Deep attention networks reveal the rules of collective motion in zebrafish

Perception of motion salience shapes the emergence of collective motions

Despite the profound implications of self-organization in animal groups for collective behaviors, understanding the fundamental principles and applying them to swarm robotics remains incomplete. Here we propose a heuristic measure of perception of motion salience (MS) to quantify relative motion changes of neighbors from first-person view. Leveraging three large bird-flocking datasets, we explore how this perception of MS relates to the structure of leader-follower (LF) relations, and further perform an individual-level correlation analysis between past perception of MS and future change rate of velocity consensus. We observe prevalence of the positive correlations in real flocks, which demonstrates that individuals will accelerate the convergence of velocity with neighbors who have higher MS. This empirical finding motivates us to introduce the concept of adaptive MS-based (AMS) interaction in swarm model. Finally, we implement AMS in a swarm of ~102 miniature robots. Swarm experiments show the significant advantage of AMS in enhancing self-organization of the swarm for smooth evacuations from confined environments.

Selective social interactions and speed-induced leadership in schooling fish

Animals moving together in groups are believed to interact among each other with effective social forces, such as attraction, repulsion, and alignment. Such forces can be inferred using "force maps," i.e., by analyzing the dependency of the acceleration of a focal individual on relevant variables. Here, we introduce a force map technique suitable for the analysis of the alignment forces experienced by individuals. After validating it using an agent-based model, we apply the force map to experimental data of schooling fish. We observe signatures of an effective alignment force with faster neighbors and an unexpected antialignment with slower neighbors. Instead of an explicit antialignment behavior, we suggest that the observed pattern is the result of a selective attention mechanism, where fish pay less attention to slower neighbors. This mechanism implies the existence of temporal leadership interactions based on relative speeds between neighbors. We present support for this hypothesis both from agent-based modeling as well as from exploring leader-follower relationships in the experimental data.

Dynamical order and many-body correlations in zebrafish show that three is a crowd

Zebrafish constitute a convenient laboratory-based biological system for studying collective behavior. It is possible to interpret a group of zebrafish as a system of interacting agents and to apply methods developed for the analysis of systems of active and even passive particles. Here, we consider the effect of group size. We focus on two- and many-body spatial correlations and dynamical order parameters to investigate the multistate behavior. For geometric reasons, the smallest group of fish which can exhibit this multistate behavior consisting of schooling, milling and swarming is three. We find that states exhibited by groups of three fish are similar to those of much larger groups, indicating that there is nothing more than a gradual change in weighting between the different states as the system size changes. Remarkably, when we consider small groups of fish sampled from a larger group, we find very little difference in the occupancy of the state with respect to isolated groups, nor is there much change in the spatial correlations between the fish. This indicates that fish interact predominantly with their nearest neighbors, perceiving the rest of the group as a fluctuating background. Therefore, the behavior of a crowd of fish is already apparent in groups of three fish.

Predicting the long-term collective behaviour of fish pairs with deep learning

Modern computing has enhanced our understanding of how social interactions shape collective behaviour in animal societies. Although analytical models dominate in studying collective behaviour, this study introduces a deep learning model to assess social interactions in the fish species Hemigrammus rhodostomus. We compare the results of our deep learning approach with experiments and with the results of a state-of-the-art analytical model. To that end, we propose a systematic methodology to assess the faithfulness of a collective motion model, exploiting a set of stringent individual and collective spatio-temporal observables. We demonstrate that machine learning (ML) models of social interactions can directly compete with their analytical counterparts in reproducing subtle experimental observables. Moreover, this work emphasizes the need for consistent validation across different timescales, and identifies key design aspects that enable our deep learning approach to capture both short- and long-term dynamics. We also show that our approach can be extended to larger groups without any retraining, and to other fish species, while retaining the same architecture of the deep learning network. Finally, we discuss the added value of ML in the context of the study of collective motion in animal groups and its potential as a complementary approach to analytical models.

