MACHINE_LEARNING

Machine learning section [Edit on GitHub]

Mentions: ⭐PAO-ML

Subsections

• TRAINING_SET

Keywords

• DESCRIPTOR

• GP_NOISE_VAR

• GP_SCALE

• METHOD

• PRIOR

• TOLERANCE

Keyword descriptions

DESCRIPTOR

Type: enum
Default: POTENTIAL

Description: Descriptor used as input for machine learning. [Edit on GitHub]

Valid values:

• POTENTIAL Eigenvalues of local potential matrix

• OVERLAP Eigenvalues of local overlap matrix

• R12 Distance between two atoms (just for testing)

References: Zhu2016

GP_NOISE_VAR

Type: real
Default: 1.00000000E-001

Description: Variance of noise used for Gaussian Process machine learning. [Edit on GitHub]

GP_SCALE

Type: real
Default: 5.00000000E-002

Description: Length scale used for Gaussian Process machine learning. [Edit on GitHub]

Mentions: ⭐PAO-ML

METHOD

Type: enum
Default: GAUSSIAN_PROCESS

Description: Machine learning scheme used to predict PAO basis sets. [Edit on GitHub]

Valid values:

• GAUSSIAN_PROCESS Gaussian Process

• NEURAL_NETWORK Neural Network

• LAZY Just rely on prior

PRIOR

Type: enum
Default: ZERO

Description: Prior used for predictions. [Edit on GitHub]

Valid values:

• ZERO Simply use zero

• MEAN Use average of training-set

TOLERANCE

Type: real
Default: 1.00000000E-002

Description: Maximum variance tolerated when making predictions. [Edit on GitHub]