maica.core

The maica.data module includes essential utilities and classes to load various chemical data from data files. All chemical data is abstracted to numerical vector or mathematical graph by this module. This module supports the following four essential data types.

• Numerical vector.

• Chemical formula.

• Molecular structure.

• Crystal structure.

In addition to the above essential data types, a type of composite data of them can be handled by this module. It is useful to process complex chemical data, such as molecule-to-molecule interactions and crystal structure in specific conditions.

Environment Variables

This is a core module of MAICA that includes environment variables to define prediction task, machine learning algorithm, and other algorithmic descriptors.

Data Formats

Variable Name	Value	Description
DATA_TYPE_VEC	vec	Numerical vector
DATA_TYPE_FORM	form	Chemical formula
DATA_TYPE_SMILES	smiles	Simplified Molecular-Input Line-Entry System
DATA_TYPE_CIF	cif	Crystallographic information file

Feature Types

Variable Name	Value	Description
FEAT_TYPE_NUM	‘num’	Numerical value
FEAT_TYPE_FORM	‘form’	Chemical formula
FEAT_TYPE_STRUCT	‘struct’	Mathematical graph

Imputation Methods

Variable Name	Value	Description
IMPUTE_MEAN	‘mean’	Fill empty value by mean
IMPUTE_ZERO	‘zero’	Fill empty value by zero
IMPUTE_KNN	‘knn’	Fill empty value by k-nearest neighbor data.

Machine Learning Algorithms

Variable Name	Value	Description
ALG_SKLEARN	‘sklearn’	Abstract algorithm of Scikit-learn models
ALG_PYTORCH	‘pytorch’	Abstract algorithm of PyTorch models
ALG_LR	‘lr’	Linear Regression
ALG_LASSO	‘lasso’	Least absolute shrinkage and selection operator
ALG_DCTR	‘dctr’	Decision tree regression
ALG_SYMR	‘symr’	Symbolic regression
ALG_KRR	‘krr’	Kernel ridge regression
ALG_KNNR	‘knnr’	K-nearest neighbor regression
ALG_SVR	‘svr’	Support vector regression
ALG_GPR	‘gpr’	Gaussian process regression
ALG_GBTR	‘gbtr’	Gradient boosting tree regression
ALG_FCNN	‘fcnn’	Fully-connected neural network
ALG_ATE	‘autoencoder’	Autoencoder
ALG_DOPNET	‘dopnet’	DopNet
ALG_GCN	‘gcn’	Graph convolutional network
ALG_GAT	‘gat’	Graph attention network
ALG_GIN	‘gin’	Graph isomorphism network
ALG_CGCNN	‘cgcnn’	Crystal graph convolutional neural network

Manifold Learning Methods

Variable Name	Value	Description
EMB_TSNE	‘tsne’	t-distributed stochastic neighbor embedding
EMB_SPECT	‘spect’	Spectral clustering

Pipeline Types of the Machine Learning Algorithms

Variable Name	Value
ALGS_SKLEARN	[ALG_SKLEARN, ALG_LR, ALG_LASSO, ALG_DCTR, ALG_SYMR, ALG_KNNR, ALG_SVR, ALG_GPR, ALG_GBTR]
ALG_PYTORCH	[ALG_PYTORCH, ALG_FCNN, ALG_DOPNET, ALG_GCN, ALG_GAT, ALG_GIN, ALG_CGCNN]

Optimization Types

Variable Name	Value	Description
PROB_MIN	‘min’	Minimization problem
PROB_MAX	‘max’	Maximization problem

Meta-heuristic Algorithms

Variable Name	Value	Description
OPT_EP	‘ep’	Evolutionary programming

System Utilities

This is a core module of MAICA that includes base utilities of the library.

set_gpu_runnable(runnable: bool)

Set GPU enable for running machine learning algorithm in GPU.

Parameters

runnable – A flag to indicate the GPU enable.

Returns

None

Example:

>>> set_gpu_runnable(True) # Enable GPU.