendritic cortical microcircuits approximate the backpropagation algorithm. in (NeurIPS*,* 2018). Advances in Neural Information Processing Systems 31 (NeurIPS 2018) Payeur, A., Guerguiev, J., Zenke, F., Richards, B. A. & Naud, R. Burst-dependent synaptic plasticity can coordinate learning in hierarchical circuits. , 1010–1019 (2021). Nat. Neurosci. 24 Bicknell, B. A. & Häusser, M. A synaptic learning rule for exploiting nonlinear dendritic computation. , 4001–4017 (2021). Neuron 109 Moldwin, T., Kalmenson, M. & Segev, I. Le clusteron gradient: un modèle de neurone qui apprend à résoudre des tâches de classification via les non-linéarités dendriques, la plasticité structurelle et la descente gradiente. Hodgkin, A. L. & Huxley, A. F. A quantitative description of membrane current and Its application to conduction and excitation in nerve. , 500–544 (1952). J. Physiol. 117 Rall, W. Theory of physiological properties of dendrites. , 1071–1092 (1962). Ann. N. Y. Acad. Sci. 96 Hines, M. L. & Carnevale, N. T. The NEURON simulation environment. , 1179–1209 (1997). Neural Comput. 9 Bower, J. M. & Beeman, D. in (eds Bower, J.M. & Beeman, D.) 17–27 (Springer New York, 1998). The Book of GENESIS: Exploring Realistic Neural Models with the GEneral NEural SImulation System Hines, M. L., Eichner, H. & Schürmann, F. Neuron splitting in compute-bound parallel network simulations enables runtime scaling with twice as many processors. , 203–210 (2008). J. Comput. Neurosci. 25 Hines, M. L., Markram, H. & Schürmann, F. Fully implicit parallel simulation of single neurons. , 439–448 (2008). J. Comput. Neurosci. 25 Ben-Shalom, R., Liberman, G. & Korngreen, A. Accelerating compartmental modeling on a graphical processing unit. , 4 (2013). Front. Neuroinform. 7 Tsuyuki, T., Yamamoto, Y. & Yamazaki, T. Efficient numerical simulation of neuron models with spatial structure on graphics processing units. In (eds Hirose894Akiraet al.) 279–285 (Springer International Publishing, 2016). Proc. 2016 International Conference on Neural Information Processing Vooturi, D. T., Kothapalli, K. & Bhalla, U. S. Parallelizing Hines Matrix Solver in Neuron Simulations on GPU. In 388–397 (IEEE, 2017). Proc. IEEE 24th International Conference on High Performance Computing (HiPC) Huber, F. Efficient tree solver for hines matrices on the GPU. Preprint at (2018). https://arxiv.org/abs/1810.12742 Korte, B. & Vygen, J. 6 edn (Springer, 2018). Combinatorial Optimization Theory and Algorithms Gebali, F. (Wiley, 2011). Algorithms and Parallel Computing Kumbhar, P. et al. CoreNEURON: An optimized compute engine for the NEURON simulator. , 63 (2019). Front. Neuroinform. 13 Urbanczik, R. & Senn, W. Learning by the dendritic prediction of somatic spiking. , 521–528 (2014). Neuron 81 Ben-Shalom, R., Aviv, A., Razon, B. & Korngreen, A. Optimizing ion channel models using a parallel genetic algorithm on graphical processors. , 183–194 (2012). J. Neurosci. Methods 206 Mascagni, M. A parallelizing algorithm for computing solutions to arbitrarily branched cable neuron models. , 105–114 (1991). J. Neurosci. Methods 36 McDougal, R. A. et al. Twenty years of modelDB and beyond: building essential modeling tools for the future of neuroscience. , 1–10 (2017). J. Comput. Neurosci. 42 Migliore, M., Messineo, L. & Ferrante, M. Dendritic Ih selectively blocks temporal summation of unsynchronized distal inputs in CA1 pyramidal neurons. , 5–13 (2004). J. Comput. Neurosci. 