Swimming behavior affects ingestion of microplastics by fish

Microplastics (< 5 mm) are widely found in organisms and have the potential harm to ecosystems. Despite their widespread prevalence in environments, there is high individual varation in the abundance of microplastics found in individuals of the same species. In the present study, juvenile cichlid fish (Chindongo demasoni) were chosen to determine the ingestion personality for microplastics in the laboratory. The visible implant fluorescent tags were used for individual recognition. The fish were fed with microplastic fiber, pellet, and food for comparison. Our results showed that the observation of the behaviors of fish could be successfully matched with subsequent measurements for each individual through the tag method in microplastic research. The difference in the abundance of fiber (0-27 items/ind.) among fish individuals was also observed in our study. Meanwhile, the abundance of fiber showed a positive correlation with the average speed and covered area of fish, which indicates the degree of activity of fish. Moreover, fish with higher speed or a front position had higher capturing times for pellet. Our results suggest that the swimming behaviors of fish affect their ingestion of microplastics, and active fish had a higher likelihood of ingesting microplastics, which might be one of the reasons for the common phenomena, i.e., great individual differences observed in microplastic studies.

Applying collective motion models to study discordant individual behaviours within a school of fish

Computational models of collective motion successfully reproduce the most common behaviours of a school of fish, using only a few elementary interactions between individuals. However, their ability to also reproduce individual behaviours that are discordant from those of the group has not yet been adequately investigated. In this paper, a self-propelled particle model using three interaction zones is considered in relation to the counter-rotation of an individual: a phenomenon observable in real schools of fish milling in a torus, when an individual moves in the same torus but in the opposite direction for a certain period of time. This study shows that the interactions of repulsion, orientation and attraction between individuals moving at constant speed in a three-dimensional space, with asynchronous updating, can generate temporary counter-rotations. The analysis of such events sheds light on the mechanisms that start the counter-rotation and those that end it. Although the contribution of the repulsion interaction is often significant to start and terminate the counter-rotation, it does not prove to be decisive. Indeed, it is observed that even when interactions between individuals are limited to attraction alone, temporary counter-rotations of individuals occur, provided the fish density along the circumference is not uniform. Some of these conclusions, deduced from the simulations performed, are visually consistent with what is observed in some underwater video recordings of milling schools of fish.

An examination of force maps targeted at orientation interactions in moving groups

Force mapping is an established method for inferring the underlying interaction rules thought to govern collective motion from trajectory data. Here we examine the ability of force maps to reconstruct interactions that govern individual's tendency to orient, or align, their heading within a moving group, one of the primary factors thought to drive collective motion, using data from three established general collective motion models. Specifically, our force maps extract how individuals adjust their direction of motion on average as a function of the distance to neighbours and relative alignment in heading with these neighbours, or in more detail as a function of the relative coordinates and relative headings of neighbours. We also examine the association between plots of local alignment and underlying alignment rules. We find that the simpler force maps that examined changes in heading as a function of neighbour distances and differences in heading can qualitatively reconstruct the form of orientation interactions, but also overestimate the spatial range over which these interactions apply. More complex force maps that examine heading changes as a function of the relative coordinates of neighbours (in two spatial dimensions), can also reveal underlying orientation interactions in some cases, but are relatively harder to interpret. Responses to neighbours in both the simpler and more complex force maps are affected by group-level patterns of motion. We also find a correlation between the sizes of regions of high alignment in local alignment plots and the size of the region over which alignment rules apply when only an alignment interaction rule is in action. However, when data derived from more complex models is analysed, the shapes of regions of high alignment are clearly influenced by emergent patterns of motion, and these regions of high alignment can appear even when there is no explicit direct mechanism that governs alignment.

A study of transfer of information in animal collectives using deep learning tools

We studied how the interactions among animals in a collective allow for the transfer of information. We performed laboratory experiments to study how zebrafish in a collective follow a subset of trained animals that move towards a light when it turns on because they expect food at that location. We built some deep learning tools to distinguish from video which are the trained and the naïve animals and to detect when each animal reacts to the light turning on. These tools gave us the data to build a model of interactions that we designed to have a balance between transparency and accuracy. The model finds a low-dimensional function that describes how a naïve animal weights neighbours depending on focal and neighbour variables. According to this low-dimensional function, neighbour speed plays an important role in the interactions. Specifically, a naïve animal weights more a neighbour in front than to the sides or behind, and more so the faster the neighbour is moving; and if the neighbour moves fast enough, the differences coming from the neighbour's relative position largely disappear. From the lens of decision-making, neighbour speed acts as confidence measure about where to go. This article is part of a discussion meeting issue 'Collective behaviour through time'.

