ctpeterson / TensorFlow-to-GVar

Simple Python code for converting a TensorFlow model into a GVar function

TensorFlow-to-GVar

Simple Python code for converting a TensorFlow model into a GVar function. First, create a TensorFlow model using Keras.

import keras
layers = layers = [
    keras.layers.Input(shape = (1,)),
    keras.layers.Dense(2, activation = "relu"),
    keras.layers.Dense(2, activation = "relu"),
    keras.layers.Dense(1, activation = "relu")
]; model = keras.Sequential(layers); model.build();

Now convert the model function into a GVar function.

gvar_model = convert_fixed_input(model)

The gvar_model function that is returned by convert_fixed_input can now be evaluated at a point x. For example, if x=3,

import gvar
print(gvar_model([3.], [gvar.gvar(1., 0.) for p in range(13)]))

will produce [19(0)]. The argument in range is the number of "trainable" parameters that belong to the Keras/TensorFlow model that we created above. The gvar_model function can now be used in a GVar analysis pipeline; most notably, the parameters of the TensorFlow model can be extracted from a least squares fit using Lsqfit.