Training

Training utilities handle data preparation, normalization, and optimization loops for NN-OpInf models.

Settings

nnopinf.training.get_default_settings([...])

Normalizers

nnopinf.training.AbsNormalizer(x)	Standard normalization class x_norm = (x - x_mean)/x_std
nnopinf.training.StandardNormalizer(x)	Standard normalization class x_norm = (x - x_mean)/x_std
nnopinf.training.MaxAbsNormalizer(x)	MaxAbs normalization class x_norm = (x)/np.amax(np.abs(x_std))
nnopinf.training.NoOpNormalizer(x)	MinMax normalization class x_norm = (x)/np.amax(np.abs(x_std))

Trainers

nnopinf.training.DataClass(training_data, ...)

Data class equipped with member vectors features and response

nnopinf.training.bfgs_step(model, ...)
nnopinf.training.bfgs_step_batch_integrated(...)
nnopinf.training.split_and_normalize(x, ...)
nnopinf.training.prepare_data(inputs, ...)	Take input data, split into test and training, and normalize
nnopinf.training.optimize_weights(model, ...)
nnopinf.training.train(model, variables, y, ...)