16 Hemond, P. et al. Distinct classes of pyramidal cells exhibit mutually exclusive firing patterns in hippocampal area CA3b. , 411–424 (2008). Hippocampus 18 Hay, E., Hill, S., Schürmann, F., Markram, H. & Segev, I. Models of neocortical layer 5b pyramidal cells capturing a wide range of dendritic and perisomatic active Properties. , e1002107 (2011). PLoS Comput. Biol. 7 Masoli, S., Solinas, S. & D’Angelo, E. Le traitement potentiel d’action dans un modèle cellulaire purkinje détaillé révèle un rôle critique pour la compartimentation axonale. Lindroos, R. et al. Neuromodulation des ganglions basaux sur plusieurs échelles temporelles et structurelles – des simulations de MSN de voie directe enquêtent sur l’apparition rapide des effets dopaminergiques et prédisent le rôle de Kv4.2. Migliore, M. et al. Synaptic clusters function as odor operators in the olfactory bulb. , 8499–8504 (2015). Proc. Natl Acad. Sci. USa 112 NVIDIA. . (2021). CUDA C++ Programming Guide https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html NVIDIA. . (2021). CUDA C++ Best Practices Guide https://docs.nvidia.com/cuda/cuda-c-best-practices-guide/index.html Harnett, M. T., Makara, J. K., Spruston, N., Kath, W. L. & Magee, J. C. Synaptic amplification by dendritic spines enhances input cooperativity. , 599–602 (2012). Nature 491 Chiu, C. Q. et al. Compartmentalization of GABAergic inhibition by dendritic spines. , 759–762 (2013). Science 340 Tønnesen, J., Katona, G., Rózsa, B. & Nägerl, U. V. Spine neck plasticity regulates compartmentalization of synapses. , 678–685 (2014). Nat. Neurosci. 17 Eyal, G. et al. Human cortical pyramidal neurons: from spines to spikes via models. , 181 (2018). Front. Cell. Neurosci. 12 Koch, C. & Zador, A. The function of dendritic spines: devices subserving biochemical rather than electrical compartmentalization. , 413–422 (1993). J. Neurosci. 13 Koch, C. Dendritic spines. In (Oxford University Press, 1999). Biophysics of Computation Rapp, M., Yarom, Y. & Segev, I. The impact of parallel fiber background activity on the cable properties of cerebellar purkinje cells. , 518–533 (1992). Neural Comput. 4 Hines, M. Efficient computation of branched nerve equations. , 69–76 (1984). Int. J. Bio-Med. Comput. 15 Nayebi, A. & Ganguli, S. Protection biologiquement inspirée des réseaux profonds contre les attaques adversaires. Pré-impression à https://arxiv.org/abs/1703.09202 (2017). Goddard, N. H. & Hood, G. Simulation à grande échelle à l'aide de la Genèse parallèle. dans le livre de la Genèse: Exploring Realistic Neural Models with the General Neural Simulation System (eds Bower James M. & Beeman David) 349-379 (Springer New York, 1998). Migliore, M., Cannia, C., Lytton, W. W., Markram, H. & Hines, M. L. Parallel network simulations with NEURON. , 119 (2006). J. Comput. Neurosci. 21 Lytton, W. W. et al. Simulation neurotechnologies for advancing brain research: parallelizing large networks in NEURON. , 2063–2090 (2016). Neural Comput. 28 Valero-Lara, P. et al. cuHinesBatch: Solving multiple Hines systems on GPUs human brain project. In 566–575 (IEEE, 2017). Proc. 2017 International Conference on Computational Science Akar, N. A. et al. Arbor—A morphologically-detailed neural network simulation library for contemporary high-performance computing architectures. In 274–282 (IEEE, 2019). Proc. 27th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP) Ben-Shalom, R. et al. NeuroGPU: Accelerating multi-compartment, biophysically detailed neuron simulations on GPUs. , 109400 (2022). J. Neurosci. Methods 366 Rempe, M. J. & Chopp, D. L. A predictor-corrector algorithm for reaction-diffusion equations associated with neural activity on branched structures. , 2139–2161 (2006). SIAM J. Sci. Comput. 