Motion leadership and local interaction in two species of freshwater fish (Danio rerio and Hyphessobrycon herbertaxelrodi)

The authors studied momentary motion leadership in small groups of black neon tetra (Hyphessobrycon herbertaxelrodi) and zebrafish (Danio rerio), its relationship with local interaction parameters, such as the acceleration and turning angle of the individuals, and the relative locations of the individuals within the group. The purpose was to know whether leadership tended to be monopolised by certain individuals or whether it was equitably shared between them and if there were differences in leadership sharing between these two species, which are known to have different degrees of cohesion and polarisation. The authors filmed groups of two, three, four and eight fishes of each species and tracked their individual motion by image analysis and trajectory extraction. In both species, motion leadership was not monopolized but egalitarian and very short lived, with leadership shifts distributed randomly over time. The duration of leadership episodes decreased as group size increased and was longer in black neon tetra than in zebrafish. Momentary leaders did not tend to be in the front positions, but closer to the centre of the group. Acceleration and turning angle were more extreme in zebrafish than in black neon tetra and in the momentary leaders than the followers in both species. In general, these differences between species and between leaders/followers were qualitatively similar with some differences in detail, indicating that the relationship between motion leadership and local interaction parameters is likely to conform to a general physical law.

Entropy and Fractal Techniques for Monitoring Fish Behaviour and Welfare in Aquacultural Precision Fish Farming-A Review

In a non-linear system, such as a biological system, the change of the output (e.g., behaviour) is not proportional to the change of the input (e.g., exposure to stressors). In addition, biological systems also change over time, i.e., they are dynamic. Non-linear dynamical analyses of biological systems have revealed hidden structures and patterns of behaviour that are not discernible by classical methods. Entropy analyses can quantify their degree of predictability and the directionality of individual interactions, while fractal dimension (FD) analyses can expose patterns of behaviour within apparently random ones. The incorporation of these techniques into the architecture of precision fish farming (PFF) and intelligent aquaculture (IA) is becoming increasingly necessary to understand and predict the evolution of the status of farmed fish. This review summarizes recent works on the application of entropy and FD techniques to selected individual and collective fish behaviours influenced by the number of fish, tagging, pain, preying/feed search, fear/anxiety (and its modulation) and positive emotional contagion (the social contagion of positive emotions). Furthermore, it presents an investigation of collective and individual interactions in shoals, an exposure of the dynamics of inter-individual relationships and hierarchies, and the identification of individuals in groups. While most of the works have been carried out using model species, we believe that they have clear applications in PFF. The review ends by describing some of the major challenges in the field, two of which are, unsurprisingly, the acquisition of high-quality, reliable raw data and the construction of large, reliable databases of non-linear behavioural data for different species and farming conditions.

Collective motions of fish originate from balanced local perceptual interactions and individual stochastics

The movement of a group of biological individuals, such as fish schools, can evolve from disordered motions to synergistic movements or even ordered patterns. However, the physical origins behind such emergent phenomena of complex systems remain elusive. Here, we established a high-precision protocol for studying the collective behavior of biological groups in quasi-two-dimensional systems. Based on our video recording of ∼600h of fish movements, we extracted a force map of the interactions between fish from their trajectories using the convolution neural network. Presumably, this force implies the fish's perception of the surrounding individuals, the environment, and their response to social information. Interestingly, the fish in our experiments were predominantly in a seemingly disordered swarm state, but their local interactions were clearly specific. Combining such local interactions with the inherent stochasticity of the fish movements, we reproduced the collective motions of the fish through simulations. We demonstrated that a delicate balance between the specific local force and the intrinsic stochasticity is essential for ordered movements. This study presents implications for self-organized systems that use basic physical characterization to produce higher-level sophistication.

Visual processing and collective motion-related decision-making in desert locusts

Collectively moving groups of animals rely on the decision-making of locally interacting individuals in order to maintain swarm cohesion. However, the complex and noisy visual environment poses a major challenge to the extraction and processing of relevant information. We addressed this challenge by studying swarming-related decision-making in desert locust last-instar nymphs. Controlled visual stimuli, in the form of random dot kinematograms, were presented to tethered locust nymphs in a trackball set-up, while monitoring movement trajectory and walking parameters. In a complementary set of experiments, the neurophysiological basis of the observed behavioural responses was explored. Our results suggest that locusts use filtering and discrimination upon encountering multiple stimuli simultaneously. Specifically, we show that locusts are sensitive to differences in speed at the individual conspecific level, and to movement coherence at the group level, and may use these to filter out non-relevant stimuli. The locusts also discriminate and assign different weights to different stimuli, with an observed interactive effect of stimulus size, relative abundance and motion direction. Our findings provide insights into the cognitive abilities of locusts in the domain of decision-making and visual-based collective motion, and support locusts as a model for investigating sensory-motor integration and motion-related decision-making in the intricate swarm environment.

