Robust adaptive neuronal state estimation

Robust adaptive nonlinear Kalman filter for joint state and parameter estimation of single neuron models

Setup

To run the simulation with the correct dependencies, it is easiest to set up a python3 virtual environment in the cloned repository:

cd RAUKF
virtualenv -p /usr/bin/python3.8 venv     # create virtual env
echo 'export PYTHONPATH="${PYTHONPATH}:<absolute path to>/RAUKF"' >> venv/bin/activate
source venv/bin/activate                # activate virtual env
pip install -r requirements.txt         # install python dependencies

Simulations

Figures 1, 4, 5:

python sim_ml_2d.py --filter ukf|ekf --input_noise oup [--robust] --sim sim_ml_2d

Example: Figures 4, 5

# Simulations
python sim_ml_2d.py --filter ukf --input_noise oup --sim sim_ml_2d
python sim_ml_2d.py --filter ukf --robust --input_noise oup --sim sim_ml_2d

# Plotting
python -i plots.py --sim sim_ml_2d
>>> plot_state_tracking(data["raukf"], data["ukf"]); plt.show()
>>> plot_parameter_tracking(data["raukf"], data["ukf"]); plt.show()

Figures 6, 7:

python sim_ml_2d.py --filter ukf --input_noise oup --y_fault uni --robust --sim sim_ml_2d_faulty

Figures 8, 9:

python sim_ml_3d.py --filter ukf --input_noise oup [--robust] --sim sim_ml_3d_mismatch

Figures 2, 3:

python -i plots.py --sweep ab|ld