Replication Data for: Optimal information injection and transfer mechanisms for active matter reservoir computing (Gaimann and Klopotek, 2025)

<p>
This repository contains raw and post-processed replication data for the publication <a href="https://doi.org/10.48550/arXiv.2509.01799">"Optimal information injection and transfer mechanisms for active matter reservoir computing" (Gaimann and Klopotek, 2025).</a>
</p>

<p>
The datasets contain physical observables recorded during non-equilibrium simulations of active matter systems (swarms) driven by an external force. These simulations serve as information processors in a reservoir computing setup.
</p>

<p>
We provide replication data for all figures and supplementary videos shown in our publication:
<ul>
<li>speed controller, with a linearly attractive driver</li>
<li>speed controller, with a linearly attractive driver, and a driver interaction strength of 2.0</li>
<li>speed controller, with a linearly attractive driver, and a Ridge parameter of 200.0</li>
<li>speed controller, with an inversely attractive driver</li>
<li>speed controller, with an inversely attractive driver, and a driver interaction strength of 2.0</li>
<li>speed controller, with an inversely attractive driver, and a Ridge parameter of 200.0</li>
<li>speed controller, with a repulsive driver (reproduction)</li>
<li>driver repulsion, with speed-controller setting of Lymburn et al. (2021)</li>
<li>driver repulsion, with speed-controller setting of Lymburn et al. (2021), and recorded kernel observations</li>
<li>driver repulsion, with speed-controller setting of Lymburn et al. (2021), with simulation box size 32.0 and observation box size 16.0</li>
<li>driver repulsion, with speed-controller setting of Lymburn et al. (2021), with simulation box size 32.0 and observation box size 32.0</li>
<li>driver repulsion, with speed-controller setting of Lymburn et al. (2021), with simulation box size 64.0 and observation box size 32.0</li>
<li>driver repulsion, with speed-controller setting of Lymburn et al. (2021), with a single agent</li>
<li>driver repulsion, with near-critically damped speed-controller setting</li>
<li>driver repulsion, with near-critically damped speed-controller setting, with a single agent</li>
<li>driver attraction (inverse)</li>
<li>driver attraction (inverse), with a single agent (near-critically damped speed-controller setting)</li>
<li>agent-agent repulsion, with a repulsive driver</li>
<li>agent-agent repulsion, with an (inversely) attractive driver</li>
<li>agent-agent repulsion, with an (inversely) attractive driver, and a driver interaction radius of 2.0</li>
<li>agent-agent repulsion, with an (inversely) attractive driver, and a Lorenz-96 driving protocol</li>
<li>agent-agent repulsion vs. number of agents, with a repulsive driver</li>
<li>agent-agent repulsion vs. number of agents, with an (inversely) attractive driver</li>
<li>agent-agent repulsion vs. number of agents, with an (inversely) attractive driver, an agent-agent repulsion radius of 1.0, and a driver interaction strength of 100.0</li>
<li>agent-agent repulsion vs. number of agents, with an (inversely) attractive driver, an agent-agent repulsion radius of 1.0, and a driver interaction strength of 11.2883789</li>
<li>Ridge parameter vs. target agent speed, with a linearly attractive driver</li>
<li>short-range agent-agent repulsion, with an agent-agent repulsion radius of 1.0</li>
<li>short-range agent-agent repulsion, with an agent-agent repulsion radius of 4.0</li>
<li>long-range agent-agent repulsion, with an agent-agent repulsion radius of 1.0</li>
<li>long-range agent-agent repulsion, with an agent-agent repulsion radius of 4.0</li>
<li>viscoelastic fluid, with a repulsive driver and a driver interaction strength of 100.0</li>
<li>viscoelastic fluid, with a repulsive driver and a driver interaction strength of 1000.0</li>
<li>undriven swarm, with a near-critically damped speed-controller setting</li>
</ul>
</p>

<p>
We note that the controlled variable (config setting) "interaction_types__driver_attraction" adds a linear (homing-style) driver attraction interaction, while the variable "interaction_types__driver_repulsion" combined with a negative value for "interaction_types__driver_repulsion__strength" adds an inversely driver attraction interaction.
</p>

<p>
By default, we use a Lorenz-63 driving protocol that was generated on the fly during the simulation. One scan uses <a href="https://darus.uni-stuttgart.de/file.xhtml?fileId=346448">this Lorenz-96 driving protocol</a>.
</p>

<p>
Each dataset typically contains 400 parameter combinations. Each parameter combination contains four files:
<ul>
<li>config.yaml: controlled variables</li>
<li>simulation_output_train.h5: physical simulation observables in first (training) run</li>
<li>simulation_output_test.h5: physical simulation observables in second (testing) run</li>
<li>reservoir_computer_output.h5: observables related to reservoir computing and time series prediction</li>
</ul>

