rossler_dynamic_3sp

Train the model for Rossler oscillator

Create data directories

Run the included Python scripts

python create_data_dirs.py

Running

First use cmake:

mkdir build
cd build
cmake ..

Input data

This requires the stochastic simulations in the rossler_stoch_sims directory.

Training

From the build directory:

make learn_sequential
cd ../bin
./learn_sequential

The data is in the data/learn_sequential directory.

The progress can be monitored with the monitor_progress_ixn_params and monitor_progress_moments notebooks in the mathematica directory.

Sampling the learned trajectories

The learned trajectories can be sampled by running from the build directory:

make sample
cd ../bin
./sample

The data is in the data/sample directory.

Activating hidden layer

To activate the hidden layer of some lattice configuration, run from the build directory:

make activate
cd ../bin
./activate

The data is in the data/activate directory.

Paper figures

Paper figures can be generated using the notebooks in the mathematica directory:

• diff_eq_rhs visualizes the learned right hand sides of the differential equations.
• sample visualizes some sampled learned lattices and the hidden layers.
• moments compares the learned moments calculated from the sampled lattices with the moments in the stoch sims in rossler_stoch_sims.
• trajs compares the learned trajectories to a simple MaxEnt model without weights.