Determining at least one node
A method including providing a graph database representation, wherein the graph database representation represents a plurality of nodes in a graph which are interconnected by respective edges, wherein each node of the plurality of the nodes represents a data sample and is assigned to at least one node feature, wherein each edge of the plurality of the edges represents a relationship between the data samples, transforming the graph database representation into a data matrix using a first machine learning algorithm suitable for graph data and an architecture as first layers of a joint machine learning architecture, determining at least one node of the plurality of nodes based on the transformed data matrix using a second machine learning algorithm and a second architecture as second layers of a joint machine learning architecture, and providing the at least one determined node. Further, a computing unit and a computer program product is provided.
1. A computer-implemented method comprising:
a. providing a graph database representation; wherein
the graph database representation represents a plurality of nodes in a graph which are interconnected by respective edges; wherein
each node of the plurality of the nodes represents a data sample and is assigned to at least one node feature; wherein
each edge of the plurality of the edges represents a relationship between the data samples;
b. inputting the graph database representation into a first neural network having a first machine learning algorithm suitable for graph data and a first architecture as first layers of a joint machine learning architecture, the first neural network transforming the graph database representation into a data matrix as an output of the first neural network;
c. inputting the output of the first neural network into a second neural network having a second machine learning algorithm and a second architecture as second layers of a joint machine learning architecture; and
d. outputting at least one determined node from the second neural network, wherein the second neural network considers a concatenation of a global graph embedding with individual node embeddings to generate a score for each data sample, and the node embeddings for each data sample are computed using other data samples within a neighborhood within the graph such that relationships between data samples are considered;
e. training the joint machine architecture using a reinforcement learning approach, in a simulation environment, wherein model weights of the second machine learning algorithm are optimized to maximize a reward function, and a subset of the labels of the training data set is artificially made not observable by the joint machine architecture or a machine learning model, the machine learning model being trained on a subset of the training data whose labels are observable, and the joint machine architecture explores and nominates a subset of data samples from the training data to be labeled, and through a selection of different data samples and subsequent observation of a corresponding reward, the joint machine architecture learns an optimal selection strategy which maximizes the reward function.
2. The computer-implemented method according to claim 1 , wherein the graph database representation is a knowledge graph.
3. The computer-implemented method according to claim 2 , wherein the knowledge graph is a k-nearest neighbor graph.
4. The computer-implemented method according to claim 1 , wherein the first neural network is a graph neural network.
5. The computer-implemented method according to claim 1 , wherein the data matrix is a node latent representation.
6. The computer-implemented method according to claim 1 , wherein the second neural network is a multilayer perceptron.
7. The computer-implemented method according to claim 1 , further comprising labelling the at least one determined node, annotating the at least one determined node and/or transmitting the at least one determined node.
8. The computer-implemented method according to claim 1 , further comprising:
training the first architecture and the second architecture using the output of the first neural network and an output of the second neural network.
9. A computer system comprising:
a processor;
a memory operatively connected to the processor,
wherein the processor is configured to:
a. provide a graph database representation; wherein
the graph database representation represents a plurality of nodes in a graph which are interconnected by respective edges; wherein
each node of the plurality of the nodes represents a data sample and is assigned to at least one node feature; wherein
each edge of the plurality of the edges represents a relationship between the data samples;
b. input the graph database representation into a first neural network having a first machine learning algorithm suitable for graph data and a first architecture as first layers of a joint machine learning architecture, the first neural network transforming the graph database representation into a data matrix as an output of the first neural network;
c. input the output of the first neural network into a second neural network having a second machine learning algorithm and a second architecture as second layers of a joint machine learning architecture; and
d. output at least one determined node from the second neural network, wherein the second neural network considers a concatenation of a global graph embedding with individual node embeddings to generate a score for each data sample, and the node embeddings for each data sample are computed using other data samples within a neighborhood within the graph such that relationships between data samples are considered;
e. train the joint machine architecture using a reinforcement learning approach, in a simulation environment, wherein model weights of the second machine learning algorithm are optimized to maximize a reward function, and a subset of the labels of the training data set is artificially made not observable by the joint machine architecture or a machine learning model, the machine learning model being trained on a subset of the training data whose labels are observable, and the joint machine architecture explores and nominates a subset of data samples from the training data to be labeled, and through a selection of different data samples and subsequent observation of a corresponding reward, the joint machine architecture learns an optimal selection strategy which maximizes the reward function.
10. A non-transitory computer program product directly loadable into an internal memory of a computer, comprising software code portions for performing the steps according to claim 1 , when said computer program product is running on a computer.