<p>
The second run has a different chaotic driving protocol, using the same underlying dynamical system (chaotic attractor) but different initial conditions. By default, for all driven simulations, physical observables are only recorded for the test run for a full reservoir computing train/test cycle. Each simulation typically consists of 1,000.00 time units (50,000 integration time steps of 0.02 time units by default). A burn-in phase of 20.0 simulation time units (1,000 integration time steps of 0.02 time units by default) takes place at the beginning of each simulation, which is not recorded. Controlled variables are stored as HDF5 attributes. At each step, we predict by default 25 integration time steps ahead (=0.45283 L63-Lyapunov times).
</p>

<p>
The simulation output files contain:
<ul>
<li>agent_observables: positions, velocities, total forces, velocity fluctuations for all agents; for the first 20.0 simulation time units</li>
<li>frame_observables: driver position (external driving trajectory / input time series), center of mass (taking periodic boundary conditions into account), agent-averaged observables, scalar polarity, scalar rotation; for the full simulation</li>
<li>histograms: binned agent observables and derived quantities; for the full simulation</li>
<li>radially_binned: radial distribution function (agent count), connected velocity correlation, cumulative velocity correlation</li>
<li>time_lags: auto-correlations of agent observables and derived quantities, two-time correlations of agent observables and derived quantities</li>
<li>reference_frame_steps: reference frames (measured in integration steps) for the recording of delay-based quantities in time_lags</li>
</ul>
</p>

<p>The reservoir computer output files contain:<ul>
<li>linear_regression_model: the weights of the linear model (readout layer)</li>
<li>observer_kernel_params: placement positions and widths of the Gaussian observation kernels</li>
<li>predictions_train: n-steps-ahead prediction using the trained linear model, on training data</li>
<li>predictions_test: n-steps-ahead prediction using the trained linear model, on testing data</li>
</ul>
</p>

<p>
Aggregates of physical observables across all parameter combinations in a single dataset are stored as CSV files for convenience; the relevant observable is indicated by the file name. Files that carry the "time_avg" tag are averaged over all simulation time steps, for the "ensemble_avg" averaged over all seeds (only one seed is used here), and for the "array_avg" averaged over all recorded entries (typically samples at different time steps). We provide the following aggregated observables that were processed to generate figures in our associated publication:
<ul>
<li>lymburn_correlation_coefficient: Correlation coefficient, predictive performance</li>
<li>agent_avg_msd_at_lyapunov_time_step=55: Agent-averaged mean squared displacement at the Lyapunov integration time step of the Lorenz-63 attractor (after 55 integration time steps of 0.02 each)</li>
<li>first_local_min.array_avg.h5?connected_velocity_correlation: First local minimum of the connected velocity correlation function, averaged over all recorded samples</li>
<li>ensemble_avg.array_avg.n_activated_kernels_threshold=0.001: Time-averaged number of activated observation kernels for agent count, firing at least above a threshold value of 0.001</li>
<li>ensemble_avg.array_avg.smallest_agent_distance_circle_around_driver: Time-averaged and agent-averaged circle area spanned by the distance between an agent i and its next-nearest neighboring agent j</li>
<li>mean_speed: Agent-averaged speed</li>
<li>scalar_polarity: Scalar polarity</li>
<li>scalar_rotation: Scalar rotation</li>
<li>attanasi_susceptibility: Dynamical susceptibility</li>
</ul>
</p>

<p>
The supplementary videos generated using this raw data are published as: Gaimann, M. U., & Klopotek, M. (2025). Supplementary Videos for: Optimal information injection and transfer mechanisms for active matter reservoir computing (Gaimann and Klopotek, 2025). DaRUS. <a href="https://doi.org/10.18419/DARUS-4806">doi:10.18419/DARUS-4806</a>
</p>

<p>Changelog</p>

<p>V2</p>
<p>
<ul>
<li>Added speed-controller parameter scan datasets and raw Lorenz-63 driver trajectory files for different mean driver speeds (1.0, 2.0, 5.0, 10.0, 15.0).</li>
<li>Added raw Lorenz-63 driver trajectory file (with burn-in phase) and the corresponding NARMA-10 time series file.</li>
<li>Added the raw random uniform driver trajectory file confined to a circle with radius 4.0, with a change interval of 1 (used to compute the short-term memory capacity) and the corresponding NARMA-10 time series file. For the other random trajectories with different change intervals, please refer to the data repository of the prequel to this work (https://doi.org/10.18419/DARUS-4620).</li>
<li>Added datasets for the computation of the short-term memory capacity and the kernel rank for parameter combinations of the speed-controller parameter scan, for an inversely attractive and a linearly attractive driver, with different driving protocols (Lorenz-63, random uniform with different change intervals). These datasets provide the full recorded Gaussian kernel observations of the simulation runs; datasets labeled with "memory-capacity" contain observables related to the memory capacity.</li>
<li>Added datasets for all parameter scans in the main text performed with better statistics (100 random seeds with different agent and driver initial conditions). Due to their large size (~1 TB each), only the configuration files and the final aggregates were uploaded; the raw results files can be reproduced on demand using the ResoBee software package and the specific git hash provided in the metadata of these datasets.</li>
<li>Updated datasets used for figures in the main text with aggregates of more performance metrics (NMSE, NRMSE, sMAPE, Pearson correlation coefficient of the y coordinate computed according to Lymburn et al. (2021)).</li>
</ul>
</p>