Social behavioral profiling by unsupervised deep learning reveals a stimulative effect of dopamine D3 agonists on zebrafish sociality

It has been a major challenge to systematically evaluate and compare how pharmacological perturbations influence social behavioral outcomes. Although some pharmacological agents are known to alter social behavior, precise description and quantification of such effects have proven difficult. We developed a scalable social behavioral assay for zebrafish named ZeChat based on unsupervised deep learning to characterize sociality at high resolution. High-dimensional and dynamic social behavioral phenotypes are automatically classified using this method. By screening a neuroactive compound library, we found that different classes of chemicals evoke distinct patterns of social behavioral fingerprints. By examining these patterns, we discovered that dopamine D3 agonists possess a social stimulative effect on zebrafish. The D3 agonists pramipexole, piribedil, and 7-hydroxy-DPAT-HBr rescued social deficits in a valproic-acid-induced zebrafish autism model. The ZeChat platform provides a promising approach for dissecting the pharmacology of social behavior and discovering novel social-modulatory compounds.

Visual recognition of social signals by a tectothalamic neural circuit

Social affiliation emerges from individual-level behavioural rules that are driven by conspecific signals^1-5. Long-distance attraction and short-distance repulsion, for example, are rules that jointly set a preferred interanimal distance in swarms^6-8. However, little is known about their perceptual mechanisms and executive neural circuits³. Here we trace the neuronal response to self-like biological motion^9,10, a visual trigger for affiliation in developing zebrafish^2,11. Unbiased activity mapping and targeted volumetric two-photon calcium imaging revealed 21 activity hotspots distributed throughout the brain as well as clustered biological-motion-tuned neurons in a multimodal, socially activated nucleus of the dorsal thalamus. Individual dorsal thalamus neurons encode local acceleration of visual stimuli mimicking typical fish kinetics but are insensitive to global or continuous motion. Electron microscopic reconstruction of dorsal thalamus neurons revealed synaptic input from the optic tectum and projections into hypothalamic areas with conserved social function^12-14. Ablation of the optic tectum or dorsal thalamus selectively disrupted social attraction without affecting short-distance repulsion. This tectothalamic pathway thus serves visual recognition of conspecifics, and dissociates neuronal control of attraction from repulsion during social affiliation, revealing a circuit underpinning collective behaviour.

Selective interaction and its effect on collective motion

Plenty of empirical evidence on biological swarms reveal that interaction between individuals is selective. Each individual’s neighbor is selected based on one or more featured factors. Based on the self-propelled model, we develop a general probability neighbor selection framework to study the effect of four typical featured factors (i.e., distance, bearing, orientation change and bearing change). In this work, two common cases are involved to comprehensively analyze the impact of the four featured factors on the collective motion. One is the flocking, the other is the responsivity to stimulus. The impact of different selection strengths of the featured factors on both cases are investigated. The effect of noise on flocking and different stimulus intensities on responsivity to stimulus are analyzed. This study allows us to get the insight of selective interaction and suggests the potential solution to overcome the trade-off between flocking and responsivity quality.

Measuring inter-individual differences in behavioural types of gilthead seabreams in the laboratory using deep learning

Deep learning allows us to automatize the acquisition of large amounts of behavioural animal data with applications for fisheries and aquaculture. In this work, we have trained an image-based deep learning algorithm, the Faster R-CNN (Faster region-based convolutional neural network), to automatically detect and track the gilthead seabream, Sparus aurata, to search for individual differences in behaviour. We collected videos using a novel Raspberry Pi high throughput recording system attached to individual experimental behavioural arenas. From the continuous recording during behavioural assays, we acquired and labelled a total of 14,000 images and used them, along with data augmentation techniques, to train the network. Then, we evaluated the performance of our network at different training levels, increasing the number of images and applying data augmentation. For every validation step, we processed more than 52,000 images, with and without the presence of the gilthead seabream, in normal and altered (i.e., after the introduction of a non-familiar object to test for explorative behaviour) behavioural arenas. The final and best version of the neural network, trained with all the images and with data augmentation, reached an accuracy of 92,79% ± 6.78% [89.24-96.34] of correct classification and 10.25 ± 61.59 pixels [6.59-13.91] of fish positioning error. Our recording system based on a Raspberry Pi and a trained convolutional neural network provides a valuable non-invasive tool to automatically track fish movements in experimental arenas and, using the trajectories obtained during behavioural tests, to assay behavioural types.

