Computer-readable recording medium, learning method, and learning apparatus
A non-transitory computer-readable recording medium stores a learning program that causes a computer to execute a machine learning process for graph data. The machine learning process includes: generating, from graph data to be subjected to learning, extended graph data where at least some of nodes included in the graph data have a value of the nodes and a value corresponding to presence or absence of an indefinite element at the nodes; and obtaining input tensor data by performing tensor decomposition of the generated extended graph data, performing deep learning with a neural network by inputting the input tensor data into the neural network upon deep learning, and learning a method of the tensor decomposition.
1. A non-transitory computer-readable recording medium having stored therein a learning program that causes a computer to execute a machine learning process for graph data, the machine learning process comprising:
generating, from graph data to be subjected to learning, extended graph data where at least some of nodes included in the graph data have a first value of the node and a second value of a wild card node, which is a node having an indefinite element,
the graph data including nodes that represent chemical elements and edges that represent bonds between the chemical elements,
the first value of the node being a node label;
obtaining input tensor data, performing deep learning with a neural network by inputting the input tensor data into the neural network upon deep learning, and learning a method of the tensor decomposition; and
outputting a discrimination result of data of a graph structure having a partial graph structure including an arbitrary label, wherein
when the graph data represent a structure of a chemical compound, the generating includes generating the extended graph data having values corresponding to bond orders between atoms of the chemical compound, wherein
an incidence matrix having identity representing a structure of a chemical compound serves as the graph data to be subjected to learning, and
the learning includes learning with the incidence matrix serving as an input,
the incidence matrix corresponding to the graph data including the wild card node.
2. A learning method implemented by a computer that executes a machine learning process for graph data, the learning method comprising:
generating, from graph data to be subjected to learning, extended graph data where at least some of nodes included in the graph data have a first value of the node and a second value of a wild card node, which is a node having an indefinite element, using a processor,
the graph data including nodes that represent chemical elements and edges that represent bonds between the chemical elements,
the first value of the node being a node label;
obtaining input tensor data by performing tensor decomposition of the generated extended graph data, performing deep learning with a neural network by inputting the input tensor data into the neural network upon deep learning, and learning a method of the tensor decomposition, using the processor; and
outputting a discrimination result of data of a graph structure having a partial graph structure including an arbitrary label, wherein
when the graph data represent a structure of a chemical compound, the generating includes generating the extended graph data having values corresponding to bond orders between atoms of the chemical compound, wherein
an incidence matrix having identity representing a structure of a chemical compound serves as the graph data to be subjected to learning, and
the learning includes learning with the incidence matrix serving as an input, the incidence matrix corresponding to the graph data including the wild card node.
3. A learning apparatus that performs machine learning for graph data, the learning apparatus comprising:
a memory; and
a processor coupled to the memory, wherein the processor executes a process comprising:
generating, from graph data to be subjected to learning, extended graph data where at least some of nodes included in the graph data have a first value of the node and a second value of a wild card node, which is a node having an indefinite element,
the graph data including nodes that represent chemical elements and edges that represent bonds between the chemical elements,
the first value of the node being a node label;
obtaining input tensor data by performing tensor decomposition of the generated extended graph data, performing deep learning with a neural network by inputting the input tensor data into the neural network upon deep learning, and learning a method of the tensor decomposition; and
outputting a discrimination result of data of a graph structure having a partial graph structure including an arbitrary label, wherein
when the graph data represent a structure of a chemical compound, the generating includes generating the extended graph data having values corresponding to bond orders between atoms of the chemical compound, wherein
an incidence matrix having identity representing a structure of a chemical compound serves as the graph data to be subjected to learning, and
the learning includes learning with the incidence matrix serving as an input, the incidence matrix corresponding to the graph data including the wild card node.