KalmanNet

April.3, 2023 Update "Network Initialization"

Use Xavier initialization for GRU layers and use He initialization for FC layers. (Can mitigate "NaN" initialization problem dramatically)

Feb.13, 2023 Update "batched"

Support a batch of sequences being processed simultaneously, leading to dramatic efficiency improvement.

Link to paper

KalmanNet: Neural Network Aided Kalman Filtering for Partially Known Dynamics

Running code

This branch simulates architecture #2 in our paper. There are main files simulating the linear and non-linear cases respectively.

• Linear case (canonical model or constant acceleration model)

python3 main_linear_canonical.py
python3 main_linear_CA.py

• Non-linear Lorenz Attractor case (Discrete-Time, decimation, or Non-linear observation function)

python3 main_lor_DT.py
python3 main_lor_decimation.py
python3 main_lor_DT_NLobs.py

Parameter settings

• Simulations/model_name/parameters.py

Contain model settings: m, n, f/F, h/H, Q and R.

• Simulations/config.py

Contain dataset size, training parameters and network settings.

• main files

Set flags, paths, etc.