Learning the rules of collective cell migration using deep attention networks

Collective, coordinated cellular motions underpin key processes in all multicellular organisms, yet it has been difficult to simultaneously express the ‘rules’ behind these motions in clear, interpretable forms that effectively capture high-dimensional cell-cell interaction dynamics in a manner that is intuitive to the researcher. Here we apply deep attention networks to analyze several canonical living tissues systems and present the underlying collective migration rules for each tissue type using only cell migration trajectory data. We use these networks to learn the behaviors of key tissue types with distinct collective behaviors—epithelial, endothelial, and metastatic breast cancer cells—and show how the results complement traditional biophysical approaches. In particular, we present attention maps indicating the relative influence of neighboring cells to the learned turning decisions of a ‘focal cell’–the primary cell of interest in a collective setting. Colloquially, we refer to this learned relative influence as ‘attention’, as it serves as a proxy for the physical parameters modifying the focal cell’s future motion as a function of each neighbor cell. These attention networks reveal distinct patterns of influence and attention unique to each model tissue. Endothelial cells exhibit tightly focused attention on their immediate forward-most neighbors, while cells in more expansile epithelial tissues are more broadly influenced by neighbors in a relatively large forward sector. Attention maps of ensembles of more mesenchymal, metastatic cells reveal completely symmetric attention patterns, indicating the lack of any particular coordination or direction of interest. Moreover, we show how attention networks are capable of detecting and learning how these rules change based on biophysical context, such as location within the tissue and cellular crowding. That these results require only cellular trajectories and no modeling assumptions highlights the potential of attention networks for providing further biological insights into complex cellular systems.

Collective behaviors are crucial to the function of multicellular life, with large-scale, coordinated cell migration enabling processes spanning organ formation to coordinated skin healing. However, we lack effective tools to discover and cleanly express collective rules at the level of an individual cell. Here, we employ a carefully structured neural network to extract collective information directly from cell trajectory data. The network is trained on data from various systems, including canonical collective cell systems (HUVEC and MDCK cells) which display visually distinct forms of collective motion, and metastatic cancer cells (MDA-MB-231) which are highly uncoordinated. Using these trained networks, we can produce attention maps for each system, which indicate how a cell within a tissue takes in information from its surrounding neighbors, as a function of weights assigned to those neighbors. Thus for a cell type in which cells tend to follow the path of the cell in front, the attention maps will display high weights for cells spatially forward of the focal cell. We present results in terms of additional metrics, such as accuracy plots and number of interacting cells, and encourage future development of improved metrics.

Dominating lengthscales of zebrafish collective behaviour

Collective behaviour in living systems is observed across many scales, from bacteria to insects, to fish shoals. Zebrafish have emerged as a model system amenable to laboratory study. Here we report a three-dimensional study of the collective dynamics of fifty zebrafish. We observed the emergence of collective behaviour changing between ordered to randomised, upon adaptation to new environmental conditions. We quantify the spatial and temporal correlation functions of the fish and identify two length scales, the persistence length and the nearest neighbour distance, that capture the essence of the behavioural changes. The ratio of the two length scales correlates robustly with the polarisation of collective motion that we explain with a reductionist model of self-propelled particles with alignment interactions.

Systematic Analysis of Emergent Collective Motion Produced by a 3D Hybrid Zonal Model

Emergent patterns of collective motion are thought to arise from local rules of interaction that govern how individuals adjust their velocity in response to the relative locations and velocities of near neighbours. Many models of collective motion apply rules of interaction over a metric scale, based on the distances to neighbouring group members. However, empirical work suggests that some species apply interactions over a topological scale, based on distance determined neighbour rank. Here, we modify an important metric model of collective motion (Couzin et al. in J Theor Biol 218(1):1-11, 2002), so that interactions relating to orienting movements with neighbours and attraction towards more distant neighbours operate over topological scales. We examine the emergent group movement patterns generated by the model as the numbers of neighbours that contribute to orientation- and attraction-based velocity adjustments vary. Like the metric form of the model, simulated groups can fragment (when interactions are influenced by less than 10-15% of the group), swarm and move in parallel, but milling does not occur. The model also generates other cohesive group movements including cases where groups exhibit directed motion without strong overall alignment of individuals. Multiple emergent states are possible for the same set of underlying model parameters in some cases, suggesting sensitivity to initial conditions, and there is evidence that emergent states of the system depend on the history of the system. Groups that do not fragment tend to stay relatively compact in terms of neighbour distances. Even if a group does fragment, individuals remain relatively close to near neighbours, avoiding complete isolation.

