Graphic neural networks are all the rage, here’s why


Machine learning and deep learning methodologies have seen tremendous advancements in recent times. GNN is a relatively recent deep learning method that belongs to the category of neural networks that work on processing data on graphs. These algorithms can look for information in graphs and predict results using the information gathered.

A chart typically represents data with two components: nodes and edges (which form a connection between two nodes). GNN can be applied to charts to make node, chart and edge level predictions.

How To Start Your Career In Data Science?

Better than CNN?

According to research, CNNs can only work on regular Euclidean data, like images (2D grids) and texts (1D sequences), while these data structures can be considered as instances of graphs. Graphics, on the other hand, are not Euclidean and can be used to study and analyze 3D data. GNNs can provide the reasoning process of the human brain, which sets them apart from other neural networks.

In standard neural networks, dependency information is only seen as the characteristic of nodes. GNNs can propagate the structure of the graph instead of using it as part of functionality.

Research also indicates that in order to present a graph completely, it is necessary to cycle through all the possible orders as input to the model, such as CNNs and RNNs. But GNNs can propagate to each node respectively ignoring the order of entry of nodes.

See also

GNN applications

  • Natural Language Processing – Graphics can be an important part of NLP applications. They can be used for text classification, information retrieval and answering questions.
  • Computer Vision – Although the application of GNN in computer vision is still growing; a lot of progress has been made. GNN algorithms can be used in image classification, human-object interaction, and multi-plane image classification, among others.
  • Physics – This article says that “a physical system can be modeled as the objects of the system and the pairwise interactions between objects”. GNNs can be used by modeling objects as nodes and pairwise interactions as edges.
  • Chemistry and Biology – GNNs have found applications in molecular fingerprints, prediction of chemical reactions, protein interface prediction, drug research, biomedical engineering and more.
  • Traffic Networks – GNNs are also used to forecast traffic movement, volume or density of roads. The nodes can be the sensors installed on the roads, while the distance between the pairs of nodes measures the edges.
  • Recommendation systems – Graphics are increasingly used in user interactions with companies’ products.

Problems remain

The paper named “Graph neural networks: a review of methods and applications » talks about a few remaining issues with GNN, although a lot of progress has been made. They are:

  • Robustness – As part of the neural network family, GNNs are also vulnerable to adversarial attacks. While an attack on images and text focuses on features, adversarial attacks on graphics take additional information into account.
  • Interpretability – The document states that it is important to apply GNN models to real world applications with reliable explanations. Like computer vision and NLP, it is important to consider the interpretability on graphics.
  • Graph Pre Training – Neural network models require a large amount of labeled data. It is expensive to get such a large amount of labeled human data. Thus, self-supervised models are offered to guide models to learn from unlabeled data available on websites. This method has been successful in CV and NLP. Emphasis was also placed on pre-training on graphics, but they have different problem parameters and focus on different aspects. As this is an emerging area, it still contains issues related to the design of pre-training tasks and the effectiveness of existing GNN models on learning structural information or functionality.
  • Complex Graph Structures – Graph structures can be flexible and complex in real-world applications. Various works have been proposed to deal with complex graph structures such as dynamic or heterogeneous graphs.

GNNs may be a very important area of ​​machine learning in the years to come. If the issues listed above can be fixed, they can also be deployed to fix big issues.

Subscribe to our newsletter

Receive the latest updates and relevant offers by sharing your email.


Join our Telegram Group. Be part of an engaging community

Sreejani Bhattacharyya

Sreejani Bhattacharyya is a journalist with a postgraduate degree in economics. When not writing, she finds herself reading about geopolitics, economics and philosophy. She can be reached at [email protected]


Source link

Previous Can Quantum Computing Reorient Our Supply Chain Problems
Next Samsung Galaxy Book Pro 360 5G: Design, technology. inspiration

No Comment

Leave a reply

Your email address will not be published. Required fields are marked *