Data-driven surrogate model of the Huxley muscle model based on Gated Recurrent Unit (GRU), Temporal Convolutional Network (TCN), Nested Long Short-Term Unit (Nested LSTM)..
collected data from finite element simulations with original Huxley muscle model
preprocessing.R is used to merge all numerical experiments and organize the data into the dataMexie.csv files which are already provided in the data directory.
initialize.py is used to convert data to time series and initialize variables for training
_Train.py files are used to create the models and train them
convert_h5_to_pb.py is used to convert saved model to pb file which can be loaded into C++ code.
postprocessing.py is used to compare original and surrogate model and draw comparison diagrams
surogat-c directory contains FEM-surrogate interface
generator directory contains the numerical experiments generator code, it's used to generate (1) isotonic contraction, (2) quick release, (3) prescribed displacements, (4) prescribed forces experiments
saved models