Precise visuomotor transformations underlying collective behavior in larval zebrafish

Complex schooling behaviors result from local interactions among individuals. Yet, how sensory signals from neighbors are analyzed in the visuomotor stream of animals is poorly understood. Here, we studied aggregation behavior in larval zebrafish and found that over development larvae transition from overdispersed groups to tight shoals. Using a virtual reality assay, we characterized the algorithms fish use to transform visual inputs from neighbors into movement decisions. We found that young larvae turn away from virtual neighbors by integrating and averaging retina-wide visual occupancy within each eye, and by using a winner-take-all strategy for binocular integration. As fish mature, their responses expand to include attraction to virtual neighbors, which is based on similar algorithms of visual integration. Using model simulations, we show that the observed algorithms accurately predict group structure over development. These findings allow us to make testable predictions regarding the neuronal circuits underlying collective behavior in zebrafish.

Collective behavior emerges from genetically controlled simple behavioral motifs in zebrafish

It is not understood how changes in the genetic makeup of individuals alter the behavior of groups of animals. Here, we find that, even at early larval stages, zebrafish regulate their proximity and alignment with each other. Two simple visual responses, one that measures relative visual field occupancy and one that accounts for global visual motion, suffice to account for the group behavior that emerges. Mutations in genes known to affect social behavior in humans perturb these simple reflexes in individual larval zebrafish and change their emergent collective behaviors in the predicted fashion. Model simulations show that changes in these two responses in individual mutant animals predict well the distinctive collective patterns that emerge in a group. Hence, group behaviors reflect in part genetically defined primitive sensorimotor “motifs,” which are evident even in young larvae.

An Open-Source Framework for Automated High-Throughput Cell Biology Experiments

Modern data analysis methods, such as optimization algorithms or deep learning have been successfully applied to a number of biotechnological and medical questions. For these methods to be efficient, a large number of high-quality and reproducible experiments needs to be conducted, requiring a high degree of automation. Here, we present an open-source hardware and low-cost framework that allows for automatic high-throughput generation of large amounts of cell biology data. Our design consists of an epifluorescent microscope with automated XY stage for moving a multiwell plate containing cells and a perfusion manifold allowing programmed application of up to eight different solutions. Our system is very flexible and can be adapted easily for individual experimental needs. To demonstrate the utility of the system, we have used it to perform high-throughput Ca2+ imaging and large-scale fluorescent labeling experiments.

Unraveling hidden interactions in complex systems with deep learning

Rich phenomena from complex systems have long intrigued researchers, and yet modeling system micro-dynamics and inferring the forms of interaction remain challenging for conventional data-driven approaches, being generally established by scientists with human ingenuity. In this study, we propose AgentNet, a model-free data-driven framework consisting of deep neural networks to reveal and analyze the hidden interactions in complex systems from observed data alone. AgentNet utilizes a graph attention network with novel variable-wise attention to model the interaction between individual agents, and employs various encoders and decoders that can be selectively applied to any desired system. Our model successfully captured a wide variety of simulated complex systems, namely cellular automata (discrete), the Vicsek model (continuous), and active Ornstein-Uhlenbeck particles (non-Markovian) in which, notably, AgentNet's visualized attention values coincided with the true variable-wise interaction strengths and exhibited collective behavior that was absent in the training data. A demonstration with empirical data from a flock of birds showed that AgentNet could identify hidden interaction ranges exhibited by real birds, which cannot be detected by conventional velocity correlation analysis. We expect our framework to open a novel path to investigating complex systems and to provide insight into general process-driven modeling.

Optimal use of simplified social information in sequential decision-making

Social animals can improve their decisions by attending to those made by others. The benefit of this social information must be balanced against the costs of obtaining and processing it. Previous work has focused on rational agents that respond optimally to a sequence of prior decisions. However, full decision sequences are potentially costly to perceive and process. As such, animals may rely on simpler social information, which will affect the social behaviour they exhibit. Here, I derive the optimal policy for agents responding to simplified forms of social information. I show how the behaviour of agents attending to the aggregate number of previous choices differs from those attending to just the most recent prior decision, and I propose a hybrid strategy that provides a highly accurate approximation to the optimal policy with the full sequence. Finally, I analyse the evolutionary stability of each strategy, showing that the hybrid strategy dominates when cognitive costs are low but non-zero, while attending to the most recent decision is dominant when costs are high. These results show that agents can employ highly effective social decision-making rules without requiring unrealistic cognitive capacities, and indicate likely ecological variation in the social information different animals attend to.

