# DeepWalk: Online Learning of Social Representations

**Authors:** Bryan Perozzi, Rami Al-Rfou', Steven Skiena  
**arXiv:** [1403.6652](https://arxiv.org/abs/1403.6652)  
**Date:** 2014-03-26  
**Summary page:** https://www.summarizepaper.com/en/paper/1403.6652v1/

## Key points

- DeepWalk is a novel approach for learning latent representations of vertices in a network.
- It encodes social relations in a continuous vector space, making it easily exploitable by statistical models.
- DeepWalk treats truncated random walks as the equivalent of sentences to learn latent representations that capture the underlying structure and relationships within the network.
- DeepWalk outperformed challenging baselines in multi-label network classification tasks on social networks like BlogCatalog, Flickr, and YouTube.
- DeepWalk's representations achieved higher $F_1$ scores than competing methods when labeled data was sparse.
- DeepWalk demonstrated scalability and efficiency as an online learning algorithm that could be easily parallelized.
- DeepWalk is suitable for various real-world applications including network classification and anomaly detection.

## Summary

DeepWalk is a novel approach for learning latent representations of vertices in a network. It encodes social relations in a continuous vector space, making it easily exploitable by statistical models. The method builds on recent advancements in language modeling and unsupervised feature learning, extending them from sequences of words to graphs. The key idea behind DeepWalk is to treat truncated random walks as the equivalent of sentences. By using local information obtained from these walks, DeepWalk learns latent representations that capture the underlying structure and relationships within the network. To evaluate the effectiveness of DeepWalk's latent representations, several multi-label network classification tasks were performed on social networks such as BlogCatalog, Flickr, and YouTube. The results showed that DeepWalk outperformed challenging baselines that had access to a global view of the network, particularly when dealing with missing information. In fact, DeepWalk's representations achieved $F_1$ scores up to 10% higher than competing methods when labeled data was sparse. Furthermore, DeepWalk demonstrated its scalability and efficiency as an online learning algorithm. It produced useful incremental results and could be easily parallelized. These qualities make it suitable for various real-world applications including network classification and anomaly detection. In conclusion,  offers a powerful approach for learning  in . Its ability to encode social relations in a continuous vector space allows for effective exploitation by statistical models. With superior performance compared to baselines and scalability advantages, has significant potential for various applications requiring accurate analysis of network data.