28 Kozloski, J. & Wagner, J. An ultrascalable solution to large-scale neural tissue simulation. , 15 (2011). Front. Neuroinform. 5 Jayant, K. et al. Targeted intracellular voltage recordings from dendritic spines using quantum-dot-coated nanopipettes. , 335–342 (2017). Nat. Nanotechnol. 12 Palmer, L. M. & Stuart, G. J. Membrane potential changes in dendritic spines during action potentials and synaptic input. , 6897–6903 (2009). J. Neurosci. 29 Nishiyama, J. & Yasuda, R. Biochemical computation for spine structural plasticity. , 63–75 (2015). Neuron 87 Yuste, R. & Bonhoeffer, T. Morphological changes in dendritic spines associated with long-term synaptic plasticity. , 1071–1089 (2001). Annu. Rev. Neurosci. 24 Holtmaat, A. & Svoboda, K. Experience-dependent structural synaptic plasticity in the mammalian brain. , 647–658 (2009). Nat. Rev. Neurosci. 10 Caroni, P., Donato, F. & Muller, D. Structural plasticity upon learning: regulation and functions. , 478–490 (2012). Nat. Rev. Neurosci. 13 Keck, T. et al. Massive restructuring of neuronal circuits during functional reorganization of adult visual cortex. , 1162 (2008). Nat. Neurosci. 11 Hofer, S. B., Mrsic-Flogel, T. D., Bonhoeffer, T. & Hübener, M. Experience leaves a lasting structural trace in cortical circuits. , 313–317 (2009). Nature 457 Trachtenberg, J. T. et al. Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. , 788–794 (2002). Nature 420 Marik, S. A., Yamahachi, H., McManus, J. N., Szabo, G. & Gilbert, C. D. Axonal dynamics of excitatory and inhibitory neurons in somatosensory cortex. , e1000395 (2010). PLoS Biol. 8 Xu, T. et al. Rapid formation and selective stabilization of synapses for enduring motor memories. , 915–919 (2009). Nature 462 Albarran, E., Raissi, A., Jáidar, O., Shatz, C. J. & Ding, J. B. Enhancing motor learning by increasing the stability of newly formed dendritic spines in the motor cortex. , 3298–3311 (2021). Neuron 109 Branco, T. & Häusser, M. Synaptic integration gradients in single cortical pyramidal cell dendrites. , 885–892 (2011). Neuron 69 Major, G., Larkum, M. E. & Schiller, J. Active properties of neocortical pyramidal neuron dendrites. , 1–24 (2013). Annu. Rev. Neurosci. 36 Gidon, A. et al. Dendritic action potentials and computation in human layer 2/3 cortical neurons. , 83–87 (2020). Science 367 Doron, M., Chindemi, G., Muller, E., Markram, H. & Segev, I. Timed synaptic inhibition shapes NMDA spikes, influencing local dendritic processing and global I/O properties of cortical neurons. , 1550–1561 (2017). Cell Rep. 21 Du, K. et al. Cell-type-specific inhibition of the dendritic plateau potential in striatal spiny projection neurons. , E7612–E7621 (2017). Proc. Natl Acad. Sci. USA 114 Smith, S. L., Smith, I. T., Branco, T. & Häusser, M. Dendritic spikes enhance stimulus selectivity in cortical neurons in vivo. , 115–120 (2013). Nature 503 Xu, N.-l et al. Nonlinear dendritic integration of sensory and motor input during an active sensing task. , 247–251 (2012). Nature 492 Takahashi, N., Oertner, T. G., Hegemann, P. & Larkum, M. E. Les dendrites corticales actives modulent la perception. Science 354, 1587–1590 (2016). Sheffield, M. E. & Dombeck, D. A. Calcium transient prevalence across the dendritic arbour predicts place field properties. , 200–204 (2015). Nature 517 Markram, H. et al. Reconstruction and simulation of neocortical microcircuitry. , 456–492 (2015). Cell 163 Billeh, Y. N. et al. Systematic integration of structural and functional data into multi-scale models of mouse primary visual cortex. , 388–403 (2020). Neuron 106 Hjorth, J. et al. The microcircuits of striatum in silico. , 202000671 (2020). Proc. Natl Acad. Sci. USA 117 Guerguiev, J., Lillicrap, T. P. & Richards, B. A. Towards deep learning with segregated dendrites. , e22901 (2017). elife 6 Iyer, A. et al. Avoiding catastrophe: active dendrites enable multi-task learning in dynamic environments. , 846219 (2022). Front. Neurorobot. 16 Jones, I. S. & Kording, K. P. Might a single neuron solve interesting machine learning problems through successive computations on its dendritic tree? , 1554–1571 (2021). Neural Comput. 33 Bird, A. D., Jedlicka, P. & Cuntz, H. Dendritic normalisation improves learning in sparsely connected artificial neural networks. , e1009202 (2021). PLoS Comput. Biol. 17 Goodfellow, I. J., Shlens, J. & Szegedy, C. Explaining and harnessing adversarial examples. In (ICLR, 2015). 3rd International Conference on Learning Representations (ICLR) Papernot, N., McDaniel, P. & Goodfellow, I. Transferability in machine learning: from phenomena to black-box attacks using adversarial samples. Preprint at (2016). https://arxiv.org/abs/1605.07277 Lecun, Y., Bottou, L., Bengio, Y. & Haffner, P. Gradient-based learning applied to document recognition. , 2278–2324 (1998). Proc. IEEE 86 Xiao, H., Rasul, K. & Vollgraf, R. Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms. Preprint at (2017). http://arxiv.org/abs/1708.07747 Bartunov, S. et al. Évaluation de l'évolutivité des algorithmes et des architectures d'apprentissage profond biologiquement motivés. dans Advances in Neural Information Processing Systems 31 (NeurIPS 2018) (NeurIPS, 2018). Rauber, J., Brendel, W. & Bethge, M. Foolbox: A Python toolbox to benchmark the robustness of machine learning models. In (2017). Reliable Machine Learning in the Wild Workshop, 34th International Conference on Machine Learning Rauber, J., Zimmermann, R., Bethge, M. & Brendel, W. Foolbox native: fast adversarial attacks to benchmark the robustness of machine learning models in PyTorch, TensorFlow, and JAX. , 2607 (2020). J. Open Source Softw. 5 Paszke, A. et al. PyTorch: An imperative style, high-performance deep learning library. In (NeurIPS, 2019). Advances in Neural Information Processing Systems 32 (NeurIPS 2019) He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In 770–778 (IEEE, 2016). Proc. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Acknowledgements The authors sincerely thank Dr. Rita Zhang, Daochen Shi and members at NVIDIA for the valuable technical support of GPU computing. This work was supported by the National Key R&D Program of China (No. 2020AAA0130400) to K.D. and T.H., National Natural Science Foundation of China (No. 61088102) to T.H., National Key R&D Program of China (No. 2022ZD01163005) to L.M., Key Area R&D Program of Guangdong Province (No. 2018B030338001) to T.H., National Natural Science Foundation of China (No. 61825101) to Y.T., Swedish Research Council (VR-M-2020-01652), Swedish e-Science Research Centre (SeRC), EU/Horizon 2020 No. 945539 (HBP SGA3), and KTH Digital Futures to J.H.K., J.H., and A.K., Swedish Research Council (VR-M-2021-01995) and EU/Horizon 2020 no. 945539 (HBP SGA3) to S.G. and A.K. Part of the simulations were enabled by resources provided by the Swedish National Infrastructure for Computing (SNIC) at PDC KTH partially funded by the Swedish Research Council through grant agreement no. 2018-05973. This paper is under CC by 4.0 Deed (Attribution 4.0 International) license. available on nature This paper is under CC by 4.0 Deed (Attribution 4.0 International) license. Disponible dans la nature