Collective states and their transitions in football

Movement, positioning and coordination of player formations is a key aspect for the performance of teams within field-based sports. The increased availability of player tracking data has given rise to numerous studies that focus on the relationship between simple descriptive statistics surrounding team formation and performance. While these existing approaches have provided a high-level a view of team-based spatial formations, there is limited research on the nature of collective movement across players within teams and the establishment of stable collective states within game play. This study draws inspiration from the analysis of collective movement in nature, such as that observed within schools of fish and flocking birds, to explore the existence of collective states within the phases of play in soccer. Order parameters and metrics describing group motion and shape are derived from player movement tracks to uncover the nature of the team's collective states and transitions. This represents a unique addition to the current body of work around the analysis of player movement in team sports. The results from this study demonstrate that sequences of ordered collective behaviours exist with relatively rapid transitions between highly aligned polar and un-ordered swarm behaviours (and vice-versa). Defensive phases of play have a higher proportion of ordered team movement than attacking phases, indicating that movements linked with attacking tactics, such as player dispersion to generate passing and shooting opportunities leads to lower overall collective order. Exploration within this study suggests that defensive tactics, such as reducing the depth or width to close passing opportunities, allows for higher team movement speeds and increased levels of collective order. This study provides a novel view of player movement by visualising the collective states present across the phases of play in football.

Parallel Fish School Tracking Based on Multiple Appearance Feature Detection

A parallel fish school tracking based on multiple-feature fish detection has been proposed in this paper to obtain accurate movement trajectories of a large number of zebrafish. Zebrafish are widely adapted in many fields as an excellent model organism. Due to the non-rigid body, similar appearance, rapid transition, and frequent occlusions, vision-based behavioral monitoring is still a challenge. A multiple appearance feature based fish detection scheme was developed by examining the fish head and center of the fish body based on shape index features. The proposed fish detection has the advantage of locating individual fishes from occlusions and estimating their motion states, which could ensure the stability of tracking multiple fishes. Moreover, a parallel tracking scheme was developed based on the SORT framework by fusing multiple features of individual fish and motion states. The proposed method was evaluated in seven video clips taken under different conditions. These videos contained various scales of fishes, different arena sizes, different frame rates, and various image resolutions. The maximal number of tracking targets reached 100 individuals. The correct tracking ratio was 98.60% to 99.86%, and the correct identification ratio ranged from 97.73% to 100%. The experimental results demonstrate that the proposed method is superior to advanced deep learning-based methods. Nevertheless, this method has real-time tracking ability, which can acquire online trajectory data without high-cost hardware configuration.

A statistical method for identifying different rules of interaction between individuals in moving animal groups

The emergent patterns of collective motion are thought to arise from application of individual-level rules that govern how individuals adjust their velocity as a function of the relative position and behaviours of their neighbours. Empirical studies have sought to determine such rules of interaction applied by 'average' individuals by aggregating data from multiple individuals across multiple trajectory sets. In reality, some individuals within a group may interact differently from others, and such individual differences can have an effect on overall group movement. However, comparisons of rules of interaction used by individuals in different contexts have been largely qualitative. Here we introduce a set of randomization methods designed to determine statistical differences in the rules of interaction between individuals. We apply these methods to a case study of leaders and followers in pairs of freely exploring eastern mosquitofish (Gambusia holbrooki). We find that each of the randomization methods is reliable in terms of: repeatability of p-values, consistency in identification of significant differences and similarity between distributions of randomization-based test statistics. We observe convergence of the distributions of randomization-based test statistics across repeat calculations, and resolution of any ambiguities regarding significant differences as the number of randomization iterations increases.

Examination of an averaging method for estimating repulsion and attraction interactions in moving groups

Groups of animals coordinate remarkable, coherent, movement patterns during periods of collective motion. Such movement patterns include the toroidal mills seen in fish shoals, highly aligned parallel motion like that of flocks of migrating birds, and the swarming of insects. Since the 1970's a wide range of collective motion models have been studied that prescribe rules of interaction between individuals, and that are capable of generating emergent patterns that are visually similar to those seen in real animal group. This does not necessarily mean that real animals apply exactly the same interactions as those prescribed in models. In more recent work, researchers have sought to infer the rules of interaction of real animals directly from tracking data, by using a number of techniques, including averaging methods. In one of the simplest formulations, the averaging methods determine the mean changes in the components of the velocity of an individual over time as a function of the relative coordinates of group mates. The averaging methods can also be used to estimate other closely related quantities including the mean relative direction of motion of group mates as a function of their relative coordinates. Since these methods for extracting interaction rules and related quantities from trajectory data are relatively new, the accuracy of these methods has had limited inspection. In this paper, we examine the ability of an averaging method to reveal prescribed rules of interaction from data generated by two individual based models for collective motion. Our work suggests that an averaging method can capture the qualitative features of underlying interactions from trajectory data alone, including repulsion and attraction effects evident in changes in speed and direction of motion, and the presence of a blind zone. However, our work also illustrates that the output from a simple averaging method can be affected by emergent group level patterns of movement, and the sizes of the regions over which repulsion and attraction effects are apparent can be distorted depending on how individuals combine interactions with multiple group mates.

Deep learning-assisted comparative analysis of animal trajectories with DeepHL

A comparative analysis of animal behavior (e.g., male vs. female groups) has been widely used to elucidate behavior specific to one group since pre-Darwinian times. However, big data generated by new sensing technologies, e.g., GPS, makes it difficult for them to contrast group differences manually. This study introduces DeepHL, a deep learning-assisted platform for the comparative analysis of animal movement data, i.e., trajectories. This software uses a deep neural network based on an attention mechanism to automatically detect segments in trajectories that are characteristic of one group. It then highlights these segments in visualized trajectories, enabling biologists to focus on these segments, and helps them reveal the underlying meaning of the highlighted segments to facilitate formulating new hypotheses. We tested the platform on a variety of trajectories of worms, insects, mice, bears, and seabirds across a scale from millimeters to hundreds of kilometers, revealing new movement features of these animals.

Decoding collective communications using information theory tools

Organisms have evolved sensory mechanisms to extract pertinent information from their environment, enabling them to assess their situation and act accordingly. For social organisms travelling in groups, like the fish in a school or the birds in a flock, sharing information can further improve their situational awareness and reaction times. Data on the benefits and costs of social coordination, however, have largely allowed our understanding of why collective behaviours have evolved to outpace our mechanistic knowledge of how they arise. Recent studies have begun to correct this imbalance through fine-scale analyses of group movement data. One approach that has received renewed attention is the use of information theoretic (IT) tools like mutual information, transfer entropy and causation entropy, which can help identify causal interactions in the type of complex, dynamical patterns often on display when organisms act collectively. Yet, there is a communications gap between studies focused on the ecological constraints and solutions of collective action with those demonstrating the promise of IT tools in this arena. We attempt to bridge this divide through a series of ecologically motivated examples designed to illustrate the benefits and challenges of using IT tools to extract deeper insights into the interaction patterns governing group-level dynamics. We summarize some of the approaches taken thus far to circumvent existing challenges in this area and we conclude with an optimistic, yet cautionary perspective.

Genetic Control of Collective Behavior in Zebrafish

Many animals, including humans, have evolved to live and move in groups. In humans, disrupted social interactions are a fundamental feature of many psychiatric disorders. However, we know little about how genes regulate social behavior. Zebrafish may serve as a powerful model to explore this question. By comparing the behavior of wild-type fish with 90 mutant lines, we show that mutations of genes associated with human psychiatric disorders can alter the collective behavior of adult zebrafish. We identify three categories of behavioral variation across mutants: "scattered," in which fish show reduced cohesion; "coordinated," in which fish swim more in aligned schools; and "huddled," in which fish form dense but disordered groups. Changes in individual interaction rules can explain these differences. This work demonstrates how emergent patterns in animal groups can be altered by genetic changes in individuals and establishes a framework for understanding the fundamentals of social information processing.

Peripheral axonal ensheathment is regulated by RalA GTPase and the exocyst complex

Axon ensheathment is fundamental for fast impulse conduction and the normal physiological functioning of the nervous system. Defects in axonal insulation lead to debilitating conditions, but, despite its importance, the molecular players responsible are poorly defined. Here, we identify RalA GTPase as a key player in axon ensheathment in Drosophila larval peripheral nerves. We demonstrate through genetic analysis that RalA action through the exocyst complex is required in wrapping glial cells to regulate their growth and development. We suggest that the RalA-exocyst pathway controls the targeting of secretory vesicles for membrane growth or for the secretion of a wrapping glia-derived factor that itself regulates growth. In summary, our findings provide a new molecular understanding of the process by which axons are ensheathed in vivo, a process that is crucial for